Download as pdf
Download as pdf
You are on page 1of 253
Universidade do Porto Faculdade de Engenharia Experimental and Numerical Analysis of the Behaviour of Structural Concrete under Fatigue Loading with Applications to Concrete Pavements Andlise Experimental e Numérica do Comportamento do Betaéo Estrutural sob Accées de Fadiga com Aplicacao a Pavimentos Rigidos Dissertagdo apresentada para obtengiio do Grau de Doutor em Engenharia Civil pela Faculdade de Engenharia da Universidade do Porto Thesis submitted to the degree of Doctor in Civil Engineering by the Faculty of Engineering of the University of Porto Paulo Barreto Cachim Dezembro de 1999 Para a Teresa, a Maria e o Afonso Quando o vento sopra alguns erguem muros para se proteger do vento. ‘outros constréem moinhos para moer 0 grao. ACKNOWLEDGEMENTS I would like to thank the facilities provided by the Laboratory of Structures of the Faculty of Engineering of the University of Porto where this research program has been carried out. I would like to thank the financial support given by the JNICT during the year of 1993 and by the PRODEP (University of Minho) from October of 1994 to October of 1997. The facilities provided by the University of Aveiro are gratefully acknowledged. This work was performed under the direct guidance of Prof. Joaquim Figueiras and Prof. Paulo Pereira. To Prof. Joaquim Figueiras I would like to express my gratitude for all the support and encouragement all over these years. I would also like to express my thanks to Prof. Paulo Pereira for the help granted to this work. Iam grateful to Domingos Ribas, Joaquim Barros and to colleagues from University of Aveiro, University of Porto and University of Minho for their support and interesting discussions. To my friends and family my thanks for their support. Finally, my very special thanks to my wife and my sons for their love, support and patience. ABSTRACT This work represents a contribution for the understanding of the concrete behaviour under repetitive and fatigue loading. The main parameters that influence the fatigue behaviour of structural concrete are identified through a selected set of experimental tests carried out over plain and steel fibre-reinforced concrete samples. It is the purpose of this work to develop a numerical model that simulates the behaviour of concrete structures under fatigue loading conditions. The model is applied to the analysis of concrete road pavements, thus, a state of the art on concrete pavements was carried out to understand the different questions that subsist around this subject. The research was devoted to the evaluation of the fatigue behaviour of plain and fibre-reinforced concrete. After the study of some of the research endeavours carried out through this subject by other researchers, an experimental program was carried out. The objectives of these experiments were twofold: first to get insight and to compare the behaviour of these two distinct materials under fatigue loading and, secondly, to collect controlled data for the calibration of a numerical model. Two types of fatigue tests were carried out, namely compressive fatigue tests on cylinders and flexural fatigue tests on prisms, both on plain and fibre-reinforced concrete. For the numerical modelling, a code was developed to analyse concrete under monotonic, cyclic and fatigue loading conditions, based on the theory of plasticity. The mode! assumes independent compression and tension behaviours. The monotonic model serves as the envelope for the cyclic and fatigue load models. For the cyclic model, two additional surfaces rule the behaviour of the mode! during the unloading and reloading process. The monotonic envelope bounds the movement of these surfaces. The fatigue model was based on visco-plasticity by assuming a fatigue limit outside which fatigue deformations occur. A model for the study of concrete pavements was developed to account for devices such as dowel bars, continuous reinforcement or loss of contact between pavement layers, and applied to in the analysis of concrete pavements. RESUMO O presente trabalho representa um contribuigdo para a compreensio do comportamento do betdio sob acgdes repetidas e fadiga. Os fenémenos ¢ os pardmetros principais que condicionam a resisténcia & fadiga do betdo estrutural sto identificados através de um Conjunto seleccionado de ensaios experimentais realizados sobre provetes de betio simples ¢ de betdo reforgado com fibras de ago. E objectivo do trabalho, o desenvolvimento de um modelo numérico que simule o comportamento de estruturas de betdo sob acgSes que induzam fadiga. Pretende-se aplicar 0 modelo a anélise do comportamento de pavimentos rodoviétios de betdo. Para a concretizagdo deste iiltimo objectivo, efectuou-se uma pesquisa sobre o estado da arte dos pavimentos de betio de forma a compreender as diferentes questdes relacionadas com este assunto. A investigagdo foi dedicada a avaliagdo do comportamento do betio simples e reforgado com fibras de ago. Apés 0 estudo de alguns dos trabalhos de investigacdo realizados por outros investigadores neste dominio, foi realizado um programa experimental. Os objectivos dos ensaios experimentais eram a aquisigo de conhecimentos e a comparagdo do comportamento dos dois tipos de materiais e, em também a recolha de dados que permitissem a calibragao do modelo numérico. Realizaram-se ensaios de fadiga em compressio e ensaios de fadiga em flexio, ambos para betdo simples e reforgado dom fibras. No que respeita a modelagdo numérica, desenvolveu-se um cédigo computacional para a anilise do betdo sob condigSes de carga monoténicas, ciclicas ¢ de fadiga, baseado na teoria da plasticidade. O modelo assume comportamentos & tracg4o e compressio completamente independentes. O modelo monoténico serve de envolvente para as acgdes ciclicas e de fadiga. Para as acgdes ciclicas, o endurecimento de duas superficies adicionais que se movem dentro da envolvente monoténica controlam o comportamento em descarga ¢ recarga. O modelo de fadiga é baseado na teoria da viscoplasticidade, assumindo-se que existe um limite para o qual no existem deformagdes por fadiga. O modelo desenvolvido aplicou-se ao estudo de pavimentos de betio. RESUME Ce travail représent une contribution a la comprehension du comportement de bétonde ciment sous des actions répetées de fatigue. Les phenomenes et les parimétres principaux qui conditionent la résistance A la fatigue du béton structure! sont identifiés atravérs d'un ensemble d’éssais réalisés sur des éprouvettes de béton simple e de béton renforcé avec des fibres de acier. L’objectif du travail est le dévelopment d’un médele numérique qui simule le comportement de structures de béton soumises A des actions qui aménent & la fatigue. Le modéle dévelopé sera apliqué a l'analyse du comportement des chaussées de béton de ciment. Pour cela, on a réalisé une recherche bibliographique sur I’état de art des chaussées en béton de ciment, de fagon A comprendre les diférentes questions que ce posent sur ce type de chaussée. La recherche a concerée I’évaluation du comportement du béton simple e celui du béton renforeé avec des fibres d’acier. Aprés l'étude des travaux de recherche de plusieurs chercheurs, il a été dévelopé un programme experimental déssais. IEs objectifs des essais de laboratoire ont été I’dcquisition de connaissances et la comparaison du comportement des deux types de matériaux et aussi le recueil de données pour permettre I’étallonage du modéle numérique. Pour les deux matériaux, ont été réalisés des essais de fatigue, en compression et en flexion. En ce qui concerne la modélisation numérique, il a été dévelopé un code numérique pour I’énalyse du comportement du béton sous des diverses conditions de charge, monotoniques, cycliques et de fatigue, basées dans la théorie de la plasticité. Le modéle dévelopé assume des comportements a traction et a la compression completement indépendents. Le modéle monotonic est utilisé comme enveloppe aux actions cycliques et de fatigue. Pour les actions cycliques, léndurance de deux surfaces aditionnelles qui se mouvent au sein de Iénveloppe monotonic, controlent le comportement en dechargement et rechargement. Le modéle de fatigue est basé dans Ia theorie de la viscoelasticité, tenant compte qu'il y a un limite pour lequel il n’y a pas des deformations de fatigue. Finalement, le modéle dévelopé a été apliqué a l'étude du comportement de chaussés de béton de ciment utilisés dans le réseau routier portugais. TABLE OF CONTENTS 1 INTRODUCTION. 1.1 Pavements throughout history. 1.2 Concrete pavements in Portugal 13 Objectives... 14 Scope of this study 2 UNDERSTANDING CONCRETE PAVEMENTS. 2.1 Types of concrete pavements.. 2.1.1" Jointed unreinforced concrete pavements (JUCP) 2.1.2 Jointed reinforced concrete pavements (JRCP) 2.1.3 Continuously reinforced concrete pavements (CRCP) 2.1.4 Roller-compacted concrete pavements (RCCP).. 2.15. Other types of concrete pavements.. 2.2 Damage mechanisms of concrete pavements 2.2.1 Traffic induced stresses 2.2.2 Temperature induced stresses . 2.2.3 Fatigue of materials 2.2.4 — Soil behaviour, water infiltration and joint efficiency 2.3 Concrete pavement analysis. 23.1 Analytical models. 2.3.2 Finite element model: 2.4 Slab thickness design for concrete pavement 2.4.1 Rational models 2.4.2 Empirical model: 2.5 Summary. 3__ FATIGUE OF PLAIN AND FIBRE-REINFORCED CONCRETE. 3.1 FIBRE-REINFORCED CONCRETE. 3.1.1 Tensile behaviour. 3.1.2 Bending behaviour. 3.1.3. Compressive behaviour. 3.2 FATIGUE BASICs.. 3.2.1 Fundamental definitions 3.22 ing fati 3.2.3 Fatigue testing 3.2.4 Modelling fatigue dat 3.2.5 Deformations 3.2.6 Damage accumulation. 3.3. FATIGUE OF CONCRETE 3.3.1 3.3.2 3.4 DESIGN OF CONCRETE STRUCTURES TO FATIGUE LOADS 3 Load spectrum. 3. Material characteristics. 3.5 SUMMARY... 4 COMPRESSIVE FATIGUE TESTS. 4.1 Materials and equipment 4.11 4.1.2 4.13 4.2 Test procedure... 4.2.1 4.2.2 4.2.3 4.3 Results of monotonic tests. 43.1 43.2 4.4 Results of 60-mm length fibres fatigue tests 44.1 4.42 443 44.4 44.5 4.5 — Results of 30-mm sel fibres a tests, 4 454 455 4.6 — Discussion 4.7 Summary . 5. FLEXURAL FATIGUE TESTS.. 5.1 Test procedure 5.11 5.1.2 5:13. 5.14 Fibres Concrete Equipment Specimens. Monotonic tests Fatigue tests. Stress-strain diagrams of monotonic tests. Relation between tangent modulus and compressive strength. Stress level. Deformations. Fatigue modulus Failure patterns Post-fatigue properties. Fatigue modulus.. Failure patterns. Post-fatigue properties... Specimens. Data acquisition devices. Monotonic tests Fatigue tests.. 5.2. Results of $25x150x150 prisms... 5.2.1 5.2.2 5.2.3 5.2.4 s25 5.2.6 5.2.7 5.3. Results of 525x150x75 prisms... 5.3.1 5.3.2 5.3.5 5.4 Discussion 5.5 Summary... Monotonic results. Stress leve Deformations... Fatigue modulus. Post-fatigue properties. Crack patterns.. Compressive tests on cubes sawed from the 525x150x150 prisms . Monotonic results. Stress level Deformations. Crack patterns. Use of sinusoidal curves for load-displacement modelling 6 NUMERICAL MODEL FOR CONCRETE. 6.1 Plasticity.. 6.1.1 Basic assumptions. : 6.1.2 Integration of the elastic-plastic constitutive me 6.1.3 The consistent tangent stiffness matrix... 6.1.4 6.1.5 Plasticity enhanced models... Visco-plasticity . 6.2 Constitutive relations for monotonic loads. 6.2.1 Application of the elastic-plastic models for concret: 6.2.2 — Concrete in compression... 6.23 Concrete in tension 6.24 — Comer regime 6.2.5 Numerical results. 6.3 Constitutive relations for cyclic loads .. 6.3.1 Model essentials. 63.2 Unloading to compression 6.3.3 Unloading to tension. 6.3.4 Numerical examples . 6.4 Constitutive relations for fatigue loads... 6.4.1 Model essentials. 64.2 Conerete in compressior 6.4.3 Concrete in tension... 6.4.4 — Some concluding remarks on the fatigue model 6.4.5 Numerical examples . 6.5 Summary....... 7” CONCRETE PAVEMENT ANALYSIS. 7.1. Finite element model TAL Concrete slat 7.1.2 — Reinforcement. 7.1.3 Sub-base modelling 7.1.4 — Soil modelling. 7.1.5 Interface and Joint behaviour. 7.2 Temperature loads... 7.3 Conerete pavement analysis. 74 Summary. 8 ConcLusions. REFERENCES .... Introduction IL 1 INTRODUCTION Concrete is probably the most widely used construction material in the present. It is used, for instance, in buildings, bridges, roads, offshore platforms and nuclear installations. Hence, it is quite understandable that the modelling of the concrete constitutive relations has been the subject of intensive research in the last twenty years. These research endeavours have resulted in an impressive increase of the scientific level of material modelling and of the efficiency of software for non-linear finite element analysis of concrete structures. However, the impact of these research activities on engineering practice is not so impressive. In fact, from the standpoint of engineering practice, the motivation for replacing the conventional methods of analyses by mechanically more sophisticated and considerably more complicated analysis methods will remain small as long as the expected increase in quality of the design is not proportional to the increase in mechanical sophistication and computational complexity. Nevertheless, non-linear finite element analysis of concrete structures has increased and, at present, it is mainly used as a tool for the assessment of the structural safety of non-standard concrete structures (Figueiras et al., 1994). Pavements are complex structures where, besides material properties, a number of other variables such as traffic, climatic conditions or pavement geometry play an important role on the structural performance. Moreover, traffic or climatic conditions are random by nature with large fluctuations in time, which may cause material fatigue. Additionally, the pavement typology strongly influences the boundary conditions, which causes radically different structural behaviours. Thereafter, only an accurate perception of all these interconnected factors will allow a correct modelling of pavement structures. During this century, enormous developments occurred in the field of road transport, where the number of vehicles and the magnitudes of the loads clearly grow beyond expectation. Thus, the fast increasing loads, the needs to recycle wasted materials and the limitations on the budget available for road construction and maintenance, are some of the reasons that lead to an intensification of pavement research. This research should result in improved material models and mathematical tools that allow a better prediction of the pavement performance than is currently the case. Simultaneously, it is also necessary to look for construction materials with enhanced properties such as strength, ductility, toughness and durability, which caused an increasing interest on materials like fibre 2 Chapter 1 reinforced concrete or high performance concrete. The little knowledge about the long- term behaviour of these materials namely the effect of repeated loading on their properties has caused an increasing interest on the fatigue performance of concrete. Additionally, the data acquired with the experiments should provide information about the main parameters that must be included in numerical simulations. 1.1 Pavements throughout history From the dawn of mankind that roads are recognised as a fundamental development factor allowing fast commercial trades and army movements. In fact, it can truly be said that the prosperity of a nation is bound up with the state of its roads, and that the roads act as palimpsest of a nation’s history (O'Flaherty, 1979). Ancient roads The invention of the wheel more than 5000 years ago (the Mesopotanians have used four-wheeled wagons as early as 3000 B.C.) made necessary the construction of special hard surfacings capable of carrying concentrated and greater loads than before. Some stone-paved roads have been built in the Antiquity. The Egyptian Pharaoh Cheops during the building of the Great Pyramid builds a causeway to facilitate the transportation of huge limestone blocks. By 2000 B.C., a dual carriageway was constructed in Crete connecting Cortina in the south to Knossos in the north Also in Babylon by 620 B.C. a processional route was built having 10 to 20 metres width and about 1220 metres long. However, the Romans were undoubtedly the major road builders of the ancient times. They have used remarkable building techniques that for centuries last as the most advanced. The Romans construct more than 75000 km of paved roads radiating from their centre at the miliarius aurem, in the Forum in the city of Rome. The Roman roads could have a total thickness of more than one metre and a structure conceptually similar to that of the moder roads. The Romans were the first civilisation to build paved roads that did not prevent travel during or after bad weather conditions and that did not raise clouds of dust during dry weather. Roman engineers, however, did not stop with just paving roads. In fact, Roman roads were higher in the middle then on the sides (see Figure 1) and often had gutters for drainage along the shoulders. The most incredible engineering feature Introduction 3 concerning the Roman road system, though, is the quality of the construction, that make them last in very good condition until nowadays. Figure 1. Actual view of the Via Appia built in 312 B.C. Ancient Roman roads consist of four layers placed over the pavimentum that is the formation level of the road (Figure 2). To build the pavimentum two parallel trenches were excavated along the edges of the road and all the top soil of loose material between them was removed until firm ground was found. The excavated cut was then filled with dry sand that is well-forced into position. [Summum dorsum] ‘Stone blocks Nucleus Gravel or rokan pottery cemented by ime Smal stones o bbe sein mortar ‘Staturen Flat stones or smal square stones Pavimentum Figure 2. Cross-section of roman road. 4 Chapter 1 Over the pavimentum, a layer called the statumen was built by setting two or three courses of large flat stones or small square stones into lime mortar. The next layer, the rudus, consists of small stones or rubble set in mortar. The nucleus usually consisted of fine concrete composed of locally available materials such as gravel, broken pottery or bricks, cemented together by lime to form a hard permanent mass of material. The final layer, called the summum dorsum, consisted of stone blocks embedded in the top of the newly laid nucleus. This type of construction is subjected to several variations because the materials used to build the roads are strongly dependent on what exists in the road’s immediate vicinity Comparing the Roman road structure to modem roads it can be seen that the statumen and the rudus may be considered as the sub-base of the road, the nucleus is the base and the wearing coarse is the summum dorsum (Jeuffroy, 1978). Since the first century B.C., the Romans used hydraulic or pozzolanie binders in pavement structures consisting of “two parts natural pozzolan and one part lime” to tie the stones of their pavements. Thus, the slabs of certain Roman roads may be considered the precursors of concrete roads. Modern Roads After the fall of the Roman Empire, due to the non-existence of great empires that need the massive transportation of troops and supplies, the need for the construction of thick, rigid and long-lasting roads almost disappears. Only almost 2000 years later, with the advent of the industrial revolution, and mainly with the appearance of the automobile, arouse again the necessity of developing the road conception and the construction techniques. By the end of the 18 century and beginning of 19" century, arise some methods of road construction and maintenance due to Trésaguet in 1775, Telford in 1803 and MacAdam in 1820. However, these road cross-sections are quite lighter than Roman roads, based on the principle that the subsoil should be able to support the load laid upon it. These light road building processes emphasise drainage, crushed stone subgrades and covers of finely crushed stone bound with water or oil, giving the familiar macadam type of design (Hay, 1979). The road cross-sections proposed by Trésaguet, Telford and MacAdam are presented in Figure 3. In the beginning of the 20" century, concrete started to be used in road construction. The first applications of concrete pavements have been made in Belgium, France, Germany Introduction 5 and United States. Concrete is usually placed in two layers, one after the other with no time Jag to ensure bond between them. The top layer was thinner and has higher cement content. The total thickness was 18 cm and transversal joints are provided every 10 to 20 metres. Convex surface eS eae 76 mm smal stones os eee 1 fsize of a walnut) 3% Oo 152-179 mm larger stones [7] 152-179 mm foundation stones Convex surface Level formation 1 238mm gravel 51 mm broken stones sa rarae] 4-101 mm (63 mm max size) 78 mm 1179 mm stones (103 mm wide) Broken stones ‘ ‘and gravel or | 81mm broken stones (25 mm max size) 2 j 4101 mm broken stones $2241 1 101 mm (76 mm max size) Figure 3. Road cross-sections due to Trésaguet (top), Telford (middle) and MacAdam (bottom) according to O°flaherty (1979). Then came World War I, and some time elapsed before cement has been systematically used in road construction. From 1933 onwards, industrialisation blossomed. Germany was the pioneer to organise big highway projects, as a means of mitigating unemployment and, like the Romans, to enable quick and safe movement of military units. The pavement thickness augment to 24 cm with 5 cm of wearing course. The transversal joints were about six metres apart. The final step in the evolution of concrete pavements was the suppression of joints what was achieved by the placement of a continuous reinforcement throughout the pavement length. This steel reinforcement has no structural functions and is disposed just to control shrinkage and thermal cracking, 1.2. Concrete pavements in Portugal The beginning of the utilisation of cement on Portuguese roads starts about 1938 and last until 1945. In 1953, a communication of Eng.° Alberto Ziiquete from the “Ministério das Obras Priblicas” (Public Ministry) presented the few cement applications on road 6 Chapter 1 construction in Portugal (Quaresma, 1992), Between cement macadam and concrete slabs more then 70 km of roads were build what was undoubtedly an important effort for those times. Some of those applications are summarised in Table 1. By those times, in spite of the advantages of the use of cement, the cost of cement was the main reason for not using it as a road material Table 1. Concrete pavements in Portugal from 1938 to 1945. Identification on. Date coc, Estadio Nacional highway 38 «1940 Slab with 22 em, 2 unreinforced layers EN8 ~ Loures/P.Loures 4 Slab with 20 cm, 2 unreinforced layers EN16 ~ Mangualde 35 Slab with 2 layers ENI 4 ~Peniche/Atouguia 5 1944 Slab with 14 em, 1 unreinforced layer ENI18 ~ Salvaterra/Benfica Ribatejo Slab with 2 layers ENI18 —Paul Magos Slab with 1 layer EM347 ~ Montemor-o-Velho/Alfarelos Slab with 1 unreinforced layer m Slab with 22 em, 2 layers (15+7) EN356 - Martinganca/Est. Martinganga 1.5, 1941 reinforced with 6 3/16" (1.5 kg/m?) ENIII-1 ~ Cidreira/Est. Velha (Coimbra) Slab with 1 unreinforced layer Only 40 years later in 1985, with the approval of the, “Novo Plano Rodoviario Nacional” (PRNBS) it was decided to use again the cement on concrete road construction. Meanwhile, only in 1962 a small experimental prestressed concrete road was built in EN 1 near Azdia, with relatively bad results. The decision to use again cement on road construction when the bituminous costs were declining are briefly explained, The first reason was the PRN85 itself, since more then 4000 km of roads were to be build or rehabilitated, the National Road Authority (JAE) decided that there was a field for application of internationally established road pavements (Botto, 1990). Second, the very fast increase of heavy loads and the increase of the single axle load from 100 KN to 120 kN lead to thicker bituminous solutions with costs compared with equivalent concrete solutions. Finally, the need to reduce the maintenance costs was also a very important motive fo use concrete pavements. The applications of cement on road construction include the use of cement in the treatment of soils, semi-rigid pavements and concrete pavements. In Figure 4, the portuguese map with the location of the currently existing concrete roads is presented as well as an enlarged map of Lisbon zone where more concrete roads have been built, Two types of concrete pavements are currently constructed in Portugal: jointed reinforced concrete pavements (JRCP) and continuously reinforced concrete pavements Introduction 7 (CRCP). The former consists of concrete slabs with five metres long with dowel bars at the transversal joints. As joints are always weak points of any structure, there is always the wish to eliminate them. Figure 4. Concrete roads in Portugal. The alternative to IRCP is the continuously reinforced pavement, which instead of joints have a continuous reinforcement all over the slab length, located at half of the slab thickness. The characterisation of the concrete pavements existing in Portugal is presented in Table 2. An interesting contradiction of concrete pavements is the fact that they are suitable for very high traffic situations but also for low volume roads such as rural roads. The reasons for this are the resistance to adverse climatic conditions, impossibility of rut formation and reduced maintenance. Some applications of concrete in low-traffic roads in Portugal have also been tried with much success (Proenga, 1998). 8 Chapter 1 Table 2. Concrete pavements in Portugal from 1988 to 1999. Lenten EER ae Tye Papeat sce IC2- Asseisseira/Alto da Serra 131988 RCP CC*LC+GA (22+15+15) 1P3 - RaivaTrouxemil 22 1991 SRCP CC +LC+GA (24+15+15) IPAIAG — Paredes/Penafiel 101991 FRCP CC+LC+GA(23+15+15) ICI/A8 - Loures/Malveira 13 1991 CRCPCC+LE+GA Q0+15+18-22) IPU/Via do Infante —Guia/S. Joto Venda 32-1993. CRCP_ CC + BC + GAC (22+4+15) ICIT/CRIL - Algés/Buraca 6 1995 CREP CC +LC+ GA Q0+15+15) ICIBICREL - Alverca/E. Nacional 301995 crop EE BE _ chetierg ICWV/A8 ~ Malveira/Torres 1996 _JRCP CC: cement concrete; BC: bituminous concrete; LC: lean concrete; SC: soll-cement; GA: graded aggregate; GAC: graded aggregate treated with cement. 1.3 Objectives The main goal of this research is to contribute for the understanding of the concrete behaviour under repetitive and fatigue loading. In this study, the behaviour of plain concrete was compared with that of steel fibre-reinforced concrete to investigate the effect of the improved ductility of fibre concrete on the fatigue behaviour of concrete. The main parameters that affect the fatigue performance of concrete were identified by a selected set of experimental tests carried out over concrete specimens. The other purpose of this research program is the development of a numerical model capable of simulating the behaviour of concrete under fatigue loading. This material model was implemented in a finite element code to analyse the behaviour of concrete road pavements, Hence, the finite element code had to be modified accordingly so that the basic features of concrete pavements such as dowel bars, continuous reinforcement, temperature effects or Joss of contact between the slab and the foundation could be accurately modelled. Although their apparent differences, the two aforementioned objectives are closely related to each other, since the knowledge acquired with the experimental tests will provide valuable information for the calibration and development of the numerical model. On the other side, the applications of the numerical model may contribute to a better understanding of the fatigue phenomenon. Introduction 9 1.4 Scope of this study This work starts with a general characterisation of concrete pavements in Chapter 2. The different types of concrete pavements are described. Their damage mechanisms, namely traffic and temperature loads, fatigue, soil behaviour, water infiltration and joint efficiency are briefly reviewed. Analysis methods with special emphasis to the finite element method and some design procedures are presented. Fatigue is a major cause of failure of concrete pavements being of major importance a correct understanding of the fatigue mechanisms. In Chapter 3, a state-of-the-art of the fatigue behaviour of concrete is presented where the performance of steel fibre-reinforced conerete is compared with that of plain concrete. The experimental work carried out to assess the fatigue behaviour of concrete is described in Chapter 4 and Chapter 5. Two types of fatigue tests were performed: compressive fatigue tests on cylinders and bending fatigue tests on prisms. The objectives of the experimental program are, first, the understanding and comparison of the performance of plain and fibre-reinforced concrete under fatigue loading and, secondly, to provide auxiliary data to calibrate the numerical model. Chapter 6 describes the developed numerical model for the analysis of concrete under repetitive loading. The plasticity theory is used as the basis for the numerical model. However, more advanced concepts such as bounding surface plasticity, generalised plasticity and visco-plasticity will have to be introduced for the repetitive load models. First, a description of the monotonic model is shown, which serves as envelope for the repetitive load models. Then, the model is developed to allow cyclic load/unload behaviour and, finally, a fatigue model for concrete is developed based on visco-plasticity theory. The assessment of the numerical model capabilities is presented in this chapter by comparing some experimental results with the model response. Some examples that allow the understanding of the model behaviour are also presented. Some applications of the numerical model to concrete pavement analysis are presented in Chapter 7. Additionally, the necessary finite element formulation and material models that were developed such as reinforcing bars, foundation, or interface behaviour, were also descried in this chapter. lo Chapter 2 2 UNDERSTANDING CONCRETE PAVEMENTS Pavement is the part of a road, street or airfield that directly supports the traffic induced loads and transmit them to the infra-structure (natural soil or earthworks). A road pavement is usually made of several layers. A wearing course ensures the absorption of tangential stresses induced by traffic while the foundation layers (usually the road base and the sub-base) do the degradation of the vertical loads. The road base guarantees the principal structural capacity while the sub-base has mainly the aptitude to avoid some inconvenients that are specific to the foundation soils. Historically, the pavements have been divided into flexible and rigid depending on the way they degrade loads to the foundation soil. In flexible pavements exists a gradual stiffness increasing from the foundation soil to the wearing course, which creates high stresses on the soil because loads are degraded over a relatively small area. On the contrary, in rigid pavements the stresses induced on the soil are smaller because the road base stiffness is much bigger than that of the soil. Thus, in rigid pavements the road base is subjected to significant tensile-bending stresses. Rigid pavements are usually made of a concrete slab that ensures the wearing course and road base functions, and by a sub-base that guarantees adequate support for the slab and for the slab construction. The sub-base is usually made of lean concrete or of cement stabilised soil. Thereafter, the behaviour of the concrete pavements is essentially dependent, but of course not entirely, on the behaviour of the concrete slab. The understanding of the concrete behaviour is thus of an extreme importance for a correct conception and analysis of concrete road pavements. Concrete pavements are usually used for very high traffic situations. This means that any maintenance works involve high cost for road users. Thus, the lifetime for concrete pavements usually ranges from 30 to 40 years that are intended to last with almost zero maintenance. Nowadays, thousands of load repetitions occur daily in the busiest highways, and these numbers tend to grow at very fast rates. Thus, fatigue plays a very important role in the durability of pavements. Fatigue is a process of progressive damage in the internal structure of the material due to load repetitions. Consequently, fatigue failure occurs not by Understanding Concrete Pavements u the occurrence of an isolated large load but by the repeated occurrence of loads that isolated are unable to cause failure. A very important feature of the tensile behaviour of concrete is its brittle behaviour after reaching the peak load. This is a major drawback in the performance of concrete pavements and some actions have to be adopted to control the crack formation and evolution, such as joints or continuous reinforcement. In addition, fibres may be added to concrete in order to control the crack opening process. In fact, the fibres that cross an opening crack may delay its aperture, and stress redistribution between sections of the slab may occur leading to an overall improved performance. The fast increasing of loads, the needs to use recycled materials and the limitations set on the budget available to build and maintain road infrastructure, are amongst others the main reasons to intensify pavement research. This research should result in improved material models and mathematical tools which allow pavement response and performance to be predicted more accurately than is currently the case (Molenaar, 1996). In this chapter, concrete pavements are summarily described so that a clearly understanding of their overall behaviour could be attained. This description starts with a characterisation of the different concrete pavements. After that, the distinct damage mechanisms such as fatigue, thermal gradients or pumping are briefly analysed. Finally, the methods currently available for the analysis and design of concrete pavements are outlined. 2.1. Types of concrete pavements ‘The main reasons for the adoption of concrete pavement solutions are its large durability and reduced maintenance needs, besides a considerable strength. Thus, concrete pavements are usually used in high traffic situations where the maintenance costs are very expensive for road users. Nevertheless, various other applications of concrete pavements are possible such as urban roads, low-traffic roads, ports, airports or industrial pavements (Cachim er al. 1995). The various types of concrete pavements attempt to take advantage of the intrinsic qualities of concrete and to reduce the problems created by its limited ductility and cement hydration shrinkage. Pavement concrete slabs could be short (3.5 to 5 m), long (12 to 18 m) or continuous. Its thickness ranges from 20 to 25 cm accordingly to foundation type, or 12 Chapter 2 traffic intensity (JAE, 1995; LCPC, 1993). Figure 5 presents some current cross-sections of some concrete pavement types (LCPC, 1993). Juce sucp RCP crop rep (Cattornian) (Thiek stab) 22.CRC pa 260C 35cc 21cc 2006 s0sc 1510 ec ILC ILC CC ~ cement concrete; BC ~ bituminous concrete; LC — lean concrete; DL ~ drained layer; SC soil-cement; Figure 5. Current cross-sections of concrete pavements (thickness in centimetres). The building costs of concrete pavements increase with the technology needed to build the pavements. Inversely, the maintenance costs decrease with the technology involved. However, the technologically advanced pavements have higher maintenance costs if errors occur during the building process. The most used concrete pavement solutions are presented such as jointed unreinforced/reinforced concrete pavements, continuously reinforced concrete pavements or roller-compacted concrete pavements. Nevertheless, several other less-common concrete pavement solutions are also described. Some of these solutions are very promising proposals that in a near future may be the standard ones. 2.1.1 Jointed unreinforced concrete pavements (JUCP) These pavements are made by short plain concrete slabs with contraction (and expansion) joints spaced around 3.5 to 5.5 metres. In some special zones, such as transition zones for additional climbing lanes, the slabs are reinforced with steel bars. Spacing of transverse joints is sometimes not uniform to avoid the resonance phenomenon. These pavements show the smallest building costs and easiness of construction. The joint efficiency of JUCP may be attained with a sub-base associated with the slab (californian pavement) or by a thickness increasing and a joint spacing reduction that reduces the crack opening (thick slab pavement). Concrete is continuously placed with a slipform paver (Figure 6) and transverse joints can be made through the insertion of forming strips into the fresh concrete or can be Understanding Concrete Pavements 13 made later by sawing concrete. In the latter case, this operation must be made before the appearance of the first shrinkage cracks. Anyway, the joint groove depth is about 25% of the slab thickness. aou oun os 1 Discharged concrete 410 Transfer hopper for top, 19 Front screed 2 Receiving hopper for layer concrete 20 Spreading auger top layer concrete 11 Super-smoother 21 Adjustable front screed 3 Levelling eylinaer 12 Stringline aigsariionifanter top 4 Swing eg 19 Sensors for levelling and layer concrete 5 Operator's platform steering, front 23 Oscillating transverse 8 Main Seana! 14 Crawler track assembly finishing beam 7 fewer 15 ‘Spreader plough 24 Sensors fer leveling and 8 Dowel bar inserter DBI 16 Front metering screed en 3 Bek cere 17 HF vibrators 25 Finished bottom yer ot layer concrete 18 Extrusion pan for bottom: sm ayer Oona 26 Finished top layer of concrete Figure 6. Schematic illustration of a slipform-paving machine. Longitudinal hinged joints are necessary between the traffic lanes to avoid longitudinal cracking. These hinges are made with steel bars spaced around 60 cm and with 12-mm of diameter. Contrary to the transverse joints where movements of the slab edges should be allowed, in the longitudinal joints the movements between the two faces of the joint should be eliminated. Thus, a perfect bonding must exist between the bars and concrete 2.1.2. Jointed reinforced concrete pavements (JRCP) its are doweled. The These pavements are similar to JUCP except that the transverse j introduction of dowel bars provides a much better joint efficiency and a much longer lifetime for the joint. The JRCP are more expensive than the JUCP, but they need less maintenance. rr Chapter 2 ‘The dowel bars should be able to transmit the shear stresses through the joint but not the axial stresses. For this reason, dowels are covered with bituminous products or placed into plastic sleeves to avoid bonding between steel and concrete. The bar diameter ranges from 25 to 32 mm and its length ranges from 60 to 80 cm. They must be placed parallel to road axis and are usually 30-cm spaced. There are two processes of placing dowels into joints: by insertion in the wet concrete (see Figure 7) or by placing them on prefabricated cages. The former alternative is cheaper and faster because dowels are inserted during the placement of concrete, however there are some concerns about the ability of this method to correctly position the dowel bars. Figure 7. _ Dowel bar insertion machine. 2.1.3 Continuously reinforced concrete pavements (CRCP) The non-existence of joints is the main characteristic of the CRCP. Cracking is controlled by the placement of a continuous steel reinforcement along the pavement (see Figure 8 and Figure 9). In this way, it is possible to keep cracks shortly spaced (approximately one metre) with small aperture (around 0.3 mm). Steel percentages ranging from 0.6% to 0.7% of the concrete sectional area are needed to keep cracks under control. Usually 16 or 20- mm corrugated steel! bars are used. The transversal reinforcement, when present, is 0.1% and is made of 12 to 16-mm bars, spaced from 40 to 100 cm (PIARC, 1994). Heavy and high volume traffic areas are particularly indicated for the CRCP solution, Building a CRCP is very expensive but its maintenance costs are very small (if the Understanding Concrete Pavements 3 pavement has been well constructed otherwise the maintenance and reparation costs are important). Corrosion of reinforcement seems to be minimum (Verhoeven, 1992). Figure 8. Building of aCRCP. Two types of execution procedures may be used to place the reinforcement. The reinforcement may be placed in advance laid on supports and the concrete is placed afterwards (Figure 9) or alternatively, the bars could be placed simultaneously with concrete with the aid of trumpets. In the last case, there is no transversal reinforcement. Figure 9. Reinforcement in a CRCP. 16 Chapter 2 2.1.4 Roller-compacted concrete pavements (RCCP) Roller-compacted concrete pavements are characterised by their low water content used during the mixing procedure. This allows the compaction by a roller instead of the traditional vibration. A two-step compaction is usually employed: a vibrated rolling followed by a rubber-tyre rolling. Cement content of RCCP is similar to that of standard vibrated concrete. To avoid widely open cracks, joints are usually used in RCCP spaced around 6 metres. The surface characteristics of the RCCP are not the best, so for high-speed traffics, a bituminous surface layer must be used. When used in strengthening works, these pavements could be immediately open to traffic. Another possibility for RCCP, is the placement of a continuous reinforcement (Kameta et al, 1994) creating a continuously reinforced roller-compacted concrete pavement. With this pavement, it is possible to associate fastness and easiness of construction with good structural performance. 2.1.5 Other types of concrete pavements Several other types of concrete pavements are available such as fibre-reinforced concrete pavements, recycled concrete pavements, cellular rigid pavements or prestressed concrete pavements. A brief description of these pavements will be presented hereafter. Fibre-reinforced concrete pavements (FRCP) As stated previously, fibres provide additional ductility to concrete. Fibre-reinforced concrete pavements may exhibit thinner slabs than standard plain concrete pavements (Paillere and Bonnet, 1975; Canovas and Martinez, 1984). Overlays of fibre-reinforced concrete have been used with relative success over old asphalt and concrete pavements (Verhoeven, 1990; Ozyildirim ef al., 1997). One of the major problems pointed out to the use of steel fibre-reinforced concrete is the corrosion of fibres. Nevertheless, some studies show that stainless-steel fibres do not show any corrosion even in chloride environment or cracked specimens. However, zinc-coated fibres show corrosion phenomena in the presence of chlorides (Rosseti et al., 1994). An interesting patented proposal, called fibre-reinforced roller-compacted concrete pavements (ROLLFIBER) has been proposed by Ficheroulle (1998). Twincone steel fibres Understanding Concrete Pavements 7 are added to concrete to provide a so-called total anchorage. No joints are sawed, and the first results show that main cracking, with 0.5 to 1.0 mm wide, occurs at 8 to 10 metres intervals. Recycled concrete pavements Building demolition waste and concrete fragments are produced in increasing quantities and some applications must be found in order to use them. Their use in pavements seems a promising solution (Cristofoletti ef al., 1994; Cuttell et al., 1997; Nakanishi ef al., 1998). Nevertheless, concrete containing recycled concrete aggregates exhibits high drying shrinkage coefficient and thermal expansion values. Thus, pavement joint layout and load transfer systems must be designed accordingly. Nowadays, with the increasing volume of garbage incineration residuals, it is necessary to give them an utilisation safe and environmentally acceptable. Their use in road pavements seems encouraging and some experimental sections are currently in study (Drouadaine et al., 1997). Cellular rigid pavements is a patented construction method (see Bright and Mays, 1996), that uses waste plastic for forms to create cells. It allows a reduction of approximately 25% of concrete when compared to a standard concrete pavement. Until now, no field experiments have been carried out and thus the cellular rigid pavement can be viewed as an emerging technology. Prestressed concrete pavements The prestressed concrete technique is not very used in road pavements because they are very expensive. Their thickness ranges from 12 to 20 cm and the slab length may reach 300 metres. The main advantages of prestressed pavements are their reduced thickness, reduced number of joints and reduced crack patter. On the contrary, the main disadvantages are the difficulties in applying the prestress in curve (both vertical and horizontal) and their cost (Jeuffroy, 1996). The thermal gradients that pavements are submitted are another problem, since the prestress value may be significantly altered and the stresses existing in the pavement are completed altered. 18 Chapter 2 2.2. Damage mechanisms of concrete pavements The calculation of stresses in a structure presupposes the knowledge of loads, material constitutive relations, structure geometry and boundary conditions. Concrete pavement loads are essentially the traffic and temperature gradients. These loads have a common characteristic: they are repeated thousands or millions of times during the pavement life. Thus, fatigue plays a very important role in the pavement performance. Besides, boundary conditions may change during the lifetime, for instance due to water infiltration through Joints, altering radically the state of stress that was assumed in design. An adequate characterisation of all these factors is very complex and some simplifications have to be made. A brief description of the most important damage mechanisms is presented below. 2.2.1 Traffic induced stresses Traffic is the major cause of concrete pavement stresses. It is therefore very important to know the number and magnitude of heavy vehicles that will pass over the pavement during its service life. Lane widths have important effects on stresses caused by traffic. Initially pavements are designed for critical stresses occurring at the comer of the slabs, when truck wheels moved over the comers formed by transverse joints at outside edges of the pavement However, lanes became larger and the critical section moved from the outside comer of the transverse joint to the transverse joint edge. Thus, maximum flexural stresses occur at the bottom of slab, parallel to the joint edge The properties of concrete and of the underlying layers also considerably affect the pavement behaviour. Regardless of much more sophisticated models, some basic equations are presented at this point to show the main factors affecting concrete pavement stresses. The well-known model for concrete pavement analysis is undoubtedly the Westergaard model presented in 1926. It considers a Winkler-type foundation where the vertical pressure over a specific point of the foundation is proportional to the vertical displacement of that point. This simplification of the model reduces dramatically the complexity of the problem and allows the determination of analytical expressions for the horizontal stress «., at a concrete slab with thickness H, created by a load P acting on a circular area of radius a, as shown in Figure 10 (Ioannides ef al., 1985). Understanding Conerete Pavements 19 Sita ETF LED 2, =0316-F (ogi) +1069) o = 0349 F (stny/a)~ 0078+1534(aj!)) Z(1-(o50)") Interior loading Edge loading Comer loading Figure 10. Types of load cases for the Westergaard model. The radius of equivalent distribution of pressure, b, referred in Figure 10 reads a= {lise +H -067SH, a<\724H an a>1L724H and the radius of relative stiffness, /, is given by EL here D, =e, wi DE (2.2) ‘The other variables, dependent on material properties, are the modulus of elasticity, E., and the Poisson’s ratio, v., of concrete and the modulus of subgrade reaction ky. 2.2.2 Temperature induced stresses Very frequently, a differential temperature exists between the top and the bottom of a pavement slab. For example (see Figure 11) during the day the temperature on the top face may be considerably higher than that of the bottom face. As the top of slab expands more than the bottom, its effect in producing a slightly convex surface is resisted by the slab weight, with the consequence that the bottom face is in tension, By night, the opposite situation occurs, with the result that the top of the slab is in tension and the bottom in compression. Temperature warping has a double effect in concrete pavement stresses. First, the flexural stresses caused by temperature may be of considerable magnitude themselves. Second, the subgrade support may be destroyed in some zones of the slab, resulting in increased traffic loads compared to those that would exist if the pavement had uniform support. 20 Chapter 2 LESS SS Day light curling Night curling Figure I. Slab curling by day and night temperature gradient. The thermal curling stresses could be obtained from the Westergaard model and may be given by (Oglesby and Hicks, 1982): E.a,AT =¢, . 3) a, =C, 2 (2.3) for the edge of a slab, and by E,a,AT 21-07)" for the interior of a slab. The parameters C, and C, may be obtained from Figure 12, AT is o.=(C, +0.C,) (2.4) the temperature differential between the top and the bottom layer, a, is the thermal expansion coefficient of concrete and L, and L, are the length of the slab in the x- and y-direction respectively. z ] ! : 7 a | eto °o 2 4 2 6 8 10 Lilt tlt Figure 12. Values of C, and C, coefficients for determination of thermal stresses. In the case of CRCP, the thermal gradient effect depends on the crack spacing, being negligible for crack spacing shorter than one metre. Its effect is mainly the increase of Understanding Concrete Pavements a stresses during the first years of life with acceleration of the process of total cracking (Acker and Guérin, 1994). Uniform temperature variations originate volume changes with the consequent slab shortening or lengthening. Following a temperature rise, the length of an unconfined concrete slab will increase. If this expansion is restrained, compressive stresses are created. These stresses alone are usually much below the compressive strength of concrete, but, in combination with other factors they may lead to blowups, in which localised regions of the slab abruptly buckle upward. When the temperature drops, the slabs become shorter, and if this movement is restrained, the appearance of tensile stresses is inevitable. Therefore, cracks may occur or increase their opening, precipitating pavement destruction. In CRCP, the crack opening process may increase the tensile stress in steel behind the elastic limit, producing plastic, irreversible deformations (Acker and Guérin, 1994) 2.2.3 Fatigue of materials Cracking may occur in a concrete pavement slab when the tensile stress, caused by traffic and environmental conditions, exceeds the tensile strength of concrete. Alternatively, cracking can also occur as a result of repeated application of stresses smaller than the tensile strength of concrete. As pavements may be subjected to millions, of load repetitions during their life, fatigue plays a major role on pavement performance. An important factor that designers must be aware is the effect of overloads. In fact, concrete may sustain millions of load repetitions if the applied stresses are small compared with the concrete strength, however, if the applied stresses are very high, only a few load cycles are allowed before concrete cracking. To account for this effect, Domenichini (1986) suggests that the design of a concrete pavement should be made to consume only partially the fatigue life of concrete. In general, fatigue laws relate the number of cycles of loading, N, needed to cause failure to the stress level, S. The stress level is the ratio of the applied stress, o, by the flexural strength of concrete, fiza (also known as modulus of rupture of concrete). A comparison of some of these formulae can be found in Majidzadeh (1988), and Stet and Frénay (1998) and is summarised in Table 3. 2 Chapter 2 Table 3. Fatigue models for concrete pavements. Organisation Fatigue model 11.73-12.08*S, S>0.55 4.2577, Portland Cement Association - PCA (USA) log(N) = joan 0.45<5<0.55 ©, S<045 Eisenmann (Germany log(N)=11.79-12.23*S Corps of Engineers - CE (USA) log(N)=11.111-10.101*S Veverka (Belgium) log(V) =20-20*S Darter (USA) log(N) =17.61-17.61*S Austin Research Engineers - ARE (USA) log(N) =23440* 5°?! Vesic (USA): log(V)=22500*s~ IRO-Mats-CUR (the Netherlands): log(N) =14.61-13.78* S +2. aa Associaton ofthe Netherlands Cement Industy- gg) 16.8(0.9-S) sso, sci VNC (the Netherlands) 1.0667-S,i, m2 ‘Sn is the effect of the minimum stress Oniq induced by thermal gradients: A graphical comparison of these formulae is presented in Figure 13 where it can been seen that tremendous differences can be encountered between the fatigue models (when necessary in the formulas, the minimum stress is taken as zero). Some of the differences in the predicted fatigue life is the result of using different theories in determining critical concrete stresses. Therefore, to obtain consistent results, the same theory used in pavement design must be used in the development of distress functions. oMR o9 FR “Eisenman ce Vere o7 Date ARE Ves —roats-cu 0s am 03 Login) Figure 13. Comparison of some fatigue models. Understanding Concrete Pavements 2 2.2.4 Soil behaviour, water infiltration and joint efficiency Before World War Il, little attention was paid to the soils on which concrete pavements were laid because the concrete slab was much stiffer than the soil and the stresses induced in the foundation were small. However, if itis true that concrete pavements may be laid on weak foundations it also true that they demand uniform foundations. In fact, if for any reason a foundation settlement occurs, the slab may loose its support. If these foundation settlements were not compatible with the elastic deformation capacity of the slab, then cracks will occur in the slab. The loss of support may be due to insufficient compaction or to zones where earthworks have very low stiffness. Water is an important cause of poor soil behaviour and may significantly speeds up the deterioration of concrete pavements. An inadequate drainage, may have the following effects (Stock, 1988): a) reduce support in some areas to zero, resulting in critical stresses in edge and corner regions; b) redistribute subgrade and sub-base materials so that serious dislocations (faulting) occur in jointed pavements; ©) lead to accumulations of water which may contribute to blowups, lead to corrosion of load transfer devices, cause uneven swelling of subgrade soils, and lead to concrete deterioration and spalling. Pumping is the ejection of water and subgrade soil through joints and cracks and along the edges of concrete pavements due to the sudden load of the slab. When free water is present, the subgrade is fine-grained, churning of the water and soil occurs, forming slurry that is expelled to the surface. As pumping proceeds, the supporting soil is flushed from beneath the pavement at the affected locations. The consequence is faulting of the joints, transverse cracking and breaking of the comers, resulting in new opportunities for pumping action. To avoid pumping erosion resistant undercourses, with cement or bituminous treatment, have to be used (Oglesby and Hicks, 1982) If the subgrade soils are expansive, water percolating through joints and cracks would lead to differential expansion of the subgrade and the pavement may curl upward at the joints and become exceedingly rough. To ensure adequate pavement behaviour during its design life, it is essential that joins behave satisfactory. As can be seen from Figure 14, there is a strong relation between the allowable traffic that a pavement can withstand and the subgrade erodability, Pry Chapter 2 climatic conditions and the existence of dowel bars (Springenschmid, 1979). When dowels are not present (JUCP) it is fundamental that special attention must be given to the sub-base materials. Figure 14. Influence of dowel bars, subgrade type and climatic conditions on the allowable traffic limit per day. 2.3. Concrete pavement analysis The analysis of a concrete pavement determines the stress and deformation that are induced in the different layers of the pavement due to the applied loads. There are a number of variables that must be taken into account such as traffic, climatic conditions, pavement geometry, material properties and even the probabilistic aspects of all these variables. Therefore, an analysis of a pavement considering all of these parameters is not an easy task, especially because most of the variables are not known in advance. The analytical methods of pavement analysis assume some hypotheses relatively to the loads, constitutive laws of materials and boundary conditions in order to build a system of equations that once solved will give the stresses and strains at the pavement. Sometimes sophisticated and complex mathematical models are needed to solve these equations. The development of automatic calculation procedures (in terms of capacity, speed and accessibility) has increased the number and diversification of these models. Understanding Concrete Pavements 25 2.3.1 Analytical models Among a variety of models, the Westergaard model and the Burmister model deserve a special attention because they have been widely used and, in fact, are the support for many of the actual design methodologies. Both models presuppose an isotropic linear-elastic response for materials and are quite different in the basic hypothesis assumed. Westergaard model (1926) This model is probably the most widely used method for concrete pavement analysis. It assumes a thin slab (Kirchoff hypothesis) resting on a Winkler foundation. This simplification reduces the problem complexity and allows the modelling of loads applied at the centre, at the comer and at the edge of a slab. These load cases are of an extreme importance in concrete pavement analysis due to the presence of joints. The equations for the three load cases have already been presented in this Chapter. Another advantage of Westergaard model is its ability to model temperature loads. ‘The Winkler type foundation (or dense liquid foundation) assumes that the stresses at a point of the foundation are independent of the stresses at the other points. Thus, there is no coupling effect between the various points of the foundation. To overcome this limitation, some improvements of the soil model has been introduced, such as the Vlassov-Pasternak model (Shi ef al., 1993) or the Kerr model (Kerr, 1994). Burmister model (1943) The model proposed by Burmister allows the calculation of stresses on an n-layered foundation, with horizontal infinite dimension. All layers have elastic solid behaviour, which avoids the Bernoulli and the Winkler hypothesis and may be bonded or unbonded. The possibility of using several layers allows a better representation of pavement structures (at least three layers: soil, sub-base and concrete slab). The main limitation of Burmister model is that the layers are infinite in plane. Thus, the model is not enabled to tackle with the edge effects. To overcome this limitation it is possible to use the Westergaard model to calculate the ratio between the edge load and the centre load and apply this correction to the Burmister model (Jeuffroy and Sauterey, 1996). This model is very used for pavement analysis, especially for flexible pavements. As an indication the programs ALIZE (LCPC), BISAR (Shell) and ELSYMS (Berkeley University) are mentioned. 26 Chapter 2 2.3.2 Finite element models The analytical models described above can only study very simple states of stress and boundary conditions. The finite element method (FEM) is a much more sophisticated method that yields the modelling of much more complex geometry, boundary conditions and loads. With the advent of microcomputers, the FEM has suffered a tremendous development and almost every engineer has access to it. In the following paragraphs, some of the principal aspects related with the FEM in concrete pavement analysis are reviewed. Concrete slab modelling The FEM models for concrete pavement analysis can be divided into 2D models and 3D models. The pavement structures are clearly of 3D type. However, some problems have been analysed with 2D models but these formulations has been restricted almost to particular situations, essentially related with CRCP (Krauthammer and Western, 1988; Zollinger et al., 1994; Kim et al., 1994). Pavements are essentially laminar structures, which conducted to the use of plate elements for concrete pavement modelling. Most of the finite element codes use this type of element for pavement modelling (Tabatabaie and Barenberg, 1978; Majidzadeh et al., 1981, 1985; Huang and Deng, 1983; Tia et al., 1987; Houben, 1994). These elements are very simple because although they model a tridimentional behaviour their formulation is in fact bidimentional. These elements have 4 to 9 nodes with 3 degrees of freedom in each node (two rotations and one translation). Plate elements are unable to model concrete compressibility. It is possible to consider several layers of different materials, which allow the consideration of the sub-base. The layers are all bonded. They are adequate when the calculation capacities are reduced. The generated finite element meshes are easy to create and the results are easy to analyse. ‘The 2D formulation of plate elements does not permit the modelling of in-plane mechanisms like the length variation of slabs due to uniform temperature gradients or bonding between reinforcement and concrete. This can be overcome by the use of shell elements that have in-plane degrees of freedom, but are more complicated and need more powerful calculation means. The shell elements have not been used very much in concrete pavement analysis. The development of workstations and personal computers in terms of speed and memory, allowed the increasing use of 3D elements (or solid elements) for pavement analysis (Ioannides, 1988; Bennani, 1993; Channakeshava ef al., 1993; Scarpas, 1994; Understanding Conerete Pavements n Zaghloul and White, 1994; Masad et al. 1996; Vepa and George, 1997; Pane er al., 1998; Kuo, 1998). These elements have three degrees of freedom in each node and have 8 to 20 nodes in each element. Besides, for an adequate pavement modulation a tremendous number of solid elements is needed. In Figure 15, a typical finite element mesh with volume elements is shown. The bigger number of elements needed is related with the fact that for a satisfactory analysis through the thickness, three to four layers of elements are needed, just for the concrete slab. The pre and post-processing of data is a hard task when solid elements are used. However, the relatively simple geometry of road pavements greatly simplifies these procedures. Volume elements also need much more time to solve the equilibrium system of equations that is the heavy job in the finite element codes. Regardless of all the inconvenients, volume elements are probably one of the better options for the analysis of ‘most of the phenomena related with concrete pavements. Figure 15. Typical finite element mesh using volume elements with edge load. Foundation models Until now, only the elements for the slab have been presented. The foundation models will be forthwith discussed. Similarly to the slab, the FEM allows a number of models for the soil foundation layers. The simplest model is undoubtedly the Winkler model that can be seen as a series of vertical springs. This model has been extensively used in finite element analysis of 28 Chapter 2 concrete pavements (Tabatabaie and Barenberg, 1978; Chou, 1983, 1984; Ozbeki ef al., 1985; Ioannides er al., 1985; Tia et al., 1987; Nishizawa ef al., 1989, 1994; Guo et al., 1995). The soil is completely defined by the modulus of reaction, Numerically, this model is extraordinary effective and easy to implement in a finite element code. As the soil stresses are exclusively dependent on the vertical deflection, there is no coupling between the several model springs. Thus, this model gives relatively bad results for corner and edge loading. In the case of central loading, it is the slab itself that assures the interaction between the several springs and better results are attained (Kerr, 1994). Another widely used soil idealisation consists of using Boussinesq or Burmister type foundations (Wang ef al., 1972; Huang and Deng, 1983; Majidzadeh ef al., 1981, 1985; Nishizawa ef al. 1985). These models are much more realistic than Winkler soil, since coupling between the several points of the soil and the several layers can be used in the case of Burmister model. Nevertheless, from the computational point of view, these models are very inefficient. First, the stiffness matrix of the soil is a full matrix (not banded) which conducts to more time and memory to solve the problem. Finally, the stiffness matrix calculation is much more complex because the soil is not represented by finite elements. Any of the two model idealisations consider the soil reduced to the top layer. Thus, they are particularly adapted to be used with plate elements for the concrete slab. However, it is possible to use these foundation models with volume elements for the slab (Channakeshava ef al., 1993). A distinct approach for the soil is the use of solid elements. These finite elements are well adapted to model the characteristics of a foundation because different materials can be used for the different layers and the stiffhess matrix is banded and easy to calculate. Unfortunately, the problems mentioned with respect to the slab modelling with these elements are also present. Nevertheless, the use of volume elements has increased in the last years because of computer development. Interface modelling A very important feature of concrete pavements that must be appropriate modelled is the interface between the several foundation layers and, mainly, between the slab and the sub-base. In fact, separation between the slab and the sub-base occurs very often as a consequence of thermal gradients or particular loading conditions. Separation between the layers always involves the use of iterative procedures because the contact area is not Understanding Concrete Pavements 29 known in advance. Slip between the layers should also be modelled since it affects considerably the results of the analysis. The best way to model these interface problems is by putting interface elements between the elements of each layer. These usually zero-thickness elements allow the modelling of phenomena like slip or separation in a relatively simple way. Several types of interface elements are available and range from simple spring elements modelling the interface between a pair of nodes, to continuum elements (Figure 16). The continuum elements may be placed between two slab elements, two solid elements or a slab and a solid element (Saxena and Dounias, 1986; Ioannides, 1994; Davids et al., 1998; Hammons, 1998; Hammons and Metcalf, 1999). Figure 16. Continuum interface elements for solid-solid interface. Joint modelling The load transfer between joints is also a very important mechanism that should be modelled. The load transfer across a joint could be achieved by aggregate interlock, if the joint is sufficiently closed, or by mechanical devices as dowel bars or keys moulded in concrete. The simplest process of joint modelling is by adding a spring-type element between two adjacent nodes. These springs could transmit translations or rotations (Tabatabaie and Barenberg, 1978; Huang and Deng, 1983; Ozbeki ef al., 1985; Tia er al., 1987). A different approach was applied by Huang and Deng (1983) that use a joint efficiency that measures the percentage of displacement that is carried out throughout the joint. A joint efficiency equal to unity means that the joint behaves like a hinge and a joint efficiency of zero means that no load transfer will occur across the joint. The load transfer across the joint is reduced to a linear constraint between two adjacent nodes. 30 Chapter 2 When dowel bars are present at the joint, the most realistic way seems to be the use of beam finite elements that model the dowel bar. The effect of the compressibility of concrete could be modelled by spring elements connecting the beam to the plate clement (Tabatabaie and Barenberg, 1978; Nishizawa ef al., 1989; Channakeshava ef al., 1993) or by correcting the shear term of the stiffness matrix (Chou, 1983; Huang and Deng, 1983) Instead of beam elements, Guala and Petterle (1994) use plate elements with a corrected stiffness so that the plate stiffiness will be the same of the various dowel bars existing at the joint. The load transfer between the slabs can also be modelled by the use of special interface elements as the one developed by Nishizawa et al. (1989) or by Scarpas (1994). The former element is capable of transmit shear, bending and torsion between two plate elements. The latter element has the ability of modelling the effect of aggregate interlock, bond between the concrete and dowel bars and shear stiffness of the dowel bar and is intended to be used between solid elements. Material modelling Most of the finite element models for concrete pavement analysis use linear elastic constitutive relations for materials. Two key-factors contribute to the use of this hypothesis: simplicity and damage fatigue laws. Nevertheless, the real soil or concrete behaviour is clearly non-linear, especially after the onset of concrete cracking. Consequently, the use of non-linear material models is essential for a true and correct modelling of the pavement response. The basic factor of non-linear material behaviour for concrete in pavements is cracking since the compressive stresses present in concrete pavements are small compared with the compressive strength of concrete, Cerioni and Montepara (1994) present an orthotropic incrementally linear relationship with equivalent uniaxial stress concept to represent the behaviour of concrete. After cracking, a smeared orthogonal fixed crack model is adopted with tension stiffening effect. An elastic-plastic model for concrete with smeared cracks is presented by Channakeshava et al. (1993) where concrete strains after cracking are separated into elastic, plastic and cracking parts. Various models for non-linear soil behaviour have been proposed in the literature. Toannides ef al. (1984) use non-linear soil behaviour based on the concept of resilient reaction modulus that represents the ratio among the soil pressure and the clastic displacement of the soil. Tia ef al. (1987), in alternative to the linear elastic model, allows Understanding Concrete Pavements 31 the use of a parabolic non-linear elastic behaviour of the soil. Elastic-plastic constitutive relations for soil have also been used (Guala and Petterle, 1994; Cerioni and Montepara, 1994). All the referred models represent the soil with a Winkler model. Load modelling It has been already established that temperature loads are very important in concrete pavement analysis. The problem that analysts should be aware is the limitations of the selected finite element type. The use of plate elements for the concrete slab (Nishizawa and Fukuda, 1994; Nishizawa et al., 1994) restricts the temperature gradient to a symmetric case. The use of shell elements, which have in-plane degrees of freedom, overcomes this limitation. Nevertheless, the more general element that allows any temperature distribution throughout its thickness is the solid element (Saxena and Dounias, 1986; Masad et al., 1996; Pane et al., 1998; Kuo, 1998). Since most of material models are linear, fatigue modelling in finite element analysis is usually made by considering the effect of a standard axle load. Then, using the fatigue curves presented in 2.2.3 and a damage accumulation law, the effect of the applied load spectrum is calculated (Majidzadeh ef al., 1985). Alternatively, the effect of different axle loads and positions may be calculated and the admissible number of load repetitions determined using the fatigue laws presented above (Nishizawa ef al., 1985). A different approach is used by Bhatti ef al. (1996) that taking into account the number of load repetitions for a particular load reduces the Young modulus of concrete. Faulting is a cause of pavement damage that is not often taken into account on pavement analysis. In fact, it is very difficult to predict the erosion of foundation caused for instance by pumping. In finite element modelling, the loss of support is modelled through the reduction or elimination of the foundation stiffness in the eroded area (Majidzadeh et al., 1981; Larralde and Chen, 1987; Krauthammer and Western, 1988; Channakeshava et al., 1993). The analysis and design of concrete pavements is traditionally based on the assumption of equivalent static load. The dynamic vehicle-pavement-foundation interaction effects are important and increase as the slab becomes thinner and the soil becomes softer (Wu and Shen, 1996; Alvappillai ef al., 1993; Hardy and Cebon, 1993). Despite of its importance, only some attempts have been made to model the dynamic behaviour of concrete pavements (Zaghloul and White, 1994; Wu and Shen, 1996). Probably the main reasons for this lack of attention given to the dynamic analysis of 32 Chapter 2 pavements are the additional complexity introduced on the traffic loads and the absence of knowledge about the dynamic material behaviour. General considerations As can be seen, the FEM has almost unlimited modelling capabilities. The selection of a specific solution for the problem to be analysed is not simple and depends on a variety of factors such as the objectives of the analysis, time limitations or available computational tools. Topics such as the use of several types of elements with different degrees-of-freedom at the nodes are a well-controlled problem. Thus, it is possible to use shell elements for the slab, interface elements for the layer separation, volume elements for the soil, together with beam elements for the dowel/reinforcing bars and interface elements for load transfer at the joints. Table 4 shows the type of elements used in some finite element analyses found in the literature, by chronological order. With respect to the geometric modelling, the analysis of Table 4 shows a trend for the use of volume and shell elements for the concrete slab and sub-base. For the foundation, the tendency seems to be the modelling of the upper centimetres of the soil with volume elements resting on a Winkler foundation. However, the use of volume elements still presents some difficulties such as the lack of knowledge about the material triaxial laws, the time of analysis (if non-linear behaviour is modelled) or the memory required to solve the system of equations. It can also be observed an increasing tendency for the use of non-linear material constitutive relations. It is obvious that the FEM is a fundamental tool for concrete pavement analysis. Its aptitude to model boundary conditions, material laws and loading conditions of enormous complexity is much bigger than that of analytical methods. Nevertheless, there are a few questions that analysts must be conscious. Thus, besides model sophistication, it is also necessary to know the physical variables where the pavement is located such as interaction laws between the materials, triaxial fatigue behaviour, traffic and environmental loads. Understanding Concrete Pavements 33 Table 4. Summary of referred finite element applications for concrete pavement analysis by chronological order. Year authors ‘analysed problem Types of utilised finite elements Sab Sub-base Soll_interface’_Jomnts™= 1972 Wang etal. ‘concrete pavement analysis, PLA Bou 1978 Tabatabaie et al, RCP PLA PLAIN SPRIBEA 1981 Malidzaden et al. Design model for Pccp PLA BUR BUR. 1983 chou Comparison between Es andks PLA wiweou 1983 Huang and Deng RCP PLA Bou SPRIBEAREL 1984 chou Stresses on precast slabs PLA win 1984 loannides et al. Non-linear model for soil PLA PLA WINN. 1985 Malidzaden et al. Design mode! PIA BUR BUR Bea, 1985 Nishizawa etal. Design mode! Pla Bou BU ‘SPRINT 1985 ozbeki et a. Pavement evaluation PLA PLAIN ‘SPRIBEAREL 1985 loannides etal. _Westergaard solutions analysed PLA win 1986 Saxena and Dounias CRCP + TEMP vol VoL VOL nT 1987 Tiaet al src PA WINE 1988 loannides Comparison of several models. PLAVOL wiNvo. 11988 Krauthammer et al. Load transfer with pumping bz © aD wD. SPR+BEA 1988 Larralde andCnen Damage of concrete pavements PLA PLA. WINANL SPR+BEA 1989 Nishizawa etal. RCP PA win SPRIBEAREL 11995 Channakeshava et al. Norvinear model for JRCP and JUCI VOLNL WIN WIN SPR BEA 11994 Cerioni and Montepar Norviinear analysis PLANE WINNL —SPRANL 1994 Guala and Petterie cRCP PLA wane Pua 1994 Ioannides MultiJayered systems PLA PLA WIN INT 11994 Nishizawa and Fukud Temperature stresses PLA win sPR 1994 Nishizawa etal. ‘Temperature stresses. PLA win 11994 Scarpas. Element for load transfer at Joints VOL wt 1994 Zaghioule White Strengthening with asphalt vol vol RG BEAINT 1994 Zollinger etal. palling in CRC. 20 RIG FOR ser 1995 Guo et al Dowel bar mode! PLA win BEA +1998 Wu and snen Dynamic analysis vou win 1996 Masad et a ‘Temperature stresses in JUCP vol. vol WIN INT it 1997 Vepa and George JUCP vo. VoL vot sR 1998 Davids et a. arc. vo. vol WINNT BEA 1998 Bhatti et al. JRCP + cRcP SHENL wan 1998 Pane et a JUCP + TEMP vou win 1998 Hammons Joints of airport pavements SHE VOL) WIN INT sR 1998 Kim et al cRcP + TEMP 20 SPR BEA 41998 Kuo Temperature stresses VoL vol. WIN 4999 Hammons and Metcal Edge stresses in JUCP SHENOL_SHENVOL_WINWIN spr ~ interface between layers; =» Joints between adjacent slabs \VOL: volume; PLA: plate; SHE: shel; WIN: winkler; BOU: Doussinesq; BEA: beam; INT: Interface ‘SPR: spring; REL: linear constraint; 020: plane 20; BUR: burmister; NL: non-linear material formulation 2.4 Slab thickness design for concrete pavements A pavement must have a thickness that is adequate to support the applied loads during its service life. The preceding topics indicate that many variables should be considered in designing conerete pavement slabs. They have also shown that many of these variables are of difficult evaluation so that design calculations give only approximate answers. 34 Chapter 2 Since direct compressive stresses in concrete pavements due to wheel loads are very small compared with the compressive strength of the material, they can be ignored. Flexural strength is a key factor in thickness design, since the flexural stress produced by a heavy wheel load often is more than half of the flexural strength of concrete In this section, some design methods are briefly presented. Design methods can roughly be divided into rational and empirical. The former methods are based on the calculation of the stresses and deformations that should be kept between allowable limits. On the other hand, empirical design methodologies are based on the observation of several road sections under controlled traffic. 2.4.1 Rational models In the rational design methodologies, the calculations of the stresses and displacements are based on theoretical, analytical or numerical models of pavement analysis (usually the Westergaard model or the finite element method). The stress and displacement values must be kept within allowable limits to ensure good pavement performance. Several factors such as fatigue laws, joint type, load distribution and safety coefficient must be taken into account. The rational design methods could be presented in two different forms. The first is by using abacus, tables or formulas that permit the designer to freely design the structure within the allowable limits of the model. Alternatively, the methods could be presented by catalogues of pavement cross-sections that are related with guidelines that consider regional characteristics and allow comparison with other solutions. Anyway, both catalogues and general methods are based on some hypotheses and calculation models that are the object of the following presentation. Almost every country has its own design guide for concrete pavements. The main differences between these design guides are essentially based on national experience on concrete pavement construction regarding traffic, soils, aggregate type or climate. Consequently, there are some linking points between these design methodologies. A general flowchart for pavement design is presented in Figure 17 (Pinelo, 1996). It can be seen that a wide range of variables is necessary for a correct pavement design These variables depend on factors like available materials, climatic conditions, traffic growth and national authority policies. Nevertheless, the basic input parameters for any design method are traffic, foundation type and concrete strength. Understanding Concrete Pavements PROJECT DATA >| TRAFFIC ADMINISTRATIVE MATERIALS CLIMATE Spectrum Life Foundation Temperature axles __Analysis period Layers Rain PAVEMENT DAMAGE MECHANISMS CHARACTERISATION Fatigue Pavement type Permanent deformation Available materials [>| Foundation erosion Geometry 2 Drainage F 1 ge DESIGN DESIGN CRITERIA g3 Thickness design \<————_| Serviceabilty limit state LA Material ia Ultimate limit state oo PROJECT ANALYSIS ECONOMICAL ENVIRONMENT CONSTRUCTIVE ADMINISTRATIVE ASPECTS ASPECTS ASPECTS ‘Administrative Materials Equipment needs Know-how costs, competition maintenance and Noise Materials strengthening Alternative Workers materials User costs Phased construction No Design OK? Yes END Figure 17, Flowchart for concrete pavement design. Catalogue design methods The design methods based on catalogues of pavement cross-sections such as the Portuguese (JAE, 1995), the French (LCPC, 1993; Caroff and Leycure, 1989), the Spanish (Rocei and Kraemer, 1990) or the Belgian (Veverka, 1986) are essentially used in Europe. ‘The traffic is divided into categories based on heavy vehicular traffic per day, on the most loaded track, in annual daily average (ADTT). In Table 5 are shown the traffic categories according to Portuguese guidelines. 36 Chapter 2 Table 5. Traffic categories according to Portuguese guidelines (JAE, 1995). Category T Ts Ts Tw Ty T Tr To ADTT <50 50-150 __150~300 _300~$00 500-800 8001200 1200-2000 > 2000 The formation level of the pavement is classified according the long-term bearing capacity of the soil and the effect of subgrade on this bearing capacity. In Table 6, the Portuguese guidelines for formation level categories are summarised. Table 6. _ Formation level categories according to Portuguese guidelines (JAE, 1995). cans iE —— Fy 30-50 30 wT ay 50~80 60 Ts, Te, Ts, Ts Fy 80~100 100 Ti, Ta, Tos Tes Ts, Ts Fy > 150 150, Ti, Ta, Ts, Ts, Ts Te The next step is the selection of the pavement type and correspondent cross-section from the catalogue. In Figure 18 and Figure 19 the cross-sections for JRCP and CRCP according to Portuguese guidelines are presented. The thicknesses of the different layers are expressed in centimetres, and for a F2 foundation. ‘CC cement concrete; LC: Tean concrete Values in em for foundation F, c23 oe Fy: +2 em CC; +5 em LC; Ts and Ts coe Fy: 2emec Teas Fa:-3emec Ts, Ts TT. T,, T; Figure 18, _JRCP cross-sections according to Portuguese guidelines (JAE, 1995), The guidelines allow that for traffic categories Ts, Ts and Ts the choice of a JUCP instead of IRCP may be investigated. Because of the specificity of traffic, categories T; (very low traffic) and Ty (very intense traffic) must be object of a more detailed study. ‘CC: cement concrete; LC: lean concrete | Values in em for foundation F, cc20 coo Fx H2emCC;+SemLC; TyandT, | oo, Fj:-2emcc | | LC15 i Ts Ts TT. TG Reinforcement: 0,6%~0.7% | Figure 19. _CRCP cross-sections according to Portuguese guidelines (JAE, 1995). Understanding Conerete Pavements 3 Regardless of the lifetime period of 20 or 30 years, an analysis period of 40 years should be assumed with consideration of several pavement alternatives. These alternatives should compare the costs in terms of construction, maintenance and inconvenients to users. Universal design methods The alternative to the pavement catalogues are usually called as “universal” design methodologies. In this case, the designer is free to choose the pavement characteristics within the allowable limits of the model. Nevertheless, these two methodologies must not be looked as competitors but rather as complementary ones. In fact, the design catalogues are very useful for alternative comparison, while the universal methods allow for studies that are more detailed and for thickness optimisation. The basic input for design methods are the load spectrum, pavement type, slab support modulus and concrete properties (Young modulus and flexural strength). The traffic loads are usually divided into classes, and the number of expected load repetitions in each class must be estimated. A strength criterion is always checked because of concrete fatigue. Some design procedures such as the Belgian (Veverka, 1986) or the VNC method (Most and Lewis, 1986) also use a rigidity criterion that bounds the slab deformation to certain limits (two to three millimetres). This limit deformation is necessary to control the slab loss of support. Temperature effects may be taken into account and must be combined with traffic. Hence, the probability of occurrence of high temperature gradients and high traffic should be analysed to properly study the fatigue effect (Faraggi er al., 1986; Inoue and Uchida, 1986; Yao et al., 1990). The type of pavement (JUCP, JRCP or CRCP) must also be known since the join behaviour is radically different in each type of pavement and, consequently, very different performances are expected. Moreover, the temperature effects are very distinct in the different types of pavements. In fact, for short slabs the temperature effect is almost negligible as can be deduced from Figure 12 (for example, the Belgian procedure does not consider temperature effects for CRCP and for pavements with joint spacing less then 6 metres). The slab support is usually considered as a whole. Thus, an equivalent value for the deformation modulus of the support must be found. Commonly, the modulus of reaction is utilised since it is used in Westergaard formulas (or their improvements). 38 Chapter 2 A special attention will be given to the Portland Cement Association method because it is a widely used method and is recommended in Portugal. As the objective of this section is just a presentation of the design methodologies, only a few design tables and abacus will be presented. A complete set of abacus and working sheets could be found in Miranda (1988). The method is based on more than 60 years of experience based on investigation, development and monitoring of pavements. This experience includes: i) theoretical analysis of pavements based on Westergaard model and finite element analysis; ii) tests done in models and in situ, conducted by the PCA laboratories and US Bureau of Public Roads; iii) study of experimental roads under controlled traffic; iv) behaviour of constructed concrete pavements. ‘The basic input parameters of the model are: i) traffic: loads, frequency and axle types (single, tandem, or triple); ii) lifetime period: mnsidered as 20 years, but passive of being changed; iii) concrete strength: flexural strength at 28 days; iv) reliability: load safety factor that is 1.2 for high heavy traffic, 1.1 for medium traffic and 1.0 for low traffic; v) slab support: modulus of subgrade reaction, taken into consideration the existence of cement stabilised sub-bases or made of lean concrete. The design is a trial-and-error procedure, wherein starting from a specified slab thickness the two criteria, fatigue and erosion, must be fulfilled. If these criteria are not accomplished, then a new thickness must be chosen. In rough terms, the fatigue criterion is determinant in low or medium traffic pavements, while erosion criterion is conditioning in cases of high traffic. It can also be observed that single axles are more severe for fatigue analysis, while tandem axles are more severe for erosion analysis (Cardoso, 1990). Starting with the modulus of reaction of foundation soils, an equivalent modulus of reaction is calculated as a function of the sub-base thickness and stabilisation with cement (Table 7). For fatigue and erosion analysis, the traffic must be divided into classes of loads per axle, Then, for each class, the load must be multiplied by the safety factor to obtain the design load P,, and the corresponding expected number of repetitions must determined. Understanding Concrete Pavements 39 Table 7. _ Modulus of reaction for PCA method. Modulus of Equivalent modulus of reaction (kPa/mm) as a function of the sub-base thickness (mn). textlnofthe i eee ee Reader, Unstabilised sub-base Stabilised sub-base kPa/mm 102, Bore ea ae a 1 aa ease 13.6 76 204 +231 2299 ~«46:1 «624 —«BAT~~«WOSSD 71 353° 380434 S16 76.0 1086141 173.7 343 $9.7 24 73.3 -86.9 1276 = 173.7253 : 814 869 896 10041167 : : : For fatigue analysis, an equivalent stress, S’, must be calculated, as a function of slab thickness and equivalent modulus of reaction. The equivalent stress is dependent on the existence of concrete shoulders. The equivalent stress is then divided by the concrete flexural strength to obtain the stress intensity factor for the three types of axles (Table 8 shows the equivalent stress for pavements with concrete shoulders). Table 8. _ Equivalent stress for pavements with concrete shoulders. A Equivalent modulus of reaction (kPamm) 736 274 543 a1 136 180 nyo) eed tale ata sea ad te [em RR 102/441 368 297/385 3.23 2.70] 3.37 291 2.54] 3.12 2.78 250/282 268 248/204 2.65 248 127|3.28 2.79 2.19]2.88 2.41 1,942.63 2.12 1.78|2.35 2.00 1.72|2.14 1.89 1.68) 2.03 1.84 1.67] 152] 2.56 2.24 1.72]2.28 1.91 1.60]1.99 1.66 1.35] 1.86 1.55 1.29]1.70 1.45 1.24161 1.40 1.23] 178| 2.08 1.86 1.41]1.83 1.59 1.23]1.63 1.37 1.09] 1.52 1.27 1.03|1.40 1.17 0.98] 1.32 1.12 0.95| 203/1.74 1.60 1.19 1.53 1.35 1.03] 1.36 1.16 0.90] 1.28 1.07 0.85] 1.17 0.98 0.80|1.12 0.93 0.77 228) 1.48 1.39 1.01] 1.31 1.18 0.88/ 1.17 1.01 0.77] 1.09 0.92 0.72| 1.01 0.84 0.68) 0.96 0.80 0.65 2541.28 1.23 0.88) 1.13 1.04 0.77] 1.01 0.89 0.68] 0.94 0.81 0.63] 0.88 0.74 0.58|0.83 0.70 0.56 278] 1.13 1.11 0.7] 0.99 0.93 0.68/0.89 0.79 0.60] 0.83 0.72 0.58|0.77 066 051/073 062 0.49) 305] 1.00 1.01 0.67] 0.88 0.84 0.6110.79 0.72 0.54] 0.74 0.86 0.50] 0.68 0.59 0.46|0.66 0.56 0.43 330]0.90 0.92 0.59] 0.79 0.77 0.5410.70 0.66 0.48) 0.66 0.59 0.45/0.61 0.54 0.41] 0.59 0.50 0.39 356] 0.81 0.84 0.52] 0.72 0.71 0.49 0.64 0.60 0.43] 0.60 0.54 0.41] 0.56 0.49 0.371053 0.46 0.35 = single axle; d = tandem axle; t= triple axle. The allowable number of load repetitions for each axle and load class is determined with equation presented in Table 3 where £. 82 Ss ', with the axle load P, expressed in KN. (2.5) Alternatively, Figure 20 can be used to calculate the allowable number of load repetitions. The predicted number of load repetitions must then be divided by the allowable number of load repetitions, to obtain a percentage of fatigue. If the sum of all percentages of fatigue is less than 100 then the pavement satisfies the fatigue criteria. 40 Chapter 2 Figure 20, Allowable number of load repetitions for PCA method (left: fatigue criterion; right: erosion criterion). Erosion analysis is performed in a similar way to fatigue analysis, but instead of the equivalent stress, an erosion factor is calculated (Table 9 shows erosion factors for pavements with concrete shoulders and dowel bars). The erosion factor depends on the existence of concrete shoulders and depends on the existence of dowel bars at the joints. With the erosion factor and with the help of Figure 20 the allowable number of load repetitions for each class and axle type is calculated. Then, the percentage of erosion is calculated and the sum of all percentages of erosion must be less than 100 in order to fulfil the erosion criterion is. Table 9. _ Erosion factor for pavements with concrete shoulders and dowel bars. A Equivalent modulus of reaction (kPammm) 136 2A 54.3 614 136. 190 mms dt ed tae sees nena steam ine dt 102/228 2.28 2.30223 2.21 22i;221 2.16 2.16)2.20 214 2142.17 213 2102.16 2.12 2.07 127| 2.08 2.13 2.18|2.05 2.05 2.08] 2.02 1.99 1.99]2.00 1.96 1.95] 1.98 1.92 1.92] 1.97 1.91 1.90 152| 1.92 2.02 2.09] 1.90 1.94 1.98] 1.86 1.67 1.88] 1.85 262 1.83]2.52 1.78 1.78|1.81 1.75 1.75 178) 1.80 1.92 2.01] 1.77 1.85 1.90] 1.74 1.77 1.80] 1.72 1.72 1.74] 1.70 1.67 1.68] 1.68 1.64 1.63 203] 1.70 1.85 1.95] 1.66 1.77 1.84] 1.63 1.68 1.73] 1.61 1.64 1.67] 1.59 1.59 1.60] 1.57 1.54 1.55] 229] 1.60 1.77 1.90] 1.57 1.70 1.79] 1.53 1.61 1.68] 1.51 1.57 1.61] 1.49 1.51 1.54] 1.47 1.47 1.48] 254 1.52 1.70 1.84] 1.48 1.63 1.73] 1.45 1.55 1.62] 1.43 1.50 1.56]1.40 1.44 1.48] 1.39 1.40 1.43] 279| 1.45 1.85 1.79] 1.41 1.57 1.69] 1.37 1.49 1.58] 1.34 1.44 1.52] 1.32 1.39 1.43] 1.30 1.34 1.38] 305] 1.38 1.59 1.75] 1.34 1.52 1.65] 1.30 1.44 1.54] 1.28 1.39 1.47] 1.25 1.33 1.39/1.23 1.29 1.34] 330] 1.32 1.54 1.71] 1.28 1.47 1.61] 1.23 1.39 1.50] 1.21 1.34 1.43] 1.19 1.28 1.35] 1.17 1.24 1.30] 366| 1.25 1.50 1,68] 1.21 1.42 1.57] 1.18 1.34 1.48]1.15 1.30 1.40] 1.13 1.24 1.32] 1.11 1.20 1.26] ‘$= single axle; d= tandem axle; t= triple axle. Understanding Concrete Pavements 4 The abacus presented in Figure 21 allows the determination of a reduced slab thickness if a sub-base made of lean concrete is used. The input data is the slab thickness calculated by the previous method in the absence of any sub-base. The output data is the thickness of the slab with a sub-base of lean concrete, as a function of sub-base thickness and sub-base flexural strength. Mods of rupture ot 40 193 172 2at 3 leaneonarae Pa) 241 ail. Ey ‘Sab hetness ney) ‘Sib irs re) susubecwmes Pa SSthibe cnc 2 2 MPa Figure 21. Reduced slab thickness in presence of a lean concrete sub-base. 2.4.2 Empirical models The empirical models are based on the observation, under controlled conditions, of real or experimental pavements. The process starts with the definition of a criterion that states the end of the pavement lifetime. Using this criterion, the lifetime of the pavement is observed. Over the population of test sections some relations between the pavement lifetime and some geometrical characteristics of pavements (such as layer thickness or slab length) and certain mechanical properties of materials are established using multiple regression techniques. These methods denote a few drawbacks. First, the lack of a theoretical model, which requires the consideration of an important number of test sections. Second, need of long evaluation periods, because they are based on the end of the pavement lifetime. Finally, their applicability is restricted to conditions similar to those used on tests. 42 Chapter 2 The usefulness of empirical methods is related with the fact that one observed parameter is the degradation or maintenance condition of the pavement, allowing the comparison between the predicted condition and the observed condition of pavement at the end of the predicted lifetime. Two of these methods are resumed below. American Association for State Highway and Transportation Officials (AASHTO) The design method proposed by AASHTO (1986) was based on the results attained on AASHTO test road. It represents the pavement lifetime by the number of equivalent standard axles that the pavement can support. The conversion from the number of axles of each type to equivalent standard axles depends on pavement thickness and on the final pavement condition, Therefore, this is an iterative method, where from an initial thickness, the design equations are solved, and the new thickness is compared with the initial. The design equation is very complex and is presented in Box 1. Box 1. AASHTO design model. log(APSI/3) “, L.624*10" (p+1)" log (4) = 7:35log(D +1) -0.06+ +(4.22-0.32p,)+ 1 M,C,(D°"s -1.132) 215.63/(D"* ~18.42/(E,/k)”) +Z,S, log] where: W qs the number of equivalent standard axles of 80 KN, Z, is a random variable related with the considered risk, Sis the standard deviation of errors related to traffic and material performance, Dis the slab thickness in inches, APSI is the difference between the initial PSI value (Present Serviceability Index), around 4 to 4.5, and the final PSI, around 2 to 2.5, for which the pavement is designed, 'M, is the modulus of rupture in psi (1 psi = 6.89476 kPa), Jis a joint efficiency coefficient, Ca is'a drainage coefficient, Ec is the concrete elastic modulus in psi, and ‘kis the foundation modulus, in pei (1 pei = 0.27145 kPa/mm). Transport and Road Research Laboratory (TRRL) This design method (Mayhew and Harding, 1987) was based on the observation of experimental sites of concrete pavements, both unreinforced and reinforced (slab length about 30 metres). The equations of the model could be found in Box 2. Understanding Concrete Pavements 43 Box 2. TRRL design model. For unreinforced concrete pavements: Ln(Z) = 5.094Ln(H) + 3.466 Ln(S) + 0.4836 Ln(M) + 0.08718 Ln(F) — 40.78 | For reinforced concrete pavements: Ln(L) = 4.786 Ln(H) + 3.171 Ln(S) +1.418Z n(R) +0.3255Ln(M)-45.15 | where Lis the cumulative traffic in millions of standard axles, His the thickness of the slab in mm, ‘Sis the 28-day mean compressive strength in MPa, ‘Mis the foundation equivalent modulus in MPa, and Fis the percentage of failed slabs. Ris the amount of reinforcement in mm*/m. From the analysis of the expressions presented in Box 2, it can be concluded that thickness is the most significant variable, accounting for 67 percent of the variation observed in performance. 2.5 Summary In this chapter, the different types of concrete pavements are described and their main characteristics outlined. Then, the main damage mechanisms of this type of pavements such as traffic, temperature, soil behaviour, fatigue of materials or join efficiency are analysed. The methodologies of analysis of concrete pavement are reviewed with particular emphasis to the finite element method. This method is nowadays the most powerful tool for structural analysis since it allows the modelling of complex boundary conditions, non-linear material behaviour and complex loading histories. Finally, an overview of design methodologies is presented. Rational and empirical design methodologies are summarised with particular attention to the design methods used in Portugal. 44 Chapter 3 3 FATIGUE OF PLAIN AND FIBRE-REINFORCED CONCRETE Concrete structures are preponderantly subjected to static loads or that for practical purposes may be considered as static. In a few cases of dynamic loads, it is possible to analyse the problem through the consideration of a dynamic factor that allows the consideration of the dynamic load as an equivalent static load. These types of loads should only occur sporadically during the design life of the structure. Nevertheless, some types of loads continually fluctuate in time and occur during the entire lifetime of the structure. These loads are called fatigue loads. Examples of such loads are traffic, wind in slender structures, sea waves, machinery in work, variable differential hydrostatic pressures and temperature. The continuous modification of the internal structure of a certain material, due to the actuation of repeated loads is called fatigue. The change in material structure may result in the initiation and propagation of micro-cracking and subsequent failure of the structure. This means that, because of fatigue, a structure may collapse with the continuing action of a load that acting once is unable to cause failure. Fatigue performance generally depends on a variety of factors such as material characteristics, geometry at the crack initiation sites, stress-strain history and environment, among other factors. At a macroscopic level, concrete is a heterogeneous material compound by aggregates enclosed by the cement paste. The mechanism of fatigue in concrete may be attributed to the progressive deterioration of the bond between the coarse aggregate and the cement paste, together with an accompanying reduction of the section of the specimen. The final failure of the specimen occurs by fracture of the cement paste. An alternative mechanism that may act simultaneously with the previous one, is attributed to the formation and propagation of small cracks in the cement paste until the resulting crack pattern weakens the section to the point where it cannot maintain the applied load (RILEM, 1984). Concrete is a brittle material and its tensile strength is nearly ten times less than its compressive strength. Thus, reinforcements must be placed when tensile forces have to be supported. However, concrete is a relatively low-cost material and may be shaped with almost every configuration. Consequently, the perfect construction material should be identical to concrete in what concerns to costs and workability, and should have sufficient Fatigue of plain and fibre-reinforced concrete 4s strength and ductility so that conventional reinforcements could be avoided. This is the fundamental concept that is behind the addition of fibres to concrete. Fibres should be capable of sustain the development of cracks in concrete supplying the necessary ductility and improving its strength. Nevertheless, this is not what really happens when concrete is reinforced with fibres. Although the ductility and the energy absorption capacity are easily increased with fibres, the increase in strength can only be attained with large percentages of fibres, which interdicts the use of these composites in current structural applications. This effect of the addition of fibres on the fatigue behaviour of concrete is strongly dependent on concrete composition, type of fibre used, orientation of fibres relatively to principal stress directions and volume percentage of fibres used in the mix. Thus, it can be seen that fatigue behaviour of fibre-reinforced concrete is a very complex phenomenon with additional difficulties, compared with plain concrete, caused by the presence of fibres. 3.1 Fibre-reinforced concrete Several materials have been used as fibres for concrete reinforcement. Fibres can broadly be classified as metallic, polymeric, mineral or natural. Metallic fibres are made of steel or stainless steel and are the most widely used. Polymeric fibres in use include acrylic, aramid, nylon, polyester, polyethylene, polypropylene and carbon fibres. Mineral type fibres are essentially the glass fibres, while a variety of naturally occurring fibres have been tried in concrete such as akwara, bamboo, coconut, jute, elephant grass, sisal, sugarcane or cellulose fibres. Table 10 summarises some properties of fibres. The geometric characteristics and properties of metallic fibres are presented in a more detailed way, since those are the fibres, used in this study. Steel fibres with various shapes are available: straight, hooked-end, crimped or twin- cone (Figure 22). There are also several cross-sections ranging from circular to rectangular. Besides its shape, another important geometric parameter of a fibre is its aspect ratio, which is defined as the ratio between the fibre length and its diameter. For non-circular sections, the diameter is determined as the diameter of a circle that has the same area of the fibre cross-section. Typical aspect ratio for steel fibres ranges from 30 to 100 or more 46 Chapter 3 Table 10. Geometrical and mechanical properties of various fibres (Balaguru and Shah, 1992). Diameter Tensile Elastic Type Fie tanh Pegi) SBE Seng Mods Sein” {x10 mm] ‘ty (Pa) __(MPa) (%) ‘Acrylic NA [13-104] 117-207-1000 146-196 7.5-50.0 Aramid 1 NA 12) 1443620 62 44 ¢ Aramid II ‘NA {10} 144 3620 17 25 E Nylon NA NA 116 965 5.17 20.0 & Polyester NA NA 134-139 896-1100 17.5 NA Polyethylene 12-50 [25-1020] 0.96 200300 5.0 3.0 Polypropylene 6-50 NA 0900.91 310-760__3.5-49 15.0 Coconut 50-350 0.104 112115 120200 19:26 1025 Sisal NA NA NA 280-568 13-26 35 3 Sugarcane NA 02.04 12-13 170-290 15-19 NA zg Bamboo 10-100 0.05-040 15 350-500 33-40 «NA to Jute 180-300 0.1-0.2 1.02-1.04 250-350 26-32 1.5-1.9 Elephant grass NA NA NA 178 49 36 Cellulose 255.0 0.015.008 1.5 700 NA NA Giass fibres 3850 NA 270-274 250035003135 25-36 Steel 25-75 0.25-1.0 7.85 350-1500 200 NA NA = Non-available or directly non-applicable property. Hii Gil | Hookesend —Flatlenedand Machined chip Mek etract Figure 22. Types of steel fibre shapes. The volume percentage of fibre reinforcement, Vmia, to ensure that reinforcement failure does not occurs immediately after matrix first structural cracking, supposing that all fibres are aligned may be given by (Naaman and Reinhardt, 1996) La Su . @G.1) Fatigue of plain and fibre-reinforced concrete 47 where fmu is the tensile strength of the matrix and fy is tensile strength of the fibre. Supposing a typical matrix tensile’ strength of 3.5 MPa and fibre tensile strength of 1400 MPa the minimum volume should be around 0.25 percent. Fibre composites commonly used in bulk field applications contain less then one percent of volume in fibres. For these fibre contents, increasing in strength (both tensile and compressive) is unlikely to be achieved. On the contrary, the post-peak behaviour is radically modified even for these low fibre contents. 3.1.1 Tensile behaviour The use of steel fibres in a brittle matrix like concrete in contents below one percent results in an insignificant increase in strength. Since the ultimate strain capacity of the fibre is higher than the ultimate strain capacity of the matrix, the matrix fails before the full capacity of the fibre is attained. The energy dissipation is increased through the debonding and pullout of the fibres that bridge the opening crack. Figure 23 shows the effect of fibres on the tensile behaviour of concrete. Localisation ‘Additional energy due to multple cracking oa Elongation Figure 23. Effect of fibres on the tensile behaviour of concrete. FRC: fibre-reinforced concrete (Vj< 1%); HPFRCC: high performance fibre-reinforced concrete (V;> 5%). Naaman and Reinhardt (1996). 48 Chapter 3 The softening behaviour of steel fibre-reinforced concrete is dependent on the mechanical and geometrical properties of fibres, as well as on the distribution and orientation of fibres. It is also dependent on matrix properties and fibre-matrix interaction (Barros, 1995). Thus, it is a very complex behaviour and is very difficult to simulate it. 3.1.2. Bending behaviour The effect of the addition of fibres to concrete under bending loads is presented in Figure 24 for several fibre contents (Balaguru and Shah, 1992). It can be seen that post- crack performance is significantly improved by the addition of fibres, even at low percentages. Increases in strength can only be attained for high fibre contents. It is also visible that above a certain volume of fibres, the strength and toughness did not increase any more. In fact, at these high fibre contents, the method of casting becomes very critical. (Won 98) Fe content 90 gin 2 Detection (mm) Figure 24, Influence of fibre content on load deflection curves in bending: hooked-end, 50-mm length fibres (Balaguru and Shah, 1992). Longer fibres provide higher increase in bending strength, because fibres tend to be aligned with the beam axis, and provide a better anchorage to concrete. Nevertheless, as longer the fibre the harder is the mixing process with a consequent deficient compaction, which may lead to a decrease in strength above a certain fibre content. Fatigue of plain and fibre-reinforced concrete 49 3.1.3 Compressive behaviour It is commonly accepted that the influence of fibres on the compressive strength is insignificant at low volume fractions. Various observations evidence that, even for the same material, the compressive strength may first rise followed by a drop with increasing fibre content. This suggests that for standard mixing procedures there is a competing process of strength improvement as well as degradation. In fact, fibres can introduce additional defects that increase the initial crack density and therefore high intensity of crack interaction. Nevertheless, it is possible to enhance compressive strength even for high fibre contents when special processing techniques are employed (Li et al., 1996). Figure 25 presents the effect of fibre volume both on compressive strength and softening behaviour. The addition of fibres increases the strain at peak load, however, the change in modulus of elasticity can be considered negligible. Stress [MPa] Fibre content: 30 kg/ni Plain concrete 0002 0004 000s 000s 0010 Strain {mm/mm} Figure 25. _ Effect of fibre content on compressive behaviour of concrete (Balaguru and Shah, 1992). The increase in ductility depends on a number of factors including fibre volume fraction, fibre geometry and matrix composition. A notorious rise of the energy absorption capacity is found for increasing fibre contents. For straight fibres, ductility increases with aspect ratio. The amount of fibres necessary to attain a certain ductility increases with increasing matrix strength (Barros, 1995). 50 Chapter 3 3.2. Fatigue basics Fatigue is a process of progressive internal changes occurring in a structure, or structural member, subjected to repeated load. The evolution of the internal structure may lead to a progressive crack growing and complete failure of the structure. For many decades, attempts have been made to gain a better understanding of the fatigue phenomenon through both theoretical and experimental research. General knowledge of fatigue mechanism and behaviour of ductile materials, such as steel, is almost complete. However, the fatigue mechanism and behaviour in brittle materials such as concrete is still lacking. Fatigue of concrete structures has received increasing interest due to several reasons. First, new types of concrete structures, such as marine structures and railroad ties, have been developed where the full design load may act for a very large number of cycles. Second, there is also an increasing recognition that repeated load may aggravate the crack width, deflection and stiffiness under service load even if it does not cause fatigue failure. Third, the augmenting application of high-strength concrete will result in more slender structures where the repeated live load forms a large part of the total load. Finally, the continuing refinement of design procedure requires a clear understanding of material behaviour, and fatigue is one of the properties of concrete that is less understood. 3.2.1 Fundamental definitions The stress level or fatigue stress is defined as the percentage of the monotonic strength that is applied to the tested specimen: 5 = Zale | G2) Sa where fy is the monotonic strength. The maximum stress level used during a fatigue test is denoted by Syuax While the minimum stress level is denoted by Syn. The ratio between the minimum and the maximum stress level, the fatigue ratio, reads G3) Fatigue of plain and fibre-reinforced concrete st The fatigue amplitude is the difference between the maximum and the minimum stress level and reads ASouax = Snax ~ Sin + (3.4) The most important fatigue property is probably the farigue life, N. Fatigue life can be defined as the number of load repetitions that a structure or structural member can sustain for a given stress level. Alternatively, fatigue strength, also represented by S, can be defined as the fraction of the monotonic strength that can be supported by a structure or structural member for a given number of cycles. Therefore, fatigue strength is closely related to fatigue life. Actually, the most common way to represent fatigue data is by mean of S-N diagrams, also known as Whéler-curves. Another important definition used in fatigue characterisation of materials is the endurance limit or fatigue limit. Endurance limit, S,,, is defined as the percentage of the static strength that can be supported indefinitely. Thus, the endurance limit is the fatigue strength when the number of cycles is infinite. In practice, values of two to ten million of cycles are considered as infinite, since this is the maximum number of load cycles that usually occurs in structures. Fatigue loading is usually classified in two categories (RILEM, 1984): low-cycle loading and high-cycle loading. Low-cycle loading is characterised by few load cycles (generally less then 1000) of high stress levels. High-cycle loading is characterised by a large number of cycles (generally more than 1000) at lower stress levels. A real load history may be constituted by both types of loading. Hsu (1981), defines a third class of fatigue loading called super-high-cycle loading that applies to the range 10 million to S00 million load cycles (Table 11). Table 1]. Fatigue load spectrum (Hsu, 1981). Low-cycle fatigue High-cycle fatigue Super-high-cycle fatigue Mass ; Highway and Structures subjected to | Airport pavements |_| isS0"0) rapid y bridges, Sea structures earthquakes and bridges transit highway pavements| 1 10" 10? 10° 108 10° 108 10” 108 10° 2 Chapter 3 3.2.2 Factors affecting fatigue performance A number of factors affect the fatigue life of concrete including the stress range, the load frequency, the concrete composition, the mixing and curing conditions. The stress range affects the fatigue life considerably. The variation of the fatigue life with respect to the stress range can be easily assessed by means of the well-known ‘Aas-Jakobsen formula (Tepfers and Kutti, 1979): ~All-R)log(N), (G.5) where f is a material parameter. Therefore, fatigue strength decreases with decreasing values of R. Small R-values mean lower Smiq for constant Snax Of, bigger Sax for constant ASwax- This reduction is even more pronounced in the case of negative R-values. The effect of frequency (1 to 15 Hz) on the fatigue life is generally considered insignificant for stress levels below 0.75. For higher stress levels, the fatigue life increases with increasing frequency (Comelissen, 1984; CEB, 1988). In the theoretical studies of Hsu (1981) a formula has been proposed wherein the frequency (in fact the period) of the applied load affects the fatigue life of concrete: 1.2-0.2R -0.133(1-0.779R)log(N)- N<1000 S= -0.053(1—-0.4452)log(7), (3.6) 1—0.0662(1-0.556R)log(N)-0.0294log(7), N >1000 where T is the period of the cycle in seconds. These expressions included normal and lightweight concrete in compression and bending and even some results of tests with stress reversals. The effect of frequency is more pronounced for frequencies below 1 Hz (Siemes, 1988). The curing conditions, the water-cement ratio, cement content and the amount of entrained air are variables that do not seem to influence the fatigue strength if it is expressed in terms of the corresponding monotonic strength (CEB, 1988). The fatigue life of conerete tested in wet conditions is lower then that of concrete tested in air (Cornelissen, 1984; Alliche, 1988; CEB, 1988; Fouré, 1998). Fatigue of plain and fibre-reinforced concrete 3 3.2.3. Fatigue testing Fatigue testing is a very time-consuming process. As an example, if a frequency of 5 Hz is used, a one million cycle test will last for almost fifty six hours. A specimen can be subjected to a large variety of loading schemes. The constant amplitude-loading scheme is the simplest, and the widely used one, and consists of subjecting the beam to a chosen constant minimum and maximum load repeatedly. Load variation with time is usually a sine wave but rectangular and triangular wave-shapes have also been reported (RILEM, 1984). The most common way to do a fatigue test program in order to establish the fatigue material properties consists of selecting a series of maximum stress levels, say 0.9-0.8-0.7-0.6-..., and then testing a number of specimens at each stress level. The minimum stress level may be chosen constant, or alternatively, it may vary in order to make the ratio R constant, Paskova and Meyer (1994) state that the minimum number of specimens tested at each stress level should be five. Nevertheless, Ramakrishnan and co- workers (Ramakrishnan ef al., 1991, 1996) use a somewhat different procedure. Starting from a maximum stress level of, say 0.9, a specimen is tested and if the specimen fails, the maximum stress level is reduced to 0.85. If the specimen does not fail in 2 million cycles, then another specimen is tested as a duplicate. The procedure stops when at a specified stress level three specimens do not fail. The minimum stress level is kept constant at 0.1. Due to the large scatter existing in concrete fatigue data, the last procedure does not provide enough information to build Whiler-diagrams and may be adequate only for the estimation of the endurance limit. It should be noted that the monotonic strength is usually determined from tests where the rate of application of load may be several times less than the rate of loading in fatigue tests. Because the monotonic strength is quite influenced by the rate of loading the resulting values of the fatigue strength obtained from several experimental programs may not be directly compared with each other. Therefore, the need for a standard fatigue-test procedure for concrete is mandatory. If after a fatigue test a specimen did not fail, then it is called a run-out specimen. A monotonic test is usually made on run-out specimens to investigate their post-fatigue properties. This post-fatigue behaviour is obviously related with the damage induced in the specimen by the fatigue loads. In fact, as during the first stage of fatigue testing the consolidation prevails over damage then it is expected that the post-fatigue properties of 34 Chapter 3 these run-out specimens will be better than the unloaded specimens (higher stiffness and strength). On the contrary, if the fatigue test stops later at a stage where severe damage has already occurred in the specimen, then the post-peak properties of the run-out specimens will be lower. 3.2.4 Modelling fatigue data As mentioned previously, Whdler curves are the most common way to represent fatigue data. As the scatter in data is usually very large, a regression analysis has to be made and appropriate confidence limits have to be calculated. As can be concluded from the work of Hsu (1981), a simple relationship between the applied fatigue stress and the fatigue life is difficult to obtain. Regression analysis is used to determine the functional relationship between variables and correlation analysis is used to determine the degree of association between the variables (Ramakrishnan and Lokvik, 1991). The procedures for calculating the regression lines and confidence limits for an S-N diagram could be found for example in CEB (1988). For regression analysis there is one dependent variable, which is uncontrolled during the experience, that depends of one or more independent variables that are measured with negligible error and are often controlled during the test (Ramakrishnan and Sivakumar, 1999). Figure 26 shows a typical S-N diagram for concrete where a regression analysis with confidence limits has been applied. The selection of the number of cycles to failure or the stress level for independent variable leads in general to different regression equations. Therefore, the choice of the appropriate independent variable is very important for the establishment of accurate regression models. Nevertheless, the correlation coefficient, r, which is a measure of the dependence between the variables (r?=1 means perfect correlation and 7?=0 means absence of any correlation) is independent of the choice of S or NV for independent variable. A power law (log(V) = 4+ B*log(S)) or a logarithmic law (log(N) = 4+B*S) is recommended to obtain a best-fit relation (4 and B are best fit parameters), see Ramakrishnan et al, (1996). The power law has the interesting property of when S tends to zero N tends to infinite. On the contrary, the logarithmic law tends to a finite value when S is zero and thus, applies just to a certain range of S values. ‘A question that emerges when building a Wholer diagram is the choice of the number of cycles to failure for the run-out specimens. If they were omitted from the Fatigue of plain and fibre-reinforced concrete 55 analysis, the regression lines would underestimate the number of cycles to failure. The other approach is to include the run-outs in the analysis considering the maximum number of cycles as their fatigue life. In this case, safer S-N curves are established because these specimens actually are able to withstand a larger number of cycles. Therefore, the common procedure is to include the run-outs in the analysis. 1.0 T le 90% Confidence limit 08 Zo g a g 06 2 a o4 80% Confidence limit 02 > Run-out specimens o 1 2 5 4 5 Fatiaue life floa(N)I Figure 26. Typical S-N diagram for concrete. The best-fit relation and the 90% confidence limits are also presented (Cornelissen, 1984). A different concept is often applied to build an S-N diagram and includes the failure probability, P, in the curve leading to an S-N-P diagram. The S-N diagrams based on these two procedures are often erroneously assumed equivalent but, actually, they are not comparable (RILEM, 1984). If at each stress level, the specimens are ranked in an increasing order of number of cycles to failure then, the probability of failure, at a number of cycles equal to or smaller than each value of NV is calculated by dividing the rank, r, of each specimen by i +1, in which i is the total number of specimens tested at a given stress level: r a 3.7) i+] GP This procedure avoids the possibility of the specimen with the largest number of cycles to have a failure probability equal to one. When the run-out specimens are included, the failure probability is given by (Koyanagawa et al., 1994) (3.8) 56 Chapter 3 where the run-out specimens are included in the number of specimens. The survival probability, Z, is accordingly given by L=1-P. G9) The two-parameter Weibull distribution function is often used in fatigue analysis of conerete (Oh, 1986; Shi ef al., 1993; Kobayashi and Hamada, 1998). For this function the survival probability reads 1=Av>n)-enf-(*)} 3.10) where a and vare parameters to be determined from data, With the distribution function determined at each stress level is then possible to determine for a certain survival probability the number of cycles to failure. An S-N relation form must be chosen and, for each probability, a regression equation must be established. The Jog-normal distribution function is also used in fatigue data analysis (ASCE, 1982b; Koyanagawa ef al., 1994). However, the respective hazard function (that is the probability that the specimen will fail in time interval ¢ + Ar on the condition that it has survived until time ‘) decreases with increasing fatigue life. This violates the physical phenomenon of progressive deterioration of materials resulting from fatigue process (ASCE, 1982b). An alternative process is the use of the McCall model that has been used successfully to predict the fatigue life of concrete (Do ef al., 1993; Kobayashi and Hamada, 1998) and reads log(L) = -aS*(log(N))" , G.11) where a, 6 and c are parameters obtained from data. 3.2.5 Deformations The evolution of deformations during a fatigue test, known as the cyclic creep curve, is schematically illustrated in Figure 27. Typically, three different stages can be observed. Initially, up to 10%~15% of the fatigue life of the specimen, a fast increase of deformation occurs with decreasing rate. This stage corresponds to a consolidation (in compression) or a relaxation (in tension) of the concrete specimen. Then, a second stage occurs where Fatigue of plain and fibre-reinforced concrete 37 deformations growth at a constant rate during most time of the test. Finally, during the last cycles (the last 10%~15% of the fatigue life) a rapid increase in deformation, with increasing rate leads to the failure of the specimen. \deldt 1 Secondary branch Deformation, ¢ Time, t Figure 27. Typical cyclic creep curve for concrete. It can be observed (Cornelissen, 1984; Hordijk er al., 1995) that the slope of the secondary branch, the secondary creep rate, increases as the number of cycles to failure decreases. Furthermore, if these variables are plotted on a log-log scale a strong correlation can be found between them. Thus, the relation between the secondary creep rate, ¢=de/dr, and the fatigue life can be expressed as 3.12) where a and 6 are constants calculated from the fatigue data. Since is constant, integration of Equation (3.12) over the entire test leads to the following deformation occurring during a fatigue test Ae=aN'*. (3.13) The values of the parameter 6 calculated by several authors range from 0.935 (Alliche, 1988) to 1.123 (Comelissen, 1984). Because b is close to unity, the quantity 1-b is close to zero and then, the value of Ae is almost constant for all specimens with the same characteristics. Thus, it seems that an intrinsic fatigue deformation capacity exists in concrete. During a fatigue test, when all the fatigue deformation capacity is dissipated, an unstable crack propagation occurs and the specimen fails. 58 Chapter 3 3.2.6 Damage accumulation Most of what has been previously written applies to constant amplitude loading Nevertheless, in actual structures the load cycles vary greatly in magnitude number and order. The simplest hypothesis to consider this multi-stage loading is known as the Palmgren-Miner (P-M) rule. It supposes that damage accumulates linearly with the number of cycles applied at a particular level and reads Da/N)=1, G14) in which n, is the number of load stages and N, is the number of cycles at a particular level. From multi-stage block loading tests, both conservative and unsafe predictions based on Palmgren-Miner rule have been reported (CEB, 1988). However, small variations of about 4% in stress calculations may lead to P-M sums of 0.38 and 13.87. It is also noted that P-M sum is dependent on the load counting method. Regardless of the simplicity of P-M hypothesis it is not always conservative (Shah, 1984). Additionally, there are evidences that fatigue failure of concrete is greatly influenced by the magnitude and sequence of applied fatigue loads (Oh, 1991) and that the increase of damage with cyclic loading is highly non-linear (CEB, 1988; Shah, 1984). Likewise, Grzybowski and Meyer (1993) state that for stress levels beyond 0.75 damage is increasingly non-linear both for plain and fibre-reinforced concrete. 3.3. Fatigue of concrete The fatigue mechanism of concrete is very complex. It may be attributed to progressive deterioration of the bond between the coarse aggregate and the cement paste and to the progression of existing micro-cracks and formation of new ones in the cement paste (RILEM, 1984; CEB, 1988). In fact, due to shrinkage, concrete has numerous micro-cracks even before any load is applied to it. These micro-cracks exist in the interface zone between coarse aggregate and cement paste and in the cement paste itself. The orientation and length of the micro-cracks are randomly distributed in concrete. Thus, a self- equilibrated stress field exists in concrete even if no extemal loads have yet been applied. The applied loads will create a local state of stress that, together with the pre-existing stress Fatigue of plain and fibre-reinforced concrete 39 field, will control the evolution of the micro-cracks and, consequently, of the fatigue performance of concrete. During fatigue loading, two opposing effects can be observed in concrete (Alliche, 1988). On one side, due to internal structure rearrangements a consolidation of the material can occur. On the other, the applied loads induce an increasing damage on the material During the initial part of a test, the consolidation prevails over the damage, which cause some hardening effects, while the opposite can be seen in the final part where the damage prevails over the consolidation. The presence of fibres has a double effect on the fatigue mechanism of concrete. First, by bridging cracks, they have a beneficial effect and secondly, by introducing additional defects they have a detrimental effect on fatigue life. Thus, it is from this competing mechanism that a beneficial, null or detrimental effect may be observed in the fatigue life of fibre-reinforced concrete compared with plain concrete. 3.3.1, Bending fatigue Flexural tests are often used to assess the tensile behaviour of concrete. The main advantage is that they tie up well with loading conditions encountered in practice. The main drawback is that stress distribution at the cross-section is not sufficiently accurately known, Nevertheless, and because they are also relatively simple to carry out, the bending tests are often preferred for the purpose. The bending fatigue tests can be divided into two general types. The first type and the most widely used, flexural tensile fatigue, is characterised by the fact that the bottom extreme fibre is subjected only to tensile stresses and therefore the bending moment acts only in one direction. In the other type of tests, flexural tensile-compressive fatigue, the bending moment acts alternately in one direction and in the other creating at the bottom extreme fibre alternating tensile and compressive stresses. ‘The geometry and layout used by several researchers for bending fatigue tests can be found in Table 12. It comes out from the observation of Table 12 that a wide range of geometry, loading apparatus and loading frequencies has been used, making direct comparisons difficult to be established. Another difficulty arises from the previously mentioned fact that the test procedures for determine the monotonic strength are not always the some with a direct influence on the calculated Whéler diagram. 60 Chapter 3 Table 12. Assessment of some bending fatigue tests layout. Geometry LxBxH Span Frequency Author Year Omens a = Observations Hatt, Crepps 1922 T62xt02x102 : 0.166 eee Williams 1943 826x102x130 ; 0.25 Soman, 18) Mecall 1958 368476x76 : 300 ce) Batson 1972 258xl01x182 81-81-81 30 R=00,-1 Raithby 1974 S0Bx102x102 152-152-152 200 (Comelissen, 1984) 1400-700 Comelissen loss 2300x150x280 1400-700, 80 Ramakrishnan 1987 35Sxl02x102. 102-102-102 200 on 1986 $00x100x100 200-100-200 42 Ramakrishnan 1991 S25x1S0xIS0 150-150-150 20.0 Alliche 1992 S60xI40x140 250250 1.0,3.0 1.0, $> 0.85 Shi 1993 Sooxigox100 150-150-150 45%) S085 Koyanagawa 1994 $30x1S0x150 150-150-150 50 5-7,8>08 Wei 1994 ssoxisoxiso —1so-1so150 S782 OR San = 0.1 Trichés 1998 $00x150x1S0 150-150-150 50 Kobayashi 1998 $30x1S0x1S0 150-150-150 50 Jun 1998 380xS0x100 170-170 20 Ramakrishnan 1999 _406x76x102__102-102-102 20.0 San 0.1 The evolution of crack growth under fatigue loading may be studied by means of pre-notched specimens. In this case, an even bigger difference occurs between the different testing procedures and specimen geometry, see Table 13. The bending fatigue behaviour of concrete is highly dependent on the minimum stress level or, in the case of reversing fatigue, on the level of the maximum compressive stress level, see Comelissen (1984). A reduction of the fatigue life occurs for lower Revalues. In case of alternating compression-tension (negative values of R), the reduction of the fatigue life with the decreasing of R is even more explicit. This effect is more pronounced for stress levels below 0.8. That can be explained by the fact that for higher stress levels the cracking behaviour is so dominated by the tensile stress that the interaction with the small compressive cracks is of only minor importance. To simulate the compression zone of a beam load, Ople and Hulsbos (RILEM, 1984) applied eccentric loads in compressive fatigue tests. They found that if the maximum stress level was defined as the ratio of the extreme fibre stress to the monotonic compressive Fatigue of plain and fibre-reinforced concrete 61 strength, the fatigue life of concrete increases with increasing eccentricity. However, if the stress level is expressed in terms of monotonic strength by corresponding eccentricity, the fatigue lives almost coincide. Table 13, Assessment of bending fatigue tests layout with notched specimens. Cooney 5 Frequene Author Year LxBxH ae eye) Notch Observations 3 (mm) (Hz) (mm?) Perdlars 1986 9poso.160 240no240 05-10 ws Sun 0? 95x38x38 47-47 eae Bazant 1991 190x38x76 95.95 0,033~0.040H/6 sgenos 318x38x127 159-159 102x38x38 47-47 5th Bazant 1993 288x38x105 131-131 100°. Hl ates 814x38x305. 381-381 ee (Ramakrishnan, 1996) Naaman 1993 400x100x100 100-100-100 1.0-50 Sain = 0.10 Snax = 0.7, 0.8,0.9 The comparison between concentric tension and flexural tension show that the fatigue strength of concrete loaded in bending is bigger than that of concrete loaded in tension (Comelissen, 1984). Comelissen sustains that a stress relaxation occurs, as in a creep test, in the fibres that undergo the greatest amount of strain. This relaxation causes a redistribution of stress in which the fibres located further inwards from the extreme fibre increasingly participate in carrying the load. This effect is more pronounced as the stress gradient increases or, in case of flexural load, as the depth of the beam decreases. The effect of the stress gradient is recognised by the CEB Model Code 90 that considers it by means of a correction factor, 7, that must be multiplied by the maximum stress in the compression zone: 15 (15-05(4/2)), (3.15) where 6, is the lower absolute value of the compressive stress within a distance no more than 300 mm from the surface and 0,2 is the larger absolute value of the compressive stress under the same situation. 62 Chapter 3 Effects of fibres The effect of the fibre addition in the fatigue life and endurance limit of concrete is a very complicated phenomenon. Along with the complexity of the fatigue mechanism itself, there is a variety of factors affecting the performance of fibre-reinforced concrete such as the fibre material, the fibre geometry and the fibre content. Thus, it is not simple to say if fibres improve, stabilise or reduce the fatigue performance of concrete. The fracture resistance mechanism of fibre-reinforced concrete can be explained by three stages. Initially, exists a sub-critical crack growth in the matrix and the fibre bridging effect begins. Then, there is a post-critical crack growth in the matrix such that the net stress intensity-factor due to applied load and fibre bridging stresses remains constant. Finally, a last stage is observed where the resistance to crack widening is provided exclusively by fibres (Jeng and Shah, 1986). Thus, it is possible that after cracking the specimen can still carry the applied load with increased deformations. This can be seen in Figure 28 that represents the evolution of the maximum and minimum deformation during a fatigue test of steel fibre-reinforced concrete. About cycle 7000, a crack open and the displacements start immediately to increase. Due to the presence of fibres, the crack growth process stabilises and failure occurs about cycle 63000. Displacement (mm) © — 10000 20000 30000 40000 50000 60000 70000 Number of cycles, Figure 28. Cyclic creep curve for fibre-reinforced concrete specimen under flexural load. A number of investigations have been made about the performance of fibre-reinforced concrete. The differences between the test procedures and the types of fibres make comparisons difficult to be made. Ina study comparing various percentages of steel fibres cut from a low-carbon sheet, Wei ef al. (1996) stated that the key to increase the fatigue resistance of high-strength Fatigue of plain and fibre-reinforced concrete 63 concrete is the increase of crack-arresting ability. This can be achieved by reducing the size and amount of original crack sources and increasing the capacity of inhibiting initiation and extension of cracks. They achieve these goals by adding silica fume and fibres to concrete. The compared effect on the fatigue behaviour of several types of fibres was studied by Ramakrishnan and co-workers (Ramakrishnan ef al., 1987a, 1987b; Ramakrishnan and Lokvik, 1991). In Figure 29 and Figure 30, several design S-N curves obtained from test data for various types of fibres (type A: steel hooked-end fibres; type B: steel straight fibres; type C: steel corrugated fibres; type D: polypropylene fibres) are plotted. It can be seen that all fibres increase the endurance limit of concrete. However, for the fibre volume of 0.5 percent, above the stress level of 0.75 all fibres exhibit smaller fatigue lives then plain concrete. For the fibre volume of 1.0 percent, only the hooked-end steel fibres have bigger fatigue life and endurance limit than plain concrete. o.g| Stress level 06 ol 0 4 Fatigue life {log(N)] Figure 29. Wholer diagrams for several types of fibre-reinforced concrete with a fibre volume of 0.5% (Ramakrishnan and Lokvik, 1991). An interesting discussion is related with a so-called optimum fibre content beyond what fibres do not improve the fatigue behaviour of concrete. In fact, Jun and Stang (1998) for bending fatigue report a reduction of the effectiveness of the addition of fibres when the fibre volume changes from 1% to 2%. In the last case, the fatigue performance of plain and fibre reinforced concrete is almost the same, Additionally, Grzybowski and Meyer (1993) 64 Chapter 3 testing fibre contents of 0.25%, 0.5%, 0.75% and 1.0% under compressive fatigue loading found that for fibre contents beyond 0.25% exists a reduction of the fatigue life of concrete. These apparently contradictory results alert for the fact that the improvement of the fatigue life of fibre-reinforced concrete under fatigue loading compared with plain conerete is not a straightforward question. 08 Stress level 06 04 0 1 2 3 4 5 6 7 Fatigue life [log(N)} Figure 30. — Whiler diagrams for several types of fibre-reinforced concrete with a fibre volume of 1.0% (Ramakrishnan and Lokvik, 1991). Until now, the fatigue performance of steel fibre-reinforced concrete has been presented as an intrinsic material property, what means that the Whéler diagrams have been constructed dividing the applied stress by the corresponding monotonic strength. Nevertheless, it has been already said that the monotonic strength of fibre-reinforced concrete is usually higher than that of plain concrete. Thus, if the stress level of fibre-reinforced concrete is calculated in relation to the monotonic strength of plain concrete, a shift to the right of the S-N diagram will occur. This means that if the properties of a particular concrete composition are known then almost certainly the addition of fibres will improve the fatigue life of that particular mix. However, the strength gain caused by the fibres must not be taken into account in the design of the structure. Fatigue of plain and fibre-reinforced concrete 65 Bending fatigue behaviour of other concrete types The fatigue behaviour of roller-compacted concrete is very dependent on the cement content. A ten-time increasing on fatigue life (log(V) +1) was observed when the cement content was increased from 120 kg/m? to 280 kg/m? (Trichés, 1998). For these higher cement contents, which in fact are close to those used in RCP, the fatigue life of roller- compacted concrete seem to be shorter than that of normal concrete (Inoue and Sawa, 1990; Kobayashi and Hamada, 1998). For concrete made with recycled aggregates, it seems that the fatigue life of roller-compacted concrete was similar to that of normal concrete (Kobayashi and Hamada, 1998). The fatigue life of concrete made with recycled aggregates was found worst than that of concrete made with normal aggregates (Kobayashi and Hamada, 1998). Moreover, it seems that the quality of the recycled aggregate type plays an important role in the fatigue life especially for stress levels above 0.7. 3.3.2 Compressive fatigue Most of what have been said regarding the behaviour of concrete in bending is also applicable for the compressive behaviour of concrete. The results of Tepfers and co- workers (Tepfers, 1979; Tepfers and Kutti, 1979) indicate that the Whdler-curves for tensile and compressive fatigue of concrete are identical. This behaviour is confirmed by the design curves proposed by the CEB Model Code 90 (CEB, 1990) that uses the same curves when Smin is zero. The fatigue strength of plain concrete depends on the maximum as well as on the minimum stress in the cycle. In contrast to steel, no fatigue limit of concrete has been found up to now. A quasi-fatigue strength corresponding to S= 0.4 has been suggested by Klausen and Weigler (1979). Similarly to what happens in flexural fatigue tests, the test procedures used by researchers are very distinct (Table 14). The applied load frequencies range from 0.1 Hz to 200 Hz, although the frequency of 1 Hz seems to be the most common one. However, testing at 1 Hz frequency takes a lot of time and do not appear to be very reasonable for high-cycle range (one million cycles will last for 11 days). Testing at constant R-values or at constant Siin-values is also a subject that must be clarified. Thus, the need for standard ‘test procedures is mandatory so that results from different researchers could be more easily compared. 66 Chapter 3 Table 14. Assessment of some compressive fatigue tests layout. aver es Geometry Frequency ; = Year cylinders (xd); prisms (BxBxH) (Hz) pacdiacaed 05,82 085 Soon = 02 shore ie Ses00 3,5<0.85 Ordinary and Lightweight Sn ~ 0.05, 0.2, 0.4 Klausen, Weigler 1979 soxi00 041,110,200, an ee Tepfers 1984 150x300 125 San = 02 Nelson 1988 127x12.7%127 1 BiaxialHigh-strength su, Yin 1988 152x38x152 1 Biaxial, Fibres Sin = 0.05 Do 1993, 100x200 1 eens: Greybowski 1993 tozx102x102 1 Fibres Sn = 0.95 Hordijk 1995 100x100x250 6 HORT ve Taliercio 1996 100x200 1 Multiaxal Paskova 1997 1o2xt02x102 1 Fibres Gao 1998 100x100x300 4 R=0.167 ‘Note: all dimensions are in millimetres. Effects of fibres The effect of fibres on the compressive fatigue behaviour of concrete leads to a competitive mechanism between the beneficial crack-bridging effect and the detrimental initial micro-cracking increasing. Thus, it is not easy to predict the effect of fibres on the compressive fatigue behaviour of concrete, Otter and Naaman (1988) compared the behaviour of plain and fibre concrete under cyclic loading and concluded that the envelope curve concept applies to both types of material. Moreover, for practical purposes, the envelope curve may be taken as the monotonic stress-strain curve. They also stated that the primary influence of fibres is on the monotonic behaviour and envelope curve, especially in the post-peak region. The tests of Yin and Hsu (1995) show that under uniaxial compression, the effect of fibres is negligible in the low-cycle region and in the determination of the endurance limit (Figure 31). The tested specimens have 152x152x38 mm and load cycles were applied at a frequency of one cycle per second. The used fibres are straight, slit-type with dimensions 25x0.01x0.022 mm? in a volume percentage of one percent. Paskova and Meyer (1997) found that fibres have a considerable beneficial effect on the fatigue life of concrete for all volume percentages of fibres. Tests were conducted over 102-mm cubes with a frequency of 1 Hz. Hooked-end fibres at volume percentages of 0.0, 0.25, 0.5, 0.75 and 1.0 percent have been used. The maximum stress level ranged from 80 Fatigue of plain and fibre-reinforced concrete 67 to 95 percent. Regardless of the fatigue strength found for every volume percentages of fibres, it was observed that for the a volume percentage of 0.75 percent a decrease of fatigue strength compared with all other volume percentages (Figure 32). Stress level ‘Outten ae snd eh Sehr Fatigue life log(N)] Figure 31. Whdler-diagram for plain and fibre concrete under uniaxial compression fatigue according to the experiments of Yin and Hsu (1995). A different conclusion was observed by Grzybowski and Meyer (1993) few years before. They observe a reduction of the fatigue life of fibre concrete compared with that of plain concrete for fibre volume percentages above 0.25 percent. 1000 100 Fatigue life 10 00 025 05 075 10 Fibre content (%) Figure 32. Number of cycles to failure as a function of steel fibre content (Paskova and Meyer, 1997). 68 Chapter 3 These results show that there is no direct relation between the volume percentage of fibres and the concrete fatigue life. Moreover, variables like the fibre type and aggregate grading may also have an important role on the fatigue behaviour of steel fibre-reinforced concrete. Multiaxial behaviour Only a few multiaxial compressive fatigue tests have been reported so far. Moreover, the available results report different conclusions. The results of the tests of Suand Hsu (1988) indicate that biaxial stress states increase the fatigue strength of concrete. The tested specimens have 152x152x38 mm? and were tested at a frequency of 1 Hz. Moreover, for all biaxial and some uniaxial fatigue tests, failure occurs by tensile splitting, with the fracture surfaces perpendicular to the direction of the maximum tensile strain (Figure 33, type 2). Under uniaxial loading, most of the fractures occur by formation of cracks in the direction of loading and perpendicular to the plane of test specimen (Figure 33, type 1). VO Figure 33. Observed failure modes in biaxial compression fatigue (principal loading direction is direction one). In a similar testing scheme, Nelson ef al. (1988) found that high-strength concrete specimens tested at maximum stress levels below 75 percent, failure occurred at a higher number of cycles as the biaxial stress ratio decreased. Further, they found that the biaxial state of stress did not enhance the fatigue strength of concrete below maximum stress levels of approximately 75 percent. Hsu and his co-workers (Yin and Hsu, 1995) also study the behaviour of steel fibre-reinforced concrete under biaxial loading states. The results show that the fatigue strength of fibre concrete is greater in biaxial compression than in uniaxial compression. They also found that the addition of fibres those not increase the endurance limit of concrete. Under biaxial compression, fibres are just effective for the low-cycle region (Figure 34), depending on the principal stress ratios. Fatigue of plain and fibre-reinforced concrete 69 A different type of multiaxial fatigue tests has been done by Taliercio and Gobbi (1996). They tested concrete cylinders at a frequency of 1 Hz under triaxial loading conditions. The influence of the ratio, r, and phase angle, g, between the confining pressure and axial loads has been studied. They conclude that fatigue life is shorter in tests where axial stress and confining pressure were in phase opposition rather in phase coincidence. Moreover, fatigue life increases as the mean lateral stress increases. o2/o,=1 Stress level 6 08 os | ° Fibre concrete © — = Plain concrete 04 0 1 2 3 4 5 6 7 Fatigue life [log(N)] Figure 34. Whdler-diagram for plain and fibre concrete under biaxial compression fatigue according to the experiments of Yin and Hsu (1995). 3.4 Design of concrete structures to fatigue loads The nature of fatigue implies that the cumulative damage due to all the load fluctuations that occurred during the service life of a structure may be ultimately responsible for its collapse. The fact that fatigue is caused by service loads but it may cause collapse makes the limit state of fatigue difficult to be defined. Therefore, in codes of practice fatigue is treated in three different ways (CEB, 1988): 70 Chapter 3 i) asa limit state in addition to the ultimate limit state and serviceability limit state; ii) as a serviceability limit state (type 1) based on stress range limitation in addition to a serviceability limit state of cracking and deformation (type II); iii)as a section in the chapter of ultimate limit state. ‘The range of application of fatigue design is limited to a small number of structures such as railway bridges, sea structures, highways and major roads and all structures where the imposed load is predominantly of cyclic character. The aim of fatigue design or fatigue verification is to avoid the risk of fatigue failure or avoid a decrease of safety of a structure or its elements (considering endurance limits of the materials), due to fatigue or service loads. 3.4.1 Load spectrum Fatigue evaluation of concrete structures demands an understanding of the nature of loading forming the excitation to the structural system and a quantitative description of that loading. In many situations, the loading causing fatigue damage is influenced to a considerable extent by the dynamic response of the structural system. Interactions of this type can cause considerable difficulty in the assessment of fatigue damage, particularly where the loading is quantified in a way that involves both the physical phenomenon and the response (CEB, 1988), In cases with superimposed loads due to different actions, such as traffic and temperature, it is necessary to verify if whether they are correlated in time or not. If they are correlated, the corresponding stresses should be added; if not, the damages can be added separately. The stresses induced by the effect of various intensities of the same load or different types of loads must be separated into discrete blocks that cause the same stresses in the structure. The stress blocks have associated with them a certain number of load repetitions that should be checked against fatigue failure. Damage accumulation for design purposes is usually considered by means of P-M sum, Equation (3.14), but where the total damage accumulated may be different from unity: X(a/N,)=D, (3.16) Fatigue of plain and fibre-reinforced concrete 1 where D is the maximum accumulated damage. The value of D found in several codes of practice range from 0.2 to 1.0 which does not mean that a reduced safety factor is employed when D=1. Thus, the safety factor could not be interpreted as 1/D (see Siemes, 1988). 3.4.2 Material characteristics The use of P-M sum for damage accumulation implies that linear material behaviour is assumed so that superposition of effects could be done. For concrete, the material characteristics are taken into consideration by means of S-N diagrams determined by tests. The CEB Model Code 90 propose the S-N relations presented in Table 15 for the range 0 6, and as, 203-0375. tog(¥,)=l0g(¥,)(03-03755,..)/AS,, log(N,)>6, and as, < 03-037: ‘Compression-tension with Sinan $0026, nu log(V) =9(1-S.au) Pure tension and Compression-tension with tog(N)=12(1-S,an) Same? 0.0296) |/ Faas nun isthe maximum compressivestress loi! Soa.co> Semin isthe minimum compressive tress |) Foe min? Taam ithe maximum tensilestres Sate 085Bf4(\~ Se/25fa.)) fay = \OMPa,Bisacoeficient depending onconcreteage J isthe characteristic compressive strength n Chapter 3 It should be noted that independently of the S-N diagram used the procedures are very similar. However, there are some differences regarding the existence of a fatigue limit and the influence of the frequency of loading. Some design codes also take into consideration the immersion of concrete in water. 3.5 Summary The fatigue behaviour of plain and steel fibre-reinforced concrete has been analysed in this chapter. Initially, the basic characteristics of fibre-reinforced concrete under bending, tension and compression have been presented. It has been seen that the major importance of fibres is in the post-peak behaviour increasing enormously the fracture energy of conerete. For the volume fibre percentages used in current practice, there is almost no strength improvement by the addition of fibres. The essential aspects of concrete fatigue have been presented subsequently. Emphasis on the evolution of the properties of concrete during fatigue testing has been taken. Moreover, the most used testing schemes have been presented as well as some procedures for modelling fatigue data due its large scatter. Flexural and compressive fatigue behaviour of concrete, for both plain and fibre- reinforced concrete, were assessed. The effect of fibres on the fatigue performance of concrete has been discussed. A competing mechanism between the initial defects caused by fibres and the crack-bridging capacity of fibres difficult the reaching of definite conclusions, Nevertheless, a tendency for the improvement of the fatigue behaviour of conerete due to the presence of fibres has been majority reported. It has been seen that there are no fatigue testing standards, which obscures the comparison of available fatigue data. Finally, the chapter describes the basic procedures for fatigue design of concrete structures. The essential parameters for fatigue design (load spectrum and material characteristics) have been presented in a condensed manner. Compressive fatigue tests B 4 COMPRESSIVE FATIGUE TESTS In the preceding chapter, the main aspects of the behaviour under fatigue loading of plain and fibre concrete has been presented. It has been seen that there is a lack of information regarding the evolution of deformations during fatigue loading, which is very important for constitutive relations modelling. An experimental research program has been carried out in Laboratory of Structures of the Faculty of Engineering of the University of Porto (LEUP). These experiments have two main objectives. First, it is intended to gain insight into the behaviour of plain and fibre concrete under fatigue loading and secondly the results of the experiments should provide reliable data for the calibration of the developed numerical model. Thus, the test program was designed to compare the performance of both types of concrete and to evaluate the relevant parameters that must be incorporated into the model. Although only a limited number of tests have been carried out, they seem to be sufficient for the intended objectives. Moreover, the fibre content and the mixing procedure used were planned to be similar to those used in practice. Two types of fatigue tests have been carried out namely compressive fatigue tests on cylinders and bending tests on prisms. Both types of tests were load-controlled between to predetermined stress levels. This chapter presents the results of the compressive fatigue tests. The discussion of the results will be mainly focused in the obtained Whéler diagrams, evolution of deformations during the tests, and final fatigue deformations. ‘Additionally, the post-fatigue properties of run-out specimens will also be presented and discussed. It should be kept in mind that the objective of this study was not the detection of “absolute” material laws but just the contribution for a better understanding of the behaviour of plain and fibre concrete under fatigue loading. 4.1. Materials and equipment This study is a continuation of a previous study carried out at LEUP by Barros (1995) where the effect of fibre reinforcement on the behaviour of concrete under monotonic loads has been studied. For this reason, the concrete composition and fibre content used were similar to those used in that study. 14 Chapter 4 It was the first time that fatigue tests have been performed in LEUP, which originated some difficulties in the setup of the test program and delayed the beginning of the first tests. To execute the experimental program, it was necessary to modify the equipment used in LEUP and acquire some equipment to perform the tests what have been done during the current investigation. It was also necessary to define the test procedures and data acquisition schemes so that reliable information could be attained with the experiments. 4.1.1 Fibres The most important improvement attained by the addition of fibres to concrete is the increase in toughness and ductility. Nevertheless, it is also possible to improve other properties such as the strength and stiffness. In practical applications, however, fibres are added in percentages about 0.5 percent in volume (about 45 kg/m’) for which is unlikely to obtain higher strengths. In pavements, the use of long steel fibres (60-mm length) has been recommended since their length ensures a better anchorage in the concrete. Hooked-end and twin-cone fibres seem to be the most effective since they provide the best anchorage to concrete. Following the previous work carried out at the LEUP (Barros, 1995) and keeping in mind the proposed objectives it was decided to use a fibre content of 45 kg/m’. This roughly corresponds to a percentage in volume of 0.56 percent. Following the producers’ instructions (Bekaert, 1990) it was also decided to utilise long fibres with 60-mm length Several authors show that the addition of these fibres greatly improves the performance of concrete slabs resting on an elastic foundation under ultimate load conditions (Barros, 1995 for soils and Falkner and Teutsch, 1993, for rubber and cork), However, see Chapter 2, for concrete road pavements the major cause of failure is fatigue loading and not the existence of large isolated loads. Accordingly, it is then necessary to study the behaviour of concrete reinforced with these fibres under fatigue loading conditions. The used fibres are Dramix ZC60/.80 and ZP30/.50. These fibres are glued into bundles (Figure 35) so that a better separation of fibres could be attained during the mixing process. The main geometric properties of these fibres are summarised in Table 16. The addition of fibres to concrete reduces its workability and requires more compaction efforts to avoid lowering the concrete quality. As a relatively reduced Compressive fatigue tests 5 water/cement ratio of 0.38 was used, a plasticiser has to be added to the mix to maintain concrete workability. To limit the number of variables analysed the only difference in the concrete composition between plain and fibre concrete is the existence of fibres Figure 35. Fibres used in this study. Table 16. Main geometric properties of the fibres used in this study. Fibre Length (mm) Diameter (mm) ‘Aspect ratio ZC60/.80 60 0.80 7 Z30/.50 30. 0.50 60 4.1.2 Concrete The material characteristics of the concrete components and the concrete composition are summarised in Table 17. The concrete composition has a water-cement ratio of 0.38 and a percentage of fibres of 1.85 percent in weight and 0.56 percent in volume. Table 17. Material characteristics and concrete composition. Materials ‘Characteristics ‘Composition (kg/m*) Cement Portland Type Il, Class 32.5 450.0 Sand (0/3 mm) : 716.0 Aggregate (0/5 mm) : 933.0 Aggregate (5/15 mm) : 533.0 Water - 171.0 Fibres Dramix ZC60/.80, ZP30/.50 45.0 Plasticiser Rheobuild 561 45 The mixing procedure used in the manufactured of concrete was as follows The aggregates (blue granite with maximum dimension of 15 mm) and the sand were washed 16 Chapter 4 and dried before the process starts. The constituents were introduced in the following order: water, cement, fibres, aggregates and sand, A small mixer with 200-dm’ capacity was used, Compaction of concrete in the moulds was made with a vibrator needle. The specimens remain in the moulds for seven days covered by humid tissues after what they were hardened at the natural environment of the laboratory (approximately 20° C of temperature and 70% of relative humidity). 4.1.3 Equipment ‘The equipment used in the current research program is constituted by a load frame, an actuator, a hydraulic group, a digital controller, a load unit control and a computer. The load frame and the actuator are loaded by a hydraulic actuator provided by the hydraulic group. The digital controller is the interface between the software and the rest of the system. The command given to the system and the test procedures are defined by adequate software. The load unit control allows manual control of the load frame and of the actuator, which is essential to place the specimens into position. The equipment is controlled by an automatic control cycle in which the control equipment (the computer and the digital controller) gave a command signal to the servovalve and hydraulic actuator. The control sensor sends back a return signal to the digital controller indicating the system response. The digital controller reacts to the difference between the command and the return signal and activates the hydraulic actuator in order to correct that difference (Figure 36). The load frame is from series 315 of MTS® and is loaded by a hydraulic actuator with vertical displacement. The capacity of the system is 2700 kN with a maximum displacement of 100 mm. It is also calibrated to be used with a maximum load of 270 kN which is useful when smaller loads are necessary in the tests. The digital controller allows that the tests could be carried out under displacement or load control. Compressive fatigue tests Return Signal Dieta Contr aurouane crc 4 Control sigral| Hidrautie pressure Figure 36, Higralie Actator Automatic control cycle of the load equipment. Figure 37, ‘View of the actuator (left) and of the load frame (right). The actuator has a capacity of 250 kN with a maximum displacement of 200 mm. It has a spherical hinge at both extremities. It is also possible to use the actuator with sensitivity up to 25 KN. 78 Chapter 4 4.2. Test procedure The search for construction materials with enhanced properties such as strength, ductility, toughness and durability has lead to an increasing interest on materials like fibre reinforced concrete or high performance concrete. The little knowledge about the long-term behaviour or the effects of repeated loading on the properties of these materials has caused a growing interest on the fatigue performance of concrete. Additionally, accurate models capable of simulate this long-term behaviour are necessary and they need reliable data for their calibration. ‘The main purpose of this study is the investigation of the most significant properties of concrete (both plain and fibre reinforced) when subjected to compressive fatigue loading, so that a numerical model for concrete fatigue can be developed and calibrated. The experimental program consists of a series of compressive fatigue tests on plain and steel fibre-reinforced concrete. It is intended to compare the performance of both materials under fatigue loading. Batches of several specimens were made, from which some specimens were used to estimate the monotonic properties and the others were tested under fatigue loading. 4.2.1 Specimens Cylinders with 150-mm diameter and 300-mm height were used. The cylinder extremities were rectified to ensure a good contact surface between the specimen and the load platens since no interface material has been used. Two series of compressive fatigue tests have been carried out. In the first one, 60-mm fibres were used. Six batches of plain concrete (labelled P1 to P6) and three batches of fibre concrete (labelled L1 to L3) were tested. In the second series of tests, where 30-mm fibres have been used, a batch of 17 cylinders of plain concrete (P9) and a batch of 17 cylinders of fibre-reinforced concrete (S9) have been tested. Table 18 summarises the number of specimens in each batch, the specimens used to assess the monotonic properties and those used for fatigue loading. Compressive fatigue tests 79 Table 18. Characteristics of the compressive batches. Batch N. specimens: N. monotonic N fatigue (Observation PI 5 5 2 2-monotonic lost P2 6 3 3 I-lost P3, PA, PS, P6 6 2 4 ul 5 2 3 60-mm 12,13 6 2 4 60-mm P9 v7 5 12 $9 7 3 12 30-mm Although the number of fatigue tests may be limited for a comprehensive fatigue analysis (Paskova and Meyer, 1994), it seemed appropriate for the planned objectives 4.2.2 Monotonic tests The monotonic tests were carried out with displacement control with a speed of 10 um/s until a 3-mm displacement is reached and then, the speed is increased to 50 jim/s until the displacement reach 18 millimetres. The values of the force and the displacements were stored every 0.4 seconds. The axial displacement of the cylinders is measured between the load platens. The stress-strain response is obtained by dividing the force by the cylinder area and the displacement by the cylinder length. The attained stress-strain curve is initially concave upwards, which implies that a cut off of this concave part has to be made in order to get the initial tangent modulus (E.)). This, is done (Figure 38) fitting a straight line to the initial part of the downwards concavity of the stress-strain curve. 4.2.3 Fatigue tests The fatigue tests were carried out with force control between two limits (with a sinusoidal force variation in time). The tests were conducted under a constant-minimum stress level, Smin, Of 10% of the monotonic strength. The maximum stress level, S, ranges from 60% to 90% of the monotonic strength. Before the cyclic process beginning, the load is monotonically applied with displacement control (50 um/s for the 60-mm fibre tests and 15 uum/s for 30-mm fibres tests) until it reaches the reference stress level. The load frequency was 2.5 Hz. 80 Chapter 4 700 600. ——Data points Regression Force [kN] 02 0 02 04 06 Displacement [mm] Figure 38. Data acquisition with calculation of initial tangent modulus (£.). During the initial monotonic loading, the data acquisition process is similar to the procedure described earlier for the monotonic tests. During the cyclic process, data is acquired 10 times during each cycle what means that the force and the displacements are read every 0.04 seconds (Figure 39). Thus, it is impossible to keep information of the entire test due to tremendous amount of information that can be generated in a fatigue test. To overcome this situation, data has to be stored intermittently, for which several data acquisition procedures have been drawn, In the current fatigue tests, three different types of acquisition procedures have been drawn in order to keep information about the whole test. Force [kN] 8 8 8 g nie Meet te He s Time Figure 39. Evolution of load with time during a fatigue test (seven load cycles, data acquired every 0.04 seconds). First, the acquisition of the initial 500 cycles of the test was made. This allows the monitoring of any problems occurring in the beginning of the test. After that, data of four Compressive fatigue tests 81 cycles is acquired at intervals of 100, 500 or 2000 cycles accordingly to the load level These four cycles are then averaged to store a unique cycle (Figure 40). ex so ena ewe z fs = 0. oa ost ows ou ons Diepiacement (rm) Figure 40. Average procedure for data acquisition. Finally, the last 300 cycles were read and stored to get information about this crucial part of the test (Figure 41). 0s 07 08a Displacement (mm Figure 41. Example of the last cycles acquisition of a fatigue test. 4.3 Results of monotonic tests Because of some difficulties during the manufacturing of all specimens and during the setup of the experimental program, the tests of some batches start almost two years after they were made. This means that for some specimens the concrete has already experienced all the hardening that it can experience during its lifetime. The age of concrete is therefore, 2 Chapter 4 a very important aspect that differentiates the batches P1 to P6 and LI to L3 from batches P9 and $9. The list of the results of the monotonic tests is presented in Table 19. It can be observed that the presence of 60-mm fibres did not improve the compressive strength, while the 30-mm fibres have improved the compressive strength of concrete. However, the results did not mean that smaller fibres improve the monotonic strength of concrete. A small increasing around seven percent of the peak strain was achieved for the 60-mm fibres when compared with the corresponding plain concrete batches (from 2.29 %y to 2.46%). Moreover, for 30-mm fibres the increase of the peak strain was bigger from 2.92% to 3.52%, which corresponds to an increase of the peak strain around twenty percent. These findings are probably related with the maturation of concrete at the time of testing. In fact, in the 60-mm series, concrete is much older than in the 30-mm series which implies that for the former series the stiffiness was bigger. Additionally, some perturbations caused by the presence of the fibres, such as self-equilibrated stresses or small cracks or voids, may have less effect in well-hardened concrete. Table 19. Summary of results of monotonic compressive tests. Batch N. tests: Time (days) fom (N/mm?) E.(Nimm’) ep oo) PI 3 73S 51.00 E, P2 2 730 56.98 30600 2470 PS 2 703 45.06 29400 2.184 Pa 2 449 53.66 729000 2318 Ps 2 490 34.43 32600 2.227 P6 2 418 53.51 30000 2.254 ti 2 700 48.67 729400 2.542 2 2 640 51.49 28300 2.450 Ls 2 29 5857 27900 2.385 3 30 40.77 19200 3.000 P9 1 34 45.16 22600 2.834 L 39 42.97 21600 2917 3 34 48.98 720500 3.500 s9 1 “4 55.80 25300 3.426 1 54 48.80 20100 3.643 4.3.1 Stress-strain diagrams of monotonic tests The stress-strain diagrams of the monotonic tests are plotted on Figure 42, Figure 43 and Figure 44. These diagrams allow an easier visualisation of the previous results. Because of Compressive fatigue tests, 83 problems occurred during the acquisition process, it was only possible to save one stress-strain diagram of batch P1. Figure 42. Stress-strain diagrams for plain concrete. It can be seen that within the same batch, the difference between the two tested specimens is not significant in respect to the monotonic strength. Nevertheless, and although the manufacturing process and the materials were identical for all batches, the differences between the batches are sometimes important. Figure 43. Stress-strain diagrams for 60-mm fibre-reinforced concrete. A reason for that was probably related with the manufacturing process that was unable to guaranty the necessary similarity between the batches. The post-peak behaviour of plain and fibre concrete is quite different. In fact, plain conerete specimens exhibit a very steep branch after the peak load while fibre concrete displays a much more smooth descent. 84 Chapter 4 Po so ‘stress [MP3] sees 8 © ames oor ors ace ones oan Stain Figure 44, — Stress-strain diagrams for plain and 30-mm fibre-reinforced concrete. Another aspect that came out from observation of the previous charts is the fact that the post-peak branch of the batch P9 is somewhat less steep than that of batches P1~P6. Similar results are observed for the fibre-reinforced specimens. Again, this may be attributed to the maturation of concrete. 4.3.2. Relation between tangent modulus and compressive strength The relation between the initial tangent modulus and the compressive strength has been investigated since if such a relation exists it will be possible to assess the monotonic strength of a concrete specimen if the initial loading branch is known. This is important because in the fatigue-tests the load was monotonic applied before the load cycles begin and the ultimate strength may be assessed by means of the initial tangent modulus. The relation between these two properties is presented in Figure 45 and Figure 46, where the results of the control specimens for flexural fatigue have also been included. § 49. Plain concrete (Pt to P8) E ay, Fire concoo(L1 013) 235 2 35] 3 0. 3 204 32 Bs = Eo=0473t"em +4.3342 © 29 c= 06022%em - 4.0183 5 tS Re = 0.7203 bs FP = 0.8502 P10 ze ns ea 2 io : Fs 40 80 60 mF 30 0 50 6 7 ‘Compressive strength [MPa] Compressive strength MPe] Figure 45. _ Relation between tangent modulus and compressive strength for series P1 to P6 and LI to L3 It can be deduced from observation of Figure 45 and Figure 46 that good correlation exists between the initial tangent modulus and the compressive strength of concrete. For the Compressive fatigue tests 85 relation attained it was assumed that the age of concrete has no effect in the relation between the two variables. ge 40 © 35 Ec =0.7158"%em - 9.7363 © PS 3 P= 3 ® R? = 0.8321 ae, 3B x = 2 OP Oo nvem- 14.402 3 o RP = 0.9751 3 io 30 40 50 60 70 Compressive strength [MPa] Figure 46. Relation between initial tangent modulus and compressive strength for series P9 and $9. 4.4 Results of 60-mm length fibres fatigue tests One of the problems of showing fatigue data is that the real value of the strength of the specimen that is being tested is not known and must be inferred from other specimens of the batch. This is one of the causes of the large scatter existing in fatigue data because of the logarithmic relation existing between the stress level and the fatigue life of concrete. As an example, a variation from S=0.65 to S=0.6 may cause a variation of the fatigue life from 129000 cycles to 691000 cycles using the formula proposed by Tepfers (see Chapter 3). This means that a reduction of eight percent in the stress level may cause a variation of 530% in the fatigue life of concrete. 44.1 Stress level The S-N fatigue data related to the compressive fatigue tests is plotted in Table 20. The data is grouped by stress level and ordered by increasing number of cycles to failure. Three plain concrete specimens and two fibre-reinforced concrete specimens did not fail during the fatigue tests. Thus, they have been monotonically tested after the fatigue test has ended. 86 Chapter 4 Table 20. Compressive fatigue data. Plain concrete Fibre concrete ‘Specimen S ‘N_ Specimen S N Specimen S N P6D 0.85 41 P6A 07 48413 L3D 0.85 38 P4A, 0.85 291 PSB 0.7 100876 L3C 08s 192 PSD 08s 448 PSC 07 175556 L2A, 0.85 798 PSA 0.85 1504 PaD 07 281670 L2B 08s 3316 P3A, 0.85 1673 P6B 07 © 671475 LIC 0.85 10099 Pec 0.85 2561 _P4Ct 0.7 1000000 “LIA 0.70 7458 PAB 0.85 4923 “PIC® 0.65 1000000 L3A 0.70 10984 P3B. 0.85 21962__ P3D* 0.65 1000000 L3B 0.70 17542 P2A 0.80 3281 2D 0.70 163145 P2c 0.80 6148 Lace 0.70 1000000 P2B 0.80 8054 LiBt 0.70 1000000 PIB 0.80 17076 PIA 0.80 51064 Note: the specimens marked with * are run-outs. A regression analysis has been made with a power law used to fit the data. The regression equations fitted to the fatigue data read log(N) = 1.3528 -25.7033 log(S) , for plain concrete, and log(N) = 1.1529-24.1578 log(S) , 4.1) (4.2) for fibre reinforced concrete. In Table 21 the results of statistical tests used in the regression procedure are shown. Table 21. Statistical tests for compressive fatigue (log(N) = a + blog(S)). 1 ‘statistics Festatisties ao = values _evalue limit —“Fevalue limit, Plain 0615 0.774 25.7033. 3.1896 8.058 —«1.729~=*«CBCSCCH.N Fibre 0967 0572 -241578 6.9628_3.467__—«1.833——12.04 73.93 s;— standard deviation of variable i. The Whiler diagrams are shown in Figure 47, where the 90% confidence level lines are included for both types of concrete. Compressive fatigue tests 87 L. neh DAS Jel 2 Ea Plain concrete Fibre conerete 00 — —_______ - 4) a aa. eee er oe ay 4“ ilgal| st Fatigue life foo(N)} augue ite fogtN) Figure 47. Whiler-diagram for plain and steel fibre-reinforced concrete with 60-mm length fibres with best fit line and 90% confidence levels included. It can be seen that much larger scatter exists in fibre concrete. Additionally, the correlation coefficients are rather low. A shorter fatigue life was found for steel fibre-reinforced concrete. Several reasons may explain these results, The main reason is related to the fact that the fatigue phenomenon is related to initial imperfections, such as micro-cracks or voids, existing in concrete, Therefore, the presence of fibres, especially relatively long ones, may be an additional cause of imperfections, creating bridges between the aggregates. A size effect can also possibly be observed, since the ratio between the cylinder diameter and the fibre length is 2.5, which is a relatively high value, Another problem arises from the fact that the fibres are initially glued into bundles with glue that is dissolved in the mixing water. As the mixing times are relatively short, some fibres remained glued creating a stronger fibre, augmenting the bridge formation problem. It has been seen previously that the initial tangent modulus is correlated with the compressive strength. Therefore, it was decided to investigate the relation, for each stress level, between the initial tangent modulus and the fatigue level. A chart showing the evolution of the fatigue life with the initial modulus for each stress level is presented in Figure 48. It is apparent from the observation of the figure that a certain degree of correlation exists between the initial tangent modulus and the fatigue life of concrete. 8 e 8 8 Plain concrete 2s 2s {nial tangent modulus (GP) “ . 3 s Fatigue ie (Log) Fatigue ite (Log) Figure 48. Relation between the initial tangent modulus and the fatigue life of concrete. 88 Chapter 4 To study this aspect in more detail, a regression analysis incorporating the fatigue life, the stress level and the initial tangent modulus has been made, To make the equations non-dimensional the initial tangent modulus was normalised with a reference value of Eco = 20 GPa. The regression equations read log(N) = -6.4658 — 24.5263 log(s) +5739 22 (4.3) for plain concrete, and log(N) = -5.5170-27.0347 tog(s)+42451 22, (4.4) for fibre reinforced concrete. In Table 22 the results of statistical tests used in the regression procedure are shown. Table 22. Statistical tests for compressive fatigue (log(V) =a +blog(S)+cE). Statistical test Parameter. ——___—_Rewression equation ___ SNE Plain SNE Fibre pre 0522 0.828 2 0.845 0.722 bevalue 724.526 27.035 ewatstes % 2.738 6.107 (parameter) value 3957 4427 ‘limit 174 1.860 ewvalue 5274 4245 t-statisties 1.824 2.040 (c-parameter) 2.892 2.081 Ld 1.860 aatsics Fevalue 49.23 1041 Folimit 355 4.46 5, standard deviation of variable i. ‘An important increase of the correlation coefficient was attained by the use of the initial tangent modulus as a regression variable, which indicates that it may be used as an assessment of the effective strength of the specimen. 4.4.2 Deformations To illustrate the force-displacement relation occurring during a fatigue test, the force-displacement curve of two particular tests are plotted in Figure 49. For specimen Compressive fatigue tests 89 L2D, individual force-displacement cycles are plotted every 40000 cycles. It can be seen that the first and the last cycles (corresponding to cycle 3 and 160000) are more separated than the three inner cycles (40000, 80000 and 120000 cycles), which corresponds to the fast increase of deformations occurring in the beginning and the end of the fatigue test. It should be noted that the maximum deformation varied from nearly 0.3 mm at the beginning of the test to almost 0.9 mm at the end of the fatigue test. This corresponds to an increase of almost 300 percent in the deformation supported by this specimen i! = 340900 80000 120000 60000 10 feo ie te ee $2 Displacement fmm Displacements (rm) Figure 49. Cyclic force-displacement curves (eft: complete curve of specimen S14, right: cyclic acquisition of specimen L2D). The cyclic creep curves were plotted in Figure 50. Big differences exist between the cyclic creep curves for each load level. Nevertheless, considering the final deformation as unity, it can be observed that for lower stress levels the increase of deformation is bigger for lower stress levels. Thus, for the stress level of 70 percent the final deformation was about three times bigger than the initial deformation, while for the stress levels of 0.8 and 0.9 this ratio was just 1.8 and 2.0, respectively. Usually a strong correlation exists between the slope of the secondary creep branch, the secondary creep rate, and the number of cycles to failure. This relation is linear when a logarithmic scale is used for both variables (Figure 51). The regression equation reads log(V) = -1.0725*log(é) - 0.6556, (4.5) for plain conerete, and log(N) =-1.0298* log(é)-0.2261, (4.6) for fibre-reinforced concrete. The results of the statistical tests are presented in Table 23 where it can be seen that excellent regression coefficients of 0.996 and very small standard deviations for the fatigue life were observed. Normaised dsplacoment Normalized daplacement Normaised displacement Normalised displacement Figure 50. Figure 51. Chapter 4 Plain concrete Fibre concrete $= 0.85 "a 3-085 rer ee a ae BETO Normalised number of cles Nermatsed numberof eles Plain concrete S=0.80 02 oe 8t Normaised number of cies Fibre concrete S=0.70 oe [Normalised numberof eyes Evolution of the maximum deformation during fatigue loading. ‘Secondary creep rate [Log(de/dt)] 1 2 3 ‘ 5 6 Fatigue ife[Log(N)} Secondary creep rate versus fatigue life. on Compressive fatigue tests OL Table 23. Statistical tests for equation log(N)=a+blog(é). tstatsies statistics T Sogn ae wil bevalue ss __tevalue it Fovalue _ Felimit Plain 0.072 0.996 1.0725 0.0163 65.761 1.746 449 4324.44 Fibre 0.070 0.996 -1.0298, 0.0233 44,188, 1.895 5.59 1952.65 s;— standard deviation of variable 1. The analysis of the results show that for the same rate of variation of the deformations, the fatigue life of fibre-reinforced concrete is bigger then that of plain concrete. On the other hand, for the same number of cycles to failure, the rate of variation of the fibre concrete is bigger. This can be understood with the help of Figure 52 where, for both types of concrete, the final displacement at failure is plotted for several load levels. As expected, the final deformation at failure is larger for fibre concrete. Consequently, if specimens of both types of concrete fail with the same number of cycles, then, the secondary creep rate must be greater for fibre concrete. 09 e ances e e gos. © @ z Bor, __ @meoe e ‘0 Plain concrete 1 Fiore concrete 6 $—-________ —— — 0s 1 15 Maximum displacement mm] Figure 52. Maximum deformation at failure as a function of the applied stress level. An important question regarding the fatigue behaviour of concrete is the existence of an envelope for deformations. This concept is generally accepted in the literature for plain and fibre-reinforced concrete (Karsan and Jirsa, 1969; CEB, 1988; Otter, 1996). This envelope usually coincides with the monotonic loading curve or is very close to it. As can be seen in Figure 53 for plain concrete the maximum deformation at failure is relatively close to the monotonic stress-strain diagram. The same conclusions apply for fibre concrete as can be observed in Figure 54, 2 Chapter 4 Batches P10 PS Batches PA oP Monotone Sess MPa} 0 0002 0004 © 0.008 aos aot 0 — 0.002000 ©0008 0008 at Stain Stain Figure 53. Comparison of fatigue deformations at failure with monotonic envelope of plain concrete. Consequently, it seems that the monotonic stress-strain diagram could be used as an envelope for the fatigue deformations at failure. Stress [MPa] of 0 0.002 0.004 0.006» 0.008) 0.01 Strain Figure 54. Comparison of fatigue deformations at failure with monotonic envelope of steel fibre-reinforced concrete. The use of the maximum displacement as a failure criterion is investigated in Figure 55 where the maximum displacement at failure is plotted against the fatigue life of concrete. ‘A trend may be observed between these variables, especially in the case of steel fibre-reinforced concrete. Compressive fatigue tests 93 Ee "0 Plain concrete | ° e & © Fibre concrete R12 ee & g e 10 ° S es 09 @ 06 3 @ 685 0 0% g y__e- 896 Sos 0 0 — a = 06 1 2 3 4 5 6 Fatigue life (Log (N)} Figure 55. Fatigue life versus deformations at failure. 4.4.3 Fatigue modulus Fatigue modulus, Zr, is defined as the ratio between the stress range and the corresponding deformations range within a load cycle: Cm ~F, Pans Pin (4.7) Emax ~ Erin The fatigue modulus may be viewed as a measure of the damage occurred in the material during a fatigue test. Since these experimental fatigue tests are performed under constant amplitude loading, the variation of the fatigue modulus is related to the evolution of the maximum and the minimum deformation during the test. It has been seen that the deformations increase during a fatigue test and that the fatigue modulus decreases during the test. Therefore, the maximum deformation increases at a faster rate than the minimum deformation. The influence of the loading rate on the stiffness of concrete is estimated by plotting the initial tangent modulus against the initial fatigue modulus (Figure 56). The initial fatigue modulus is defined as the fatigue modulus of the first load cycle. A regression equation passing through the point the origin of the coordinate system show in the figure Was fitted to the data. The regression equation was plotted on the figure. Both types of concrete are included in the regression equation since it was found that both materials experienced the very similar rate effects. A ratio of 1.20 was found between the initial fatigue modulus and the initial tangent modulus. 94 Chapter 4 = 45000 Regression equation $ passing through (0,0) 3 40000 3 8 e E 36000 S °~ ae 30000- Plain concrete g @Fibre concrete = 25000-—— = 25000 27000 29000 31000 33000 35000 Initial tangent modulus {MPa} Figure 56. Relation between initial tangent modulus and initial fatigue modulus for 60-mm length fibres. The effect of fatigue loading on the fatigue modulus is shown in Figure 57 where the number of cycles is normalised with the maximum number of cycles and the fatigue modulus is normalised with the initial fatigue modulus. The observed reduction of fatigue modulus is about 30% for both types of concrete. The coefficients of variation are 12.2% for plain concrete and 15.6% for steel fibre-reinforced concrete. Regardless of these similar results, some particularities can be observed. For plain concrete, it appears that the lowest stress levels correspond to bigger reductions of the fatigue modulus. Additionally, for fibre-concrete the observation of the results corresponding to the stress level of 85 percent show large dispersions on the reduction of the fatigue modulus. Similarly to what has been done with the evolution of the displacements, the relation between the rate of variation of the fatigue modulus during the secondary branch and the fatigue life has been studied (Figure 58). The observation of Figure 58 shows an excellent correlation between the rate of variation of the fatigue modulus during the second stage and the number of cycles to failure. The regression equations for the fatigue life versus fatigue modulus variation is log(V) = -0.9973 log(-d£/dN) + 3.7037, (4.8) for plain conerete, and log(V) =-0.9552 log(-dé/dN) +3720, (49) Compressive fatigue tests 9s for fibre-reinforced concrete. The results of the statistical tests are presented in Table 24. 3 oe z a. | i Petes a asec ca ° 02 ee 1 Nomis number tes Nomaise numberof is i ry oe | noe oe Pai ee tay —— on ee a ae Normalsed number tyes as oe ie 7“ gos i or Plain concrete or Fiore concrete | ‘Sueore | Sear aa A a oe : war ene ge a eae ae ae age Nomalea ruber fies Nomaleed number cf les Figure 57. Evolution of the fatigue modulus during fatigue loading. Figure 58. (© Fibre concrete | -2, | OPlain concrete 3 4 5 6 7 Fatigue lite [Log(N)] Rate of variation of fatigue modulus versus fatigue life. 96 Chapter 4 Table 24. Statistical tests for equation log(N) =a+blog(-dE/dN) A statistics Type a bevalue Sy tvalue —__tlimit Felimit Plain 0.150 0.983 -0.9973 0.033 30.207 1.746 912.44 Fibre 0.132 0.987 -0.9552_0.041_——23.394 1.895, $47.30 s,— standard deviation of variable i. - 4.4.4 Failure patterns The failure of the specimens occurs abruptly, with several pieces of concrete being projected from the specimen. The presence of fibres has an advantageous effect since it helps to keep the specimen together. This can be observed in Figure 59, where five representative failure pattems for plain and fibre concrete are shown. Figure 59. Failure patterns for plain concrete (the three left cylinders) and fibre concrete (the two right cylinders). 4.4.5 Post-fatigue properties A monotonic test has been made on the cylinders that do not fail after a million of cycles in order to establish the differences between the pre-fatigue and post-fatigue behaviour. In Figure 60 the stress-strain curve of the control specimens and that of the run-out specimens are plotted to visualise the differences between the two situations. Compressive fatigue tests 97 sean Sren Figure 60. _ Stress-strain comparison of pre and post-fatigue behaviour of concrete. As the run-outs are just three cylinders of plain concrete and two of fibre concrete, the obtained results must be observed with some caution. The increase in strength and stiffness of fibre concrete is quite evident in these charts. Average increases of 6% in the compressive strength of plain concrete and of 15% in the case of fibre concrete were found. The initial tangent modulus increases 10% in case of plain concrete and 16% in case of fibre concrete. From Figure 61, it can be seen that the softening behaviour of concrete was not influenced by the fatigue loading. This may be explained by the fact that, at least partially, the softening behaviour of concrete in compressive tests is a structural phenomenon and not a material property. : renee re =rewe $0 * oie icant Figure 61. Comparison of softening behaviour of concrete in pre and post-fatigue conditions. 98 Chapter 4 The observed hardening effects may be explained by Figure 62 where the cyclic creep curves of the run-out specimens are shown. The normalised displacement is calculated as the ratio between the displacement and the maximum displacement of the test while the normalised number of cycles is the ratio between the number of cycles and the maximum, number of eycles of the test. z10 . jos ao - is {= Bos 3 Plan concrete aca Fore cone 202 Go2 200 - 2 00 — a 00 02 04 08 08 19 oo 02 04 08 8 10 Nomalised number ote Normaises umber ofeycies Figure 62. Normalised cyclic creep curves for compressive run-out specimens, It can be seen that the rate of variation of the deformations is permanently decreasing. It seems that the specimens are not yet at a phase where damage prevails over the consolidation. Thus, as previously stated, the fatigue loading improves the monotonic properties of these run-out specimens. 4.5. Results of 30-mm length fibres fatigue tests Facing the results found for 60-mm fibre reinforced concrete, it was decided to investigate the behaviour of smaller fibres in the fatigue life of concrete. These fibres had a diameter of the cylinder to length of the fibres ratio of five, which is twice as much as that of the 60-mm fibres. Additionally, as 30-mm fibres are much smaller, many more fibres were placed in a cylinder to achieve the same fibre content (approximately five 30-mm fibres had the same weight as one 60-mm fibre). Hence, some of the reasons that may lead to a smaller fatigue life of fibre reinforced concrete were eliminated. A similar test procedure was used in this case, except that the number of cycles necessary to stop the test was reduced to 500000 cycles. The reason to do this is to limit the test duration to 55 hours (a little more than two days) instead of the four days needed to achieve one million cycles. Thus, a specimen is considered as a run-out if it can support 500000 cycles without collapsing. Compressive fatigue tests 9 4.5.1 Stress level The S-N fatigue data related to these compressive fatigue tests is shown in Table 25. Four stress levels have been tested, namely 0.9, 0.8, 0.7 and 0.6, with 3 specimens tested at each stress level. It can be seen that at the stress level of 60 Percent all the specimens are run-out. The data is grouped by stress level and ordered by increasing number of cycles to failure. Table 25. Compressive fatigue data for 30-mm fibres. Plain concrete Fibre concrete Specimen Ss N Specimen 5 N POB: 09 133 SOB 09 82 P9A 09 332 S94 09 2 POC. 09 413, sac 09 1023 P9D 08 186 SoD 08 766 POE 08 1439 S9E 08 2491 POF 0.8 2143 SOF 08 70278 POT 07 3499 S9G 07 6167 POH 07 26047 S91 07 12256 9G 07 34960 soy 07 13689 Ps 06 500000 SoH 07 23077 POK* 06 500000 SOK 500000 POL* 06 500000 soL* 300000 Note: the specimens marked with * are run-out. The data is plotted in Figure 63 together with the fitted regression line and the 90 percent confidence limits. 10 10 Dini a ae augue te fo Feng een Figure 63. Whéler-diagrams for plain and steel fibre-reinforced concrete with 30-mm length fibres (best fit line and 90% confidence levels included). 100 Chapter 4 The regression equations fitted to the data using a power law are, log(V) = 1.3292 -19.1663log(S), (4.10) for plain conerete and log( NV) = 1.6208 -17.6600log(S) , GAD for fibre reinforced concrete. Again, the number of cycles was chosen as the dependent variable. The statistical tests to assess the adequacy of the model are shown in Table 26. ‘The standard deviation of the logarithm of the number of cycles encountered for fibre concrete was 1.6 times bigger than the standard deviation for plain concrete. Table 26. Statistical tests for compressive fatigue (log(V) = a +blog(S)+cE). ; Regression equation Statistical test Parameter - SNPlain S-N-EPlain —‘S-NFibre__S-N-E Fibre Sosa 0.253 0.274 0.409 0.462 P 0924 0.967 0.781 0.890 ‘bevalue “19.166 718.359 -17.660 “15.858 ‘statistics % 1.738 1.29 2.960 2.294 (&-parameter) value 11.026 14,939 5.965 6.912 ‘limit 1.812 1.833, 1.812 1.833 e-value > 5.218 - 5.288 statistics % 1.521 : La (c-parameter) value - 3.431 - 2.981 t-limit : 1.833, : 1.833 Fevalue 121.58 132.14 35.59 36.27 F-statisties Felimit 4.96 4.26 4.96 4.26 ‘s\~ standard deviation of variable i. In Figure 64, the relation between the initial tangent modulus and the fatigue life is plotted for the different stress levels (at S = 0.6 all the specimens are run-out). For each stress level a relation between the initial tangent modulus and the fatigue life can be seen, with the former increasing as the initial tangent modulus increases. Consequently, it was also decided to include the initial tangent modulus in the regression equations. In order to made the equations non-dimensional, the initial tangent modulus was divided by a reference modulus, assumed to be Feo = 20 GPa. Compressive fatigue tests, to1 F 25000 = 00 *K — @S=0.7(P) 3 +x ms-0.8(P) #21000 - axe e AS=09(P) . a .e x8=07 (F) Se aeege a a S=0.8 (F) 5 7000 | . +8=0.9(F) 3 rime est 1 2 3 4 5 8 7 Logarithm of the number of cycles Figure 64. _ Relation between initial tangent modulus and the fatigue life as a function of the stress level for plain (P) and fibre-reinforced (F) concrete. The regression equation for plain concrete reads E, log(V) =18358Blog(S) +521757-¢ -3777, (4.12) and the equation for fibre-reinforced concrete reads log(V) = -158581log(S) +52878 ie -37718. (4.13) The statistical tests performed for a confidence level of ninety five percent are presented in Table 26. It is clear from the observation of Table 26 that a significant increase in the correlation coefficient is achieved by the introduction of the initial tangent modulus on the regression equation. If the initial tangent modulus is used to assess the monotonic strength of conerete using the equations presented in Figure 46 the following S-N relation is found por plain concrete log(N) = 1.1777 -19.4817 log(S) (4.14) and the next equation for fibre concrete log() = 15734 ~16.7296log(S) (4.15) with the regression coefficient being in this case r? = 0.968 and r? = 0.884 for plain an fibre concrete respectively, which are very close to the values found in the S-N-E relations of equations 4.12 and 4.13. From Figure 65 it can be seen that the new S-N relations attained with the use of the corrected stress level yield smaller fatigue lives for concrete. It can also 102 Chapter 4 be seen that the fatigue life of concrete reinforced with 30-mm length fibres has been bigger than that of plain concrete in this series of tests. 1.0 SE Plain concrete (Eq. 410) 2+ Plain concrete (Eq, 4.14) _ 09 «Fibre concrete (Eq. 4.11) g ‘o~ Fibre concrete (Ea. 4.15) s g 08 g a 07 06 — - 1 2 3 4 5 6 Fatigue life (log(N)] Figure 65. _S-N equations derived for 30-mm fibres tests. 4.5.2 Deformations The evolution of the stress-strain relations during a fatigue test is assessed in the following paragraphs. It is apparent from Figure 66 that near failure the stress-strain cycles become bigger, with larger amount of dissipated energy. This is confirmed by Figure 67 where the energy dissipated during each cycle is plotted against the number of cycles. ‘Stress (MPo] ‘Stess [MPa] © 0001 0002 0.003 0.004 0005 0008 © 0001 0002 0.003 0.004 005 008 Strain Skin Figure 66. _Stress-strain diagrams during fatigue loading (left: plain concrete; right: fibre reinforced concrete). Compressive fatigue tests 103 ssn 3 700 i Em 2 2 am 5 8 x0 a Greco; Plain concrete: = Som 209 ‘00 — ee ol 0 coco 000mm a0 ° m =m mm Number of eyies Number of cycles zon sm Ss z Fiber concrete E E son Sra = 0.9 eae 2 3 = B 000 o w Fiber concrete ‘00 Set ene so ° 000 to000 18000 000028000 . = a“ = = 1000 Number of cycles Number of cytes Figure 67. Energy dissipated during load cycles for 30-mm fibre concrete. To estimate the adequacy of the monotonic envelope as the envelope for fatigue deformations at failure, the maximum deformations at failure are plotted over the monotonic stress-strain diagrams (Figure 68). From the analysis of that figure it is clear that the current results confirm the previous results that the monotonic stress-strain diagram can be used as an envelope for fatigue deformations at failure. Figure 68. Comparison of fatigue deformations at failure with monotonic envelope The cyclic creep curves shown on Figure 69 indicate that for lower stress levels a bigger increase of deformation occurs. This means that considering the final deformation as unity, the initial deformation if less for lower stress levels. The relation between the secondary creep rate and the fatigue life of concrete can be assessed with the help of Figure 70. Again, it can be seen that a very strong relation exists 104 Chapter 4 between these two variables. When both variables are plotted in a logarithmic scale, a linear relation is exhibited between them. In this case, the regression equations read, log( NV) = -0.9975log(é)- 0.0770, (4.16) for plain concrete and log( NV) = -0.9850 log(é) +0.076 , (4.17) for fibre-reinforced concrete. The coefficient of correlation of 0.99 indicates the high degree of correlation between the fatigue life and the secondary creep rate. The statistical tests for these equations are indicated in Table 27. __—_—— |: Normalised displacement os on Plain concrete Fibre concrete oz S=0.90 o2 $0.90 oo ——>= Fatigue life (log(N)] Figure 123. Comparison of regression equations for 150-mm prisms for 2 Hz and 5 Hz frequencies. When a comparison was made between the fatigue life of plain concrete and the first-crack fatigue life of fibre concrete, a reduced fatigue life for fibre concrete was observed. On the Flexural fatigue tests 153 contrary, the total fatigue life of fibre concrete was bigger than that of plain concrete. However, it must be noted that these differences occur only for stress levels above 0.8 Reference must be made at this point to the fact the value used to stop the fatigue tests was different for the two load frequencies (four hundred thousand cycles for 2 Hz and one million cycles for 5 Hz). As the run-outs are included in the regression equation, the comparisons made for lower stress levels should be carried out with some caution. Considering just the case of fibre concrete, it can be observed that the first-crack fatigue life for the load frequency of 2Hz was smaller than for the 5 Hz frequency. However if the total fatigue life was considered, this observation was partially inverted: for stress levels above 0.85 the fatigue life of fibre concrete tested at 2 Hz was bigger than that tested at 5 Hz. For smaller stress levels, the previous comments regarding the inclusion of the run-outs in the regression equations apply. As expected, for plain concrete the results observed for both load-frequencies were identical. In fact, the differences observed in the fatigue life of fibre reinforced concrete for the two frequencies were difficult to understand and probably were not with the frequency of the test. The major difference existing between the specimens tested at the two frequencies is the age of concrete when the tests were carried out. Even in this case, the age itself did not seemed to be a cause for the differences because the results for plain conerete were almost the same independently of the age of concrete. Thus, a possible explanation may be that fibres caused some additional defects, such as internal stresses, that will cease with the ageing of concrete. Another reason related with the first-crack fatigue life of concrete may explain the differences observed for fibre concrete, The difference observed between the first-crack fatigue life and the total fatigue life of concrete tested at 2 Hz was related with the number and orientation of fibres that crossed the cracks. Hence, these fibres may be a cause for the weakness of the section, but once the cracked is formed, they became active increasing the fatigue life of the specimen. The results of this investigation are compared with the results of other researches in Figure 124 for plain concrete. The equation presented for the 150-mm prisms includes both frequencies. The results of the 75-mm prisms were also included for comparison regardless of their large scatter, The attained results agreed well with the results of the other researchers being a kind of an upper bound for them together with the results of Koyanagawa (1994). The difference observed between the curve with smaller fatigue life and that of larger fatigue life is approximately log(N). This is a good indication of the 134 Chapter $ dispersion observed in the fatigue results. The regression equation obtained for the 75-mm had an excellent agreement with the other experimental results. 10 ° a8 Fos 3 Bo current (801mm) Curent 7S) 02. © Koyaragawa (1984) —x--Comelssen 1884) 1 2 3 4 Fatigue life [log(N)] Figure 124. Comparison of flexural-fatigue data for plain concrete with the results of other investigations. For fibre reinforced concrete (Figure 125), the current results also agreed well with most of the studies in this field. It is interesting to note that regardless of the large scatter existing in the data, the regression equations for the 150-mm and for the 75-mm prisms are almost coincident. 10 Curent (50s) 0.56% —— Cument (75-mm) - 0.86% 02- Wei (1904) 2 Batson (1972) “a Ramakrishnan-0.5% (1991) = Ramakrishnan-1% (1991) a sun-1% (1988) 1 2 3 4 5 6 7 Fatigue ie (log(N)} Figure 125. Comparison of flexural-fatigue data for steel fibre-reinforced concrete with the results of other investigations. With respect to the final deformation at failure, both plain and fibre concrete seem to have the monotonic envelope as the envelope for it. The relation between the secondary creep rate and the fatigue life of concrete was found independent of the load frequency for both Flexural fatigue tests Iss ‘materials. Additionally, this relation was very similar for both materials. A slope close to 0.90 was observed for this relation when both variables are plotted on a logarithmic scale. The results observed for the 75-mm prisms were in good agreement with those observed for the 150-mm prisms. 2.0 2 Plain (150-mm) S20 = Fibre (150-mm) 3 —2-Plain (75-mm) S40 Fibre 76-mm) 2 3 4 5 6 Fatigue life (log(N)] Figure 126. Comparison of the eyelic creep rate for the 150-mm and for the 75-mm prisms. Similarly to what happens in the compressive tests, the monotonic tests carried out on the Tun-out specimens show an increasing of strength and stiffness for both materials. Besides the strength and stiffness improvement, the post-fatigue behaviour also exhibits a linear stress-strain curve up to a higher percentage of the ultimate strength. 5.5 Summary The described experimental program pointed out some of the most relevant characteristics of the behaviour of plain and fibre concrete under flexural fatigue. The effect of the addition of fibres in the flexural fatigue-life of concrete was small. The competing mechanism between the imperfections caused by fibres and the crack-bridging effect had an extreme importance for the fatigue life of fibre-reinforced concrete. It is necessary to understand when the first prevails over the latter based on factors such as fibre type, fibre content or concrete composition. An excellent agreement was found between the secondary creep rate and the number of cycles to failure, The existence of an envelope for deformations was also observed which means that the monotonic stress-strain curves might be used as a deformations 136 Chapter 5 failure criterion for concrete under fatigue loading. The fatigue modulus is also a very important property to accurately model individual load-unload cycles. The rate of variation the fatigue modulus during a test is strongly correlated with the number of cycles to failure. An increase of some monotonic properties, such as the initial modulus and strength, of the run-out specimens was observed. The addition of fibres to concrete provides an increase in the deformation at failure. The necessity of standard testing methods for the evaluation of the fatigue properties of concrete is mandatory. Otherwise, the comparison of different studies becomes quite complicated, which made the understanding of the fatigue behaviour more difficult. More studies are necessary to fully understand the effect of the fibre addition in the flexural fatigue life of conerete. ‘Numerical model for concrete 137 6 NUMERICAL MODEL FOR CONCRETE In general, the stress-strain response of concrete is very complex as could be noticed in the preceding chapters. It is characterised by strong non-linearity in the deformational behaviour, mainly for high stress levels, which are augmented when load reversals occur. A reasonably large number of real loads, such of traffic, wind or sea wave actions, fluctuate with time, and some of them may also reverse their action or induce stress reversals at the structural elements. These stress fluctuations or reversals cause obvious additional difficulties for the modelling of concrete behaviour. As an attempt to understand the physical behaviour of concrete, much effort has been devoted to experimental investigations. However, these investigations are often rather complicated because of the strong dependence of the behaviour of concrete on the type of loading, since concrete behaves dramatically different in compression and in tension Moreover, the influence of the testing device on the tested specimen is sometimes not negligible and may have a considerable influence on the experimental results (RILEM, 1997). Until now, there is no constitutive law for concrete that is generally accepted. Accordingly, a large number of material models based on different theories have been Published. Most of these constitutive models are characterised by acceptable deviations of ‘numerical results from experimental data for a certain class of loading paths. However, for more general loading histories these deviations may be unacceptable. The numerical modelling of concrete is complicated by the fact that the material behaviour ranges from brittle to ductile depending on the state of stress. Hence, the different types of failure are usually treated separately, but a unified treatment of cracking and crushing offers some computational advantages (Onate et al., 1987; Feenstra, 1993). At the structural level, the finite element method is generally used to solve the global equilibrium equations. These equations are established with the principle of virtual displacements, which represents a weak form of the equilibrium conditions and the static boundary conditions. For an extensive introduction to the finite element method, the reader is referred, for instance, to the books of Zienckiwicz and Taylor (1989) and Ofiate (1992). Since, in general, the mechanical behaviour of concrete structures is non-linear, an incremental-iterative procedure is usually adopted. This means that the external load is applied in increments and that the structural response for each increment is determined 158 Chapter 6 with an iterative procedure, generally with the use of the Newton-Raphson method or with any of its numerous variations. An explanation of the non-linear solution strategies is beyond the scope of this report, so the reader if referred to the works of Owen and Hinton (1980) and of Criesfield (1991), for a comprehensive introduction to this theme. In this chapter, an elastic-plastic model for concrete under monotonic and fatigue loading conditions is presented. First, an overview of basic elastic-plastic relations is presented. It includes the description of the return mapping procedure for the integration of the elastic-plastic equations (a fully implicit Euler backward scheme) and the derivation of the consistent tangent stiffness matrix used in the incremental-iterative procedure. The consistent tangent stiffness matrix plays a crucial role in the performance of the Newton-Raphson method. Next, some enhancements of the plasticity theory are presented in order to achieve better results in the cyclic loading modelling. The bounding surface plasticity theory and the generalised plasticity theory are briefly explained. These theories allow the unloading/reloading process to be non-linear making thus possible a better simulation of the concrete behaviour under load reversals. This theoretical explanation is concluded with a brief overview of visco-plasticity that is used in the fatigue load model. Next, the monotonic load model is described, which has two loading surfaces (one for conerete in compression and the other for concrete in tension) with independent hardening parameters. A linear/exponential hardening law is used for concrete in compression and an exponential softening law is used for concrete in tension. The equivalent length concept is used for both mechanisms in order to provide a relatively mesh independent solution on the softening branch of the stress-strain curve. ‘The model for the modelling of cyclic loading conditions is presented afterwards. It uses for the unloading/reloading process an additional surface, the unloading surface that provides the direction of plastic flow. This surface, which is similar to the loading surfaces, evolves controlled by a kinematic hardening rule. The model is divided into unloading to tension and into unloading to compression whenever the stress increments are tensile or compressive, respectively. Finally, the fatigue load modelling is presented. It assumes that an initial monotonic load is applied to the structure until the maximum applied load is reached. Then a cycle-dependent analysis is carried out assuming a visco-plastic type model. Numerical model for concrete 159 6.1 Plasticity The theory of plasticity has been extensively used for the modelling of the constitutive relations of concrete, particularly under compressive states of stress (see Chen, 1982: ASCE, 1982a). The principal assumption of plasticity theory is the decomposition of strains into an elastic reversible part and into a plastic irreversible part. The onset of plastic strains appearance is bounded by a surface defined in the stress space, the yield surface. If the stress state at a point remains inside the yield surface the stress-strain relation is linear elastic otherwise this relation is non-linear which gives origin to plastic strains, This definition implies that unloading from the yield surface is linear. The plasticity theory, in its classical configuration is thus only suitable for the modelling of monotonic loading conditions. 6.1.1 Basic assumptions In what follows, the basic relations of the theory of plasticity will be summarised. For a comprehensive introduction to plasticity theory reference to the work of Han and Chen (1988) is made. In the theory of plasticity, the strains are expressed as a function of stresses and of some hardening variables that serve as a means to describe the material degradation. Considering that the hardening variables describe irreversible material behaviour, their evolution is governed by rate equations. The total strain rate, , is decomposed into an elastic reversible part, 2°, and into a plastic irreversible part, &”, as ba ae? (6.1) Equation (6.1) constitutes the basic relation of theory of plasticity. The elastic strain rate is related to the stress rate, o, by an elastic constitutive law which is generally assumed linear and isotropic. This relation reads (62) where D is the elastic (constitutive) stiffness matrix. The plastic strain rate is defined by (6.3) 160 Chapter 6 where i. is a positive scalar named plastic multiplier rate and g is the plastic potential, a function of the stress state (and eventually of some hardening variables) which defines the direction of the plastic strain rate vector. The yield surface defines the elastic limits of a material point under arbitrary states of stress, beyond which irreversible strains appear. This surface must be a function of the stress state at that point, o, and of some hardening variables, « and x, that define the evolution of the yield surface, i.e. the loading surface, and reads S(o,a,«)=0, (6.4) where cis the hardening parameter and @ is the back-stress vector defining the centre of the yield surface. The yield function is defined in a way such that if f <0 then the material exhibits a linear elastic behaviour. Loading/unloading conditions can be conveniently established in standard Kuhn-Tucker form by means of the conditions h20 fs0 if=0 (6.5) ‘The requirement that during yield the point must stay on the yield surface, the consistency condition, yields (6.6) where His the plastic modulus. The plastic modulus governs the plastic strain rates in the same way as Young's modulus governs the elastic strain rates, since substitution of Equation (6.6) into Equation (6.3) gives 6.7) (68) The functional form of the yield functions used in this work is f(o,a,x)=((0-a)'P(o-a))'" +(6-«)' p(x) -z(«)=0, 69) Numerical model for concrete 161 where P is a projection matrix, p is a projection vector and z is a scalar function of the intemal parameter that is usually the effective stress. The general expression used for the yield surfaces encompasses several types of widely used yield functions such as the Drucker-Prager, Von Mises or Rankine yield functions and can be extended to a variety of other surfaces. Therefore, the proposed yield functions may have curved meridians but the traces in the deviatoric planes are circumferences. Because of the load surface definition, three basic hardening types can be defined accordingly to the dependency of the yield function on the hardening variables (Table 42) The hardening parameter x controls the shape and the size of the load surface while the back-stress vector a, controls the location of the centre of the load surface. Table 42. Basic hardening types. Load surface variables Hardening types f(a) ‘No hardening (ideal plasticity) f(a.) Isotropic hardening t(o.a) Kinematic hardening S(¢,a,«) Mixed hardening To completely define the elastic-plastic equations it is now necessary to define appropriate rules for the evolution of the hardening variables. Isotropic hardening Kinematic hardening Mixed hardening Figure 127. Basic hardening types. The hardening parameter rate, <, is usually associated with the equivalent plastic-strain ate 6” or with the plastic work rate, 17”. The former hypothesis is called strain hardening while the last is called work hardening, respectively. |n the case of strain hardening, % =e", the equivalent plastic-strain rate can be defined from the plastic strain rate as 162 Chapter 6 ve koh = Me") 2") (6.10) The scalar M is a coupling hardening parameter that ranges from zero to one. Alternatively, the equivalent plastic-strain rate can be derived from the plastic work per unit of volume, which gives oe K=é" = M. . (6.11) where & is the effective stress. If the evolution of the hardening parameter is defined by the work hardening hypothesis, x = W’’,, then, the following expression holds &=W? = Mo'é?. (6.12) The evolution of the back-stress vector, in the case of kinematic hardening, is usually connected to the flow direction (Prager's rule) Peon =(I- Min? = a =(I- MYA (6.13) or to the current state of stress (Ziegler's rule) a@=(I- M)AH?(c-a). (6.14) The scalar H”? is the kinematic hardening modulus. Mixed hardening can be introduced through the coupling parameter M. M =1 means full isotropic hardening while M= 0 means full kinematic hardening 6.1.2 Integration of the elastic-plastic constitutive equations ‘The basic problem in computational plasticity is to update the elastic-plastic equations in a consistent manner. At time ¢ (increment n) the total strains, the internal variables, the back- stress vector, the plastic strains and the stresses are known. Also the total strain increment for time ¢+Ar (increment n+1) is known, Ae’,\', from the loading regime (for simplicity the superscripts related to time are dropped from now on). Therefore, it remains to update the stresses, the internal variables and the back-stress vector, since the decomposition of Numerical model for concrete 163 the total strains into the elastic and into the inelastic components are obtained according to equations (6.1) and (6.3). This problem could be solved by using an implicit Euler backward integration algorithm that is stable and accurate even if the yield surface is highly distorted (De Borst and Feenstra, 1990). The first step is the computation of the elastic trial state o,,, which reads Ora = Fn + DAE, (6.15) and then checking the yield criterion with of,,, a,, and x,. If the yield criterion is violated then a plastic correction must be made and the stresses must be returned to the load surface. Otherwise, the load step is linear elastic. The return-mapping algorithm is described in Box 3. Equations (6.1) through (6.3) can be combined in order to obtain the stress update. The back-stress vector update can be made with the integration of Equation (6.13) or Equation (6.14). Integration of Equation (6.10), (6.11) or (6.12) gives the hardening parameter update. Finally, it is necessary to satisfy the yield condition (6.9). With these assumptions, the following system of non-linear equations holds (Hofstetter and Mang, 1995) , 7 a P(e ea) + Aba 5 =O Biyg hy ~ AOE (FpatsQ yyy) =O, (6.16) Kyat Ky Ak na (TpateKyrr Ages) = 0 Fooi(Frer A nair®nst) = 0 The unknowns of this system of non-linear equations are the stresses, the hardening Parameter, the back-stress vector and the plastic multiplier at iteration n+J. In order to solve the iterative procedure a Newton-Raphson method is used. 164 Chapter 6 Box 3, Return-mapping algorithm (single surface plasticity). ——_——_—_$__$ INPUT: Fy 1b yrKqr Eqn Ey MEna OUTPUT: ysis GnstsKnets Snot Ene 1, Evaluate elastic trial state (predictor) and initialise the counter AA =0 = 2, Evaluate the residuals and calculate current failure condition (equation 12) —D"(o9), -o441)- Aen? 0 -al) + af + Aa, inst * On =n), Hh HRY nel -f0 ret IF (f S TOLRF and [y{?\| < TOLRR) GOTO 7 3, Calculate the Jacobian matrix J), (Equation 13) 4, Evaluate the change in stresses, back-stress vector, internal variables and plastic multiplier sol"? (ied) Sans sd) OK (et) hier 5. Update the stresses, back-stress vector, internal variables and plastic multiplier Fit? = Opa HOO ais) = aye, + 54h, Kye? = Rye FO. (= AAO 4 gue Ade? = Akg + Oger Og Us) agi Aen? = Aaa 241 6. Increment the counter izi+l GoTO2 7. Evaluate the consistent tangent stiffness matrix 8, RETURN The residuals of any iteration are calculated directly from (6.16) and are used to calculate the unknowns update if convergence is not achieved. The Jacobian needed to solve the problem is obtained through differentiation of Equation (6.16) and reads Numerical model for concrete les ORs OG ChY7 OAs BAe | 6.17) ~2an, In some cases Equations (6.16) and (6.17) can be greatly simplified. This means that this system of non-linear equations with fourteen equations (three-dimensional cases, 8 equations in two-dimensional cases) can be reduced to a non-linear equation with only one unknown (Feenstra, 1993; Lourengo, 1996). These simplifications are explained below when the various models are detailed. 6.1.3 The consistent tangent stiffness matrix The non-linear equilibrium equations at the structural level are solved using the Newton-Raphson method that involves the linearization of these equations. The linearization of the equations results in the tangent stiffness matrix which plays a crucial role in the performance and robustness of the Newton-Raphson method (Simo and Taylor, 1985), and must be obtained by consistent linearization of the stress update at the end of iteration n+1: (6.18) The return mapping contains, as unknowns, the stresses, c,,,, and some other state variables, namely the back stress vector, @,,,, and the intemal parameter, x,,,, and the plastic multiplier, A2.,,,, which are now collected in a vector ¢,,,. Then, linearization of Equation (6.16) yields (Lourengo, 1996) aie fasta fees) -f7} wn 166 Chapter 6 where J,,, is the Jacobian used to solve the return mapping, n, is the number of stress is ay components and , is the number of state variables. To solve the above system of equations, the Jacobian matrix must be inverted, which gives: {igo tour’ {s [Se SS} 620) Finally, the consistent tangent stiffness matrix reads o ie (6.21) ‘These procedure is particularly interesting to use in numerical implementations since the Jacobian used to solve the return mapping can be directly used to calculate the consistent tangent stiffness matrix. 6.1.4 Plasticity enhanced models ‘The models derived from the classical theory of plasticity offer a good approximation of the deformational behaviour evidenced by concrete subjected to monotonic loading and simultaneously guarantee that all the principles of continuum mechanics are respected. However, for cyclic loading conditions these models are unable to characterise some of the most important features evidenced by concrete such as the stiffness degradation due to progressive damage of the intemal structure of concrete and the energy dissipation that occurs during the load cycles. To overcome these difficulties some improvements have been introduced in the classical plasticity theory. In what follows the boundary surface plasticity and the generalised plasticity theory are presented. Both include the classical plasticity as a special case. Generalised plasticity also includes the boundary surface theory as a particular case. Bounding surface plasticity ‘To enhance the scope and flexibility of the classical theory of plasticity the boundary surface concept has been introduced. Initially introduced for the modelling of the behaviour of metals (Tseng and Lee, 1983), it was later used for soils (Mroz ef al., 1979; Poorooshasb and Pietruzezak, 1985) and concrete (Fardis er al., 1983, Vermeer and de Borst, 1984; Yang ef al., 1985; Chen and Buyukozturk, 1985). Numerical model for concrete 167 The bounding surface is a surface defined in the stress space that encloses all the admissible stress states of the material and evolves as a function of one or more internal variables. Disregarding the specific solutions used for the different models, the basic idea is to express the plastic strain rate as a function of the distance, 4 between the current stress point, o, and its image, E, on the bounding surface. Therefore, an appropriate mapping rule, which is part of the constitutive model, from each point representative of the state of stress to an image point defined on the bounding surface must be established. Similarly to what happens in classical plasticity, the bounding surface, F, can be expressed as a function of the image stress, ©, of a back stress vector, A, and of a hardening parameter, x, by F(,A,x)=0. (6.22) The load surface, f, may contact with the bounding surface tangentially at a point or even become identical to it but can never intersect with it (Dafalias, 1986) The consistency condition is now applied to the image point on the bounding surface, so that the plastic modulus, H’, now reads 1(OF , | oF -5{ A+ Ze). 6.23) 7(Seardee) (6.23) Finally, the actual plastic modulus, H, is defined through a function relating it to Hand to the distance, ¢, between E and o-. This function is often assumed of the form H=H'+m(C), (6.24) and is such that for ¢>0 one has H > H', while for £=0 the two moduli become equal. The classical plasticity approach can be derived from the bounding surface plasticity, simply assuming that the load surface coincides with the bounding surface. Thereafter, the function m(€) is zero Generalised plasticity A subsequent enhancement of plasticity theory is achieved with the generalised plasticity concept. Introduced in 1984 by Zienkiewicz and Mrdz, it has been applied to model the soil behaviour under complex loading conditions (Zienkiewicz et al., 1985; Pastor ef al., 1985, 1990). 168 Chapter 6 The generalised plasticity theory is entirely based on the definition of tangent moduli, and no yield or plastic potential surfaces need to be explicitly defined. However, if such surfaces exist, as in the case of classical plasticity, generalised plasticity can still be applied, If irreversible, inelastic, strains are to be produced, the constitutive matrix has to be dependent on the direction of stress increment otherwise application of & followed by -& will produce an increment of strain c?e+C"(-6)=0, and hence no permanent deformations can be induced. For practical purposes, two directions provide a sufficiently accurate description of material behaviour, namely, loading and unloading, A direction vector m defined in the stress space (and depending on o and some state variables) is then used for distinguish between loading and unloading. Of course, this vector defines a surface that is equivalent to that used in classical plasticity but this surface need never be explicitly defined. Loading and unloading can be defined by n’o* >0, for loading, (6.25.a) n’ 6" =0, for neutral loading, (25.b) n'6* <0, for unloading, (25.c) where G* = Dé. The incremental stress-strain relations read b=Cyyo, (6.26) where the subscript L stands for loading and U for unloading. Continuity between loading and unloading stages requires that the constitutive matrices C, and Cy have the form (Pastor ef al., 1990) (6.27) where C=D"!, ng, and mgy are arbitrary vectors and H, and Hy are plastic moduli corresponding to loading and unloading. Under neutral loading &' =Co=é", since ‘Numerical model for concrete 169 according to Equation (25.b) n’&=0. Thus, the material behaviour is reversible and can be regarded as elastic. ‘The increment of deformation caused by & can then be assumed to have an elastic, é, and a inelastic component, é? , given b: 1pO1 gi y é (cogonann'boesee, (6.28.a) &=Co, (28.b) Deyn. (28.c) Ay To fully characterise the material behaviour the following laws must be established: loading and unloading directions n, plastic flow direction ng, and ngy and plastic moduli H; and Hy. All of these quantities are, or may be, dependent on the stress history and phenomenon such as degradation and memory of previous loading events must be taken into account. The classical plasticity and bounding surface plasticity can be derived from generalised plasticity if the material laws are established according to Table 43. In these cases, appropriate evolution laws for & and & must be established. Table 43. Derivation of plasticity models from the generalised plasticity theory Plasticity theory 2 Ta rm He Classical £ 8 - +00 ao ao Bounding surface a a +0 6.1.5. Visco-plasticity In all inelastic deformations, time rate effects are always present in some extent. Whether or not their exclusion has a significant influence on the prediction of the material behaviour depends upon several factors. In situations where a time-dependent behaviour is clearly important, this may be modelled by using visco-plasticity theory (Owen and Hinton, 1980). 170 Chapter 6 Similarly to standard plasticity, the total strain rate may be decomposed into an elastic part and into a permanent, irreversible, part. However, this permanent part of the deformation is now given by ae , A=6(8) (6.29) where ¢is a viscosity parameter that represents a measure of the increment of permanent deformation per cycle and ¢ is a function of the state of stress and of the hardening parameter. The ()-function reads ¢ if ¢>0 and zero otherwise. The stress rate is calculated in the usual way by o=D(é-é?)=Dé". (6.30) The integration of these equations lead to the following system of non-linear equations (Zienkiewicz and Taylor, 1989) Ae — Mig Aba net Ae Monat 3 ret Ky ~ BK gat (Ayer Kner Anat )=0- (631) a oAt K, Mya -9 (Frais) = 9 where the similarities with classic plasticity are evident, especially if gis taken to be equal tofand if Ar=o or G=00. It must be noted that with this formulation the state of stress could stay outside the current fatigue load-function. However, if a rate dependent load function, fjzrae, is defined 85 Some = (OK wt it may be said that the stresses must satisfy the condition Syonue =0 and the similarities with classic plasticity are complete This system can be solved using a standard Newton-Raphson algorithm with the Jacobian necessary to solve it being given by Numerical model for concrete im (6.32) a0, The residual vector reads ME no + DEQ + AA gas 8b, iF ee eee) i Tei = ap bet + Bret) al (6.33) Ret TK, FAR The last term in the first set of equations, 5¢,+1, is the strain increment due to the time change Av. 6.2 Constitutive relations for monotonic loads In a typical stress-strain curve of a concrete specimen loaded in uniaxial compression (Figure 128) three different deformation-stages are observed until the peak stress is reached. In the first stage, up to about 30% of the ultimate compressive stress, f, the cracks existing in conerete before loading remain nearly unchanged leading to an almost linear elastic response of the specimen. Beyond this limit and up to about 75% of f, cracks start to increase in length, width and number and start to link each other. Consequently, material non-linearity becomes more evident. The third stage, above 75% of f,, is characterised by formation of cracks through the mortar that join the cracks between mortar and aggregates. At this stage, crack formation is no longer a stable process. Beyond the peak stress a strain softening behaviour is observed. However, this softening branch of concrete is highly dependent upon the size of the specimen and the boundary condition of the experiments (Van Mier, 1984; RILEM, 1997). Moreover, it seems that the softening branch does not reflect a material property but rather represents a structural property (Van Mier, 1984). Anyway, it is assumed in this study that the compressive softening of concrete can be represented by compressive fracture energy, Ga 12 Chapter 6 so that objective analysis with regard to mesh refinement is obtained. Although the basis for the present definition is only numerical, some evidence exists on the existence of a local and non-local compressive fracture energy (Vonk, 1992). ca an ok Figure 128. Uniaxial stress-strain response of concrete (left: compression; right: tension). When loaded in tension (Figure 128), the limit of elasticity of concrete is about 60 to 80% of the ultimate tensile strength, fy. Above this level, the bond micro-cracks start to grow very fast and to join each other. The aggregate-mortar interface has a significantly lower tensile strength than mortar, which causes the low tensile strength of concrete. The deformational behaviour of concrete in tension is quite brittle. The ratio between the concrete tensile strength and the compressive strength, k., ranges from 0.08 to 0.12. Beyond the peak stress, concrete exhibits a significant strain-softening behaviour. Deformations are strongly localised and the strain distribution is no longer continuous. Crack propagation is an unstable process. The energy dissipated beyond the peak stress is the fracture energy of concrete G,. The fracture energy is mainly dependent on the size of the aggregates and on the compressive strength of concrete (CEB, 1990). The total compressive fracture energy found in the experiments of Vonk (1992) is about 50-100 times the tensile fracture energy. For multiaxial states of stress, a failure surface must be defined (Figure 129), Concrete have a failure surface with curved meridians, indicating that the shear strength of the material is affected by the hydrostatic pressure. The meridians open in the direction of compressive hydrostatic pressure, meaning that a pure compressive hydrostatic pressure cannot cause failure. The tensile meridian, p,, the shear meridian p, and the compressive meridian, pe, satisfy the condition

You might also like