Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Paper: Carlton

Ordinary Meeting
A paper to be presented and discussed at the Institution of Structural Engineers on Thursday I1 March 1993 at 6 pm.

Application of the finite element method to


structural engineering problems
D. Carlton, BSC, CEng, MBCS
Taywood Engineering Ltd

Dale Carlton has been associated with computer- ourselves just how far these methods were probed on some very primitive
assisred engineering at Taylor Woodrow for 30
years, The early days were spent in developing computers. Thus a frame analysis program containedthe elastic instability
analytical solutionsfor prestressed concrete functions (S and c functions) which could be utilised in an iterative process
pressurevessels,which ultimately led to the to predict the collapse of,say, a steel frame under elastic instability - an
introduction of the finite element method. Since
then, he has managed the introduction of CAD early example of buckling analysiswith an implied geometrical stiffness
and other aspects associated with the use of formulation. In pressure vessel design, utilising a few basic components
computers. Presently, he manages a hybrid team
of analysts and system developers. which permitted only a few variables (i.e. thicknesses, together with unit
load cases to represent prestress), it was possible, with the use of linear
programing techniques, to produce minimum construction cost designs
based on the cost of concrete, shuttering and prestress.
An early example of a finite element approach can be traced to our
developments for the Wylfa pressure vessel which used50 to 100 elements
Synopsis of a specific type to analyse such a structure and permitted loadcases to
The paper briefly traces the development of the finite element be combined in order to provide detailed analysisinformation toassist the
method and considers its relevance to the work of the design of this structure’. The development of that software in the early
structural engineer. It draws attention to the role of the ’60s was performed at the IBM Data Centre on a 7090 computer, and it
analyst and the necessity to pay due attention to the modelling was at that time and on that machine that Cheung, working witkzienkiewicz
activity. Recent developments aimed at defining accuracy by at Swansea, was developing one of the first finite element systems3.
the use of adaptive elements are noted, as well as the role of In the search for understanding the behaviour of such structures and
NAFEMS in ensuring that quality assurance is firmly in place. to generalise the approach, it was recognised that, with the emerging
The computer technology requirements are defined and computer power making it possibleto solve more and more equations, it
examples given with respect to large-scale analytical work and waspossible to dividethestructureintomanysmallercomponents.
the use of FEM in the prediction of the failure behaviour of Clough4 recognised that a most useful element wouldbe a triangle, since
reinforced concrete structures. almost any shape could be constructed using this, and this approach led
to the first finite element formulation.As an alternativewe pursued the
The beginning development of a mesh-based system which was strictly a finite difference
The early historyof the development of the finiteelement method through approach. It is important to recognise that both approaches used an
the 1960s is well described in a paper by Zienkiewicz’ where the ingenuity engineer’s conceptof behaviour. Thus in the triangular element approach
of the engineer seeking solutions to the then current problems is clearly the concentrationwas on the stiffness methodwith the direct solution of
illustrated. The methodis traceable to that time and occurred because of the equations. The alternative approach was, surprisingly, based on the
the developmentof computer technology. At the end of the 1950s the first formulation of equations of motion which, if processed through timewith
commercialcomputerswere being producedand,shortlyaftertheir a damping function,led to a static solution. This latter method was called
introduction, high level programinglanguagessuchasALGOLand ‘dynamic relaxation’, coined by Otter in 19655.
FORTRAN were made available and were accessible to the engineering Dynamic relaxation would todayberegardedasan explicit time-
community at large. Prior to that time the use of computers forengineering integration approach, while the solution procedure would be an accelerated
analysis had been available only to large aeronautical companies and some relaxation method. The stiffness approach postulated in theearly finite
major universities. Access to computerswas enhanced by the opening of element methods based on basic engineering principles was later recognised
computer bureaux which provided the opportunity to buy time on machines as a minimum energy theorem.
which were much larger than an organisation could justify. The two methods are today brought together in the general finite element
One major engineering challenge at that time was the development of system. Elastic problemsusually utilise a direct solutionof equations since
nuclear power stations, which necessitated a good understanding of the this permits many loadcases to be processed using the reduced equation
behaviour of engineering structures, particularly those associated with the matrix. It has, however, now been found that an iterative equation solver
reactor, including the pressure containment structure, firstlyin steel and can be more effective for non-linear problems6.
then subsequently in prestressed concrete.

Development of method Today’s finite element systems


With the introduction of computers, it became possible to solve many Most of the current FE systems commercially available consist of three
simultaneous equations- i.e. by hand, say, three, by calculator maybe programs:
five or six but, with computer, lOs, loos, lOOOs, limited mainly by the storage
-preprocessor
size.
-displacement and stress analysis
Thus the first excursions into solving problems with multiple redundancy -postprocessor
commenced with having available knownsolutionsforstructural
components, such as beams and shells. These solutionsinvolved knowledge
Preprocessor
of the reactions causedby applying loads to such components, as well as Most preprocessors have the following capabilities:
boundary deformations and reactions caused by applying unit loads to
specific degrees of freedom, i.e. the assessment of the stiffness. Thisled -generation of complete input file for FE program
specifically to frame analysis software, as first produced by Livesley, -mesh generation capabilities (lines, surfaces, volume)
available commercially in the late 1950s. -generation of nodes and elements
During the 1960s problems associated with pressure vessel design for -geometry plotting capabilities
nuclear reactorsled firstly to solutionsbeing based on combining known -checking facilities
solutions for cylinders,flat plates, domes. It is interesting to remind -equation optimisation

The Structural EngineerNolume 711No.4116 February 1993 55


Paper: Carlton

FE program facility, which can graphically display the object and assist in superimposing
The main function of this program is to carry out a solution procedure a mesh on that object. It is obvious that a link between a CAD system,
and produce results for a given problem. The programconsists mainly of which is being used in the design process, and a modelling system is an
the following: attractive proposition, but this link is not without difficulties.
The input data arecompleted by defining restraints, material properties
( I ) Extensive element library (e.g. truss, plane-stress, beam, pipe, spring,
and load applications, together with analysis control information.
plate/shell elements. Some have user-supplied element facility)
The processing phase involves submitting the data tothe system to obtain
(2) Material models (linear isotropic elastic, linear orthotropic elastic,
results. For a large analysis,thisstep mustbe taken with careful
non-linear elastic, concrete/rock/soil models, plasticity models, creep)
consideration. There is usually a phase where the data is checked by the
(3) Solution capabilities (statiddynamic, frequency solution, response
system to ensure that it is complete as far as is required by the program.
spectrum, linearised buckling, material/geometric non-linearities, large
It is at this time that the analyst should be taking as many views of the
deformation, large strains, etc.)
data as is practicable to ensure that he has defined the problem correctly.
(4) Equationsolutionmethods (direct solution techniques: Gauss
Once errors and omissions are corrected, it may be necessary to process
elimination, frontal solution, etc.)
the problem to just prior to the equation-solving phase. At this point it
( 5 ) Iterative schemes (Newton Raphson, full Newton with and without
is usual thatinformation is supplied by thesoftware whichgives an
line search, BFGS, automatic-time-stepping-method, etc.)
indication of the equation-solving time, by virtue of number of degrees
(6) Initial conditions (displacement, velocity, acceleration, temperature,
of freedom, maximum band width, and the mass storage requirements.
imperfection, restart facilities)
This is essential information to ensure that the solution will be obtained
(7) Boundary conditions and constraints (generalised constraints between
within a reasonable time, i.e. less than 24 h of processed elapsed time, and
degrees of freedom, rigid links, prescribed displacements, skew system)
that there is an adequate amountof free disk space to accept the results.
(8) Mass and damping option
Some care must be taken with this latter requirement because computer
(9) Loadingoption
operating systems do not normally allow a single file span over disks;
(10) Response spectrum analyses
therefore, either there must be a special feature within the finite element
(1 1) Output (displacements, velocity, acceleration, stresses at nodes or
system or the software shouldbe run on a machine that has this feature.
Gauss points, strains, nodal temperature, reactions; state of stress
information (cracking, crushing, yielding)) After processing a structural analysis problem the finiteelement system
will produce nodal deflections and stress and strain information either at
Postprocessor element nodes or, more probably, at the element integration points, (Gauss
This processes the results in order to make them more useful and usually points). The results can then be viewed by postprocessing software, some
possess the following features: of which may be part of the finite element system, or by other software
such as the modelling software or by user-written software.
-ability to extract results and use in an arithmetic expression The output of a large finite element analysis is voluminous and usually
-plots of original/deformed mesh cannot be comprehended from a printout; thus thepostprocessing software
-plots of mode shapes inevitably uses a graphical presentation of selected results. Typically, the
-plots of contour (U,€, heat fluxes) graphical presentation takes the formof plotting the deflected shape, vector
-display vector plots plots ofthe stresses usually global and principal, andcontour plots of these
-graphical presentation stresses. The trend now is that the contour plots are colour shaded which
can be viewed or plotted on a suitable plotter.
The process The user may require output in different forms depending on thecriteria
The analysis process involves the use of FE programs toprovide the design stipulated in the design specification or standard. Thus, if the structure
engineer with valid information to design a structure that is safe for the consists of beam or truss members, force information, usually provided
environment in which it must operate. The total analysis process results by the software, may berequired, while for volume structures such as the
in a FE outputwhich isinterpretable in terms of design decisions and consists ground or heavy prestressed sections such as a pressure vessel the stress
of the following steps: information may be adequate. However, for slab and shell structures it
is inevitable that thedesign will require stress resultant information which
-converting the designed structure into an idealised structure
is not always provided by the software. Systems for processing results for
-converting the idealised structure into a geometrical model
such structures are usually developed by the user organisation.
-interpreting the geometrical model as a meshed model
-converting the meshed model into solution model
Linear or non-linear analysis
-running the solution model to create a numerical solution
Design standards inevitably consider limit statesof serviceability and
-using the numerical solution to perform a results interpretation
ultimate. The materials that the structural engineer uses have non-linear
-with a satisfactory result set, performing the postprocessing
stress-strain relationshipsand a material suchas concrete exhibits changes
-with an adequateset of postprocessed results, obtain a set of structural
in behaviour with time because of gain in strength and creep, both related
characteristics that describe the actual behaviourof the real structure
to the chemical process between the cement and the water. These aspects
subject to the actual design load suggest that a non-linear analysis is desirable. Because, however, a structure
The process of solving an engineering problem using the finite element has to withstand a variety of possible loads and combinationsof such loads,
method is to model the topological aspects of the problem using an assembly it is usually impractical to handle theproblem by non-linear methods except
of elements selected to reflect accurately the physical nature of the problem. for special circumstances.
Thus, if the problem is a structure, the elements must not only reflect the The use of linear methods of analysis for, say, the design of reinforced
shape of the structure but also ensure the correct flow of stress from the concrete structures can openup philosophical discussions as to its validity
loaded zones to the reaction zones. since, when part of a structure cracks (which it may do at the level of
An analyst is thus mapping theelements onto a structure and, in order serviceability load factors), inevitably there will be a redistribution of the
to describe this mapping within the system, by the data input,has to define stresses and stress resultants, possibly therefore causing increases of these
the position of all the nodes and how the nodes are joined to form the in other partsof the structure. The designer should be encouraged, therefore,
elements. This phase of the work can be verytedious. A finite element system to examine carefully not only the local force (stress resultant) behaviour
usually provides some help in the process by a number of means, e.g. with respect tothe dispositionof the reinforcementbutalso a gross
behaviour which may span many partsof the structure(i.e. a cross-section)
(1) If an element has mid side nodes, these can be calculated by the program. involving many element resultants.
(2) The use of mesh generators such thatzones can be defined and a zone The concrete creep problemis also usually dealt with by a linear analysis
can be divided automatically into elements with predefined divisions approach. The understandingof creep of concrete is that, provided that
set along its boundaries. the stress is below 0.4 of the concrete strength, the strainsare proportional
(3) An extension of the abovewhereby the zones are defined as substructures to stress. Thestrains, however, under a unit stress (referred to as specific
or superelements. strains) vary according to age of loading, time under load, and temperature.
It is possible that the process can be assisted by the development of small In many circumstances, it is possible to arrive at an effective modulus of
special-purpose programs, but todayit is most likely that this initial phase elasticity based on the specific strain relevant to the age of loading, time
or preprocessing phase will utilise a modelling system, similar to a CAD under load, and temperature. The reciprocal of the specific strain is the

56 The Structural Engineer/Volume 71 /No.4/ 16 February 1993


Paper: Carlton

effective modulus. In some structures, loading may be applied at different of engineers who had that‘feel’ for the results and could satisfy themselves
times, and the modulus may vary through the structure. Therefore, each that what was being obtained from the computer was reasonable. Today
load has its own set of effective moduli. Most finite element systems do some of that ‘feel’ may be missing. The problem is compounded by the
not permit load combinationsto be made between identical structures with ease of use of the products, the ever-increasing powerof computers, and
different material properties and, under these circumstances, users may the dramatic fall in real costs. With today’s finite element systems and cost-
have to produce their own software for this purpose. effective computing, the models being analysed are becoming increasingly
Finite element systems are available with non-linear analysis capability complex to such a degree that even the most competent engineer may be
and canbe used under special circumstances provided that a suitable material unable to make qualitative judgment on validity the ofthe answers obtained.
model existsor theuser provides one throughan end-user subroutine. The This situation has been recognised, and a major step was taken in 1983
kinds of problem that might be pursued using a non-linear analysis are: with the formation of the National Agency for Finite Element Methods
& Standards(NAFEMS).Theagencyhasproducedanumber of
-ground stability or settlement problems
publications, the most notable being Guidelines to finite element practice
-behaviour of mass concrete poursinvolving thermal creep and cracking
and A finite element primer, both of which are essential reading for all
-gross overload analysis of structures eithersteel or concrete,prestressed
associated withfinite element work, whether as analysts or project managers.
or reinforced
The first level of responsibility rests firmly with the software developers,
A non-linear analysis is, however, time-consuming with respect to the who must ensurethat the behaviour of the elements, when properly used,
computer when compared with a linear analysis and also requires special produces accurate results. Comprehensive verification material is essential,
skills fromthe analystincontrollingtheprocessas well as a clear and comparative testing with other systems would be welcome. For the
understanding of the material models and the possible limitations of the latter, NAFEMS has developeda series ofbenchmarks andmany developers
system. have produced results’. The developer also has a responsibilityto inform
the users of the system of known mistakes in the software or misleading
The computer technology information within the documentation.
As was stated at the beginning, the use of the finite element system was After the developer has done all it can to deliver a high quality product,
dictated by the availability of computer technology, and its growth in use it is in the hands of the user to use it properly.
and the continuous development has been maintained by the changes in The NAFEMS guidelines point outthat a finite element analysisis only
computer technology. as good as:
The trigger point for bringing the finite element systems inhouse was
-the model of the structure (i.e. the mesh and the elements)
thearrival of theminicomputerintheearly ’80s. Since that timethe
-the assumptions embedded in the properties used for each element
computing power of relatively accessible computers has increasedby a factor
-the representation of the external loads and constraints in terms of
of at least 80, while price has decreasedby a factor of approximatelyfour.
the discrete boundary variables
The result is a price/performance improvement in excess of 300 times.
A finite element analyst requires the following machine environment: By definition a finite element analysisis only an approximation, and it is
usually impractical to model all details of the real problem; thus, further
-a high performing CPU
approximations and assumptions are required at the outset. It is important
-a large amount of RAM
that the analyst appreciates the mode of behaviour of the problem and selects
-a graphics display screen
elements and boundary conditions thatbest reflect that behaviour.If there
-high volume disk capacity
is doubt, consult the verification material and conduct small tests with mesh
-a graphics plotting facility
refinements to test convergence. This is usually beneficial, but there are
Most finite element systems are available on PCswhich have a reasonable circumstances where convergence can be misleading. Equilibrium checks
CPU performance at the upper end of the available processors. Basic utilising boundary stress resultants rather thanreactions can very often assist
machines need to be enhancedwith more memory and moredisk capacity, in validating the accuracy of solution. It is likely in, say, a linear static
perhaps a larger screen and connection to a plotter or a postscript laser analysis that deflections will be better than, say, bending moments which,
printer. in turn, will be better than shear.
In a more demanding situation, greater than casual use, it is likely that Given a suitable element choice, the skill of the analyst is to develop
a RISC-based workstation will be used with a RAM memory inexcess of the mesh, grading themesh to reflect the proper behaviourof the structure
32MB and disk capacity measured in gigabytes. and avoiding element aspect ratiosthat invalidate the use of that element.
The graphic facility is more likely to be colour, with the capability to Because the processof mesh creations is so important, the researchers of
accept screen dumps, plotting protocols (HPGL), and postscript. finite element systems have been exploring strategies that may, in due course,
The analystusually functions within a teamwhere it is important toaccept make this step contain more certainty about the error of the approximation
and deliver information. Data interchangeis thus crucial, and this is best and improve the best approximation. Such methods are termed adaptive
achieved by the use of a network. Such a network maywell have to share procedures, and the aim is to produce a mesh within a specified bound
data between at least PCs and workstations, and today that is best achieved of local or global error.
by a TCPAP-based network with NFS facility. There are two main approaches for adaptive procedure^*.^*'^. The h-
In many instances the systems in use are pushed to their limits, and it version achieves a required accuracyby using a fixed typeof element and
is disastrous if results are lost after intensive processing; thus, asis good successively reducing the elementsize. Thep-version achieves accuracyby
practice, systems should be backed up frequentlyand results archived. A using a fixed mesh and gradually raising the polynomial order of the
networkfacilitymustlogicallycontainagoodbackup and archiving equations describing each mesh element to approximate more closely the
mechanism. model’s behaviour. A third approach combines the two methods to produce
If there is plenty of activity within the finite element area, it may be the h-p version. These approaches are, however, iterativeand are therefore
desirable to include a batchprocessing machine, such as vector
a processor. similar to convergence tests. Thekey issue for their acceptanceis to define
This machine needs to be at least five times better in performance than useful error estimation algorithms based on energy and local stresses. In
any workstation. addition to the adaptive methods, alternative methods based on force
formulations rather than displacement formulations are being pursued.
Accuracy of solutions A goodexample of theforceformulation is theintegratedforce
The route from the postulated problem to the final results for design method”.
validation purposes is fraught with many obstacles which may deviate the
results from an acceptable level of accuracy. As software becomes easier Uses in structural engineering
to use and more automation is built into the process, it is possible that The fundamental basis embedded in the finite element methodis used in
the degree af competence of auser may become regardedas aless significant many similar products, such as frame analysis. A design office would
factor. This is a very dangerous attitude to adopt. It is, however, not normally use this subsetof the method to assist with the design of building
unusual, since this trend hasbeen perceivedin all areas of engineeringwhere frames, particularlywhere there were unusual framing arrangementswhich
the user can use tools provided via the computerwhich are there to assist invalidated the useof simpler methods. For many years, bridges were
him. During the eraof approximately 30 years when all forms of tools have analysed by frame analysis software using, if necessary, a grillage analogy.
been developing, whether they be FORTRAN programs, spreadsheets, Today, more specific software exists which incorporates some relevant
analysis software and CAD systems, there has usually been a generation elements but, importantly, contains traffic loading modules and design

The Structural Engineer/Volume 71 /No.4/16 February 1993 57


Paper: Carlton

informationpostprocessingmodules.Perhaps we are now missing a structure analysed elastically and designed non-linearly. In such problems
convenient facility to generate influence lines? or similar problems, it may be desirable to check a controlling load case
Structures that inevitably require the assistance of a finite element analysis with the designed reinforcement or prestress to understand how the structure
are usually those which contain slabs and walls subject to Some unusual could be expected to behave at theultimate limit state. An obvious example
loading situation. There can be difficulty in interpreting results in line with in the field of nuclear engineering is the prediction of the ultimate pressure-
design Code requirements. carrying capacity of a pressure vessel or a containment structure.
The methodis useful for theassessment of stresses arising out of thermal There are a number of papers12913v’4 that describe possible reinforced
effects provided that the analyst is aware of the creep behaviour when the concrete models. There are basically two approaches with respect to the
material is concrete. element formulation. In one approach thereinforcement is treated as truss
Two examples are cited which briefly illustrate the use of the method elements attached to the nodes of concrete elements. The concrete element
in structural engineering and highlight certain issues concerned with its is a plane element or a solid element. In the other approach, which is more
application. specific, an element is a reinforced concrete plate element whose behaviour
is obtained by considering it as a layered element. This latter approach is
Large linear analysis very similar to that used in the check procedure described previously. In
A recent example of a large linear analysis concerns the work on the caisson both cases it is the modelling of the concrete that presents the major
design for a bridge structure. In thisbridge there are63 caissons of similar development problem since it is necessary to describe its behaviour under
shape but ofvarying depths. A representative number of the caissons were multiaxial stress states, with non-linear stress-strain relationships and
analysed by a finite element system and the results postprocessed in order requiring treatment of the post-crushing and post-cracking behaviour. The
to assess the amount and disposition of the reinforcement. last aspect, along with shear carrying capacity dueto aggregate interlock,
A typical caisson has the order of 1375 elements and 31 000 degrees of causes the most debate.
freedom. It is subjected to a variety of loads which are combined to produce Commercially available software, if it possesses a concrete element, is
load cases for serviceability, ultimate andaccident situations. The running most likely to use the first form of implementation. Thisis probably more
time for such an analysis on a SUN Sparc workstation is of theorder of 24 h. beneficial to the analyst because it does not presuppose a linear strain
The loads are taken out by reactions on the ground but, instead of variation across a section. For instance, the ADINA code uses thisapproach.
modelling the ground, the reactions were calculated using the principles Taywood Engineering has used this aspect of the abovecode in a number
of Bent Hansen andNavier which simulated extreme situations. The ground of situations, predominantly for nuclear containment structures but also
reactions were calculated from initial reactions atfixed points calculated for tunnels and shafts. Two containment structures were also built as models
by the finite element system. A balanced set was then achieved, with any and subject to a ‘round robin’ of pretest predictions by a number of
small difference balanced by the system. organisations. It is undoubtedly quite a challenge to match predictions with
Certain procedures were followed in mesh generation to ensure logical measurement. Within thepredictions there is a degree of scatter, and there
node numberingwhich obviously related to different parts of the structure is never certainty about theaccuracy of all measurements, particularly strain
and progressed in an orderly fashion. Theelements were also grouped into gauges in regionsof interest which, because of theirlocation, may be subject
the parts of the structure. to damage during the progress of cracking. One of the benefits of an
Thick shell elements were used over most of thestructure. In someplaces, analytical approach, however, is that assumptions can be varied and usually
solid elements were used, and use was made of generalised constraints to therefore a bracketing of the measured behaviour can beachieved which,
couple nodes together where the section was thick and between solid and in truth, is the essence of design.
shell elements. This paper is not a place to discuss specific results, but two issues from
The outputof such an analysis consists of deflectionsat nodes and stresses the analysis are worth noting.
at Gauss points. For shell elements thestresses are provided at the top and In one of the containment structures prestressing was present. Where
bottom surfaces and the midplane. Such output is particularly unfriendly the prestressing is straight, it is a relatively simple matter to imposeinitial
for the reinforcementdesigner. Inhouse software, originally developed for strains in the prestress elements and totrack the subsequent effectsof the
similar work on a reactor containment structure,was used to convert the prestressing tendons, which act similar to reinforcement, be it bonded or
stress informationinto stress resultantinformation, i.e. direct forces, unbonded. Multicurved tendons, particularly unbonded ones, do, however,
bending moments, inplane shear and torsion, and through thickness shear. pose more difficulty because essentially their behaviour is controlled by
The element grouping then permitted sets of load case results filesfor -pecific contact between the tendons and the concrete, which involves frictional
parts of the structure to be identified. Further postprocessing could be behaviour.Moreanalytical research is required to produce a useful
carried out on those files, factoring, more combinations, extrapolation, algorithm for this solution.
and (if necessary) integration of boundary forces. The resultant set of files The other issue concerns crack width prediction. In the finite element
thus constituted thedesign information. Thatdesign information was then code the behaviour of an element is controlled largely by what happens
used to assess steel quantities against standard requirements. at the integration points. Thus if, say, the tensile stress in a particular
There are two approachesin dealing with the force fields applied to a direction at an integration point exceeds a preset limit, a crack indicator
plate element.In onecase equivalent forcesets can be produced,using the is set. It is usual then in the formulation of post-cracking behaviour to
Wood Armer approach,which relate to the design of top and bottom steel permit this maximum tensile stressto diminish to zero over a certain increase
in specific directions. The second approach,which was used in this work, in strain. Eventually, a strain limit is reached, when it is deemed that a
is based on the general non-linear analysis of cracked reinforced concrete. crack has been formed and thereis no further resistance to tensile loading
Thus thedesign stress resultants are treated as forces actinga unit on area in that direction. The behaviourin this tensile region can be characterised
of material of thespecific thickness, and a non-linear analysis of just that as elastic behaviour up tocrack initiation, followed by a phase association
area, taking account of the stress-strain behaviour of the concrete and steel with crack formationuntil a maximum strain is reached signifying a ‘real’
and permitting the cracking of the concrete,is performed to obtain strains crack. The normal postprocessing will display ‘cracks’ at every Gauss point
and stresses inthe concrete and steel. For specified amounts of reinforcement which have reachedthe maximum stress limitand usually therefore illustrates
the program will assess various ratios between actual stresses and permitted many apparent cracks. A better interpretation is achieved if cracks are
stresses, as well as predicting crack widths. The outputof the program can displayed only when exceedingthe maximum strain. Some estimate of crack
then be graphically presented to display the variations of the parameters width can be achieved by integrating strains over these apparent cracks
for the various load cases. Algorithms have been produced which will and dividing by a reasonable estimate of crack spacing which is largely a
intelligently ‘design’ the steel. function of reinforcement spacing and cover.
ThepresentprogramincorporatesBritish,NorwegianandDanish
standards. Future directions
This problem typifies the use of finite element systems for the design There is no reason to doubt that computertechnology will continue to keep
of this typeof structure and also for concrete offshore platforms - large on breaking the price/performance records; thus the cost-effective capability
problems withlong
running-times andan extensive amount of of the finite element method will continue to grow.
postprocessing work all demanding a high degree of analytical skill and Now there is strong emphasis on quality assurance of the software and
good housekeeping. its application. Itwill be important to produce reliable models and reliable
solutionprocedures.Finite element systems need to focus onthe
Non-linear structural analysis measurement of error. The adaptive element approach is moving in this
In the preceding section the problem
was essentially a reinforced concrete direction. It may well be important to recognise that, in the real world,

58 The Structural Engineer/Volume 71 /No.4/ 16 February 1993


Paper:
Carlton
Informal study groups

there are uncertainties in information and thus designers may wellbe


interested in statistical solutions rather than deterministic solutions.
The design Codes or standardsmay well need enhancement in order to
Informal study groups
recognise the sophisticated user. In particular, the British Code for concrete
leaves a lot unstated when dealing with complex force situations.
There is stillroom for improvements with respectto processing the output
data. It may be that the STEP modelling for data exchange will provide
opportunities for CADsystems, modelling systems, finite element systems, The purpose of the Study Group scheme is to create
and postprocessing systems, to utilise a common standard. opportunities for members of the Institution to exchange ideas
The progress of the finite element method is rapid and thus the user and work on deepening and developing their knowledge of
community cannot ignore the necessity to ensure a programme of continuous structural engineering, thus stimulating a greater interest in
education so that analysts and engineers retain their ‘feel’for comprehending and promoting the art and science of structural engineering.
the results. Members wishing to take part in the work of a Study
Group or who require further information about a Study
Concluding remarks Group should write to the appropriate Convener.
With today’s computer technology the finite element method is readily
available to the practicing structural engineer, but it must be used withsome
caution by engineers who appreciate the methods andhave been properly History of Structural Engineering
trained to use it. Convener: Frank Newby, MA(Cantab), FEng, FIStructE, HonFRIBA,
Generally, the outputis not in the most useful form for structural design, 27 Mayfield Avenue, London W4 1PN
and it may therefore benecessary to develop extra postprocessing software The Structural Engineer, March 1973, p1 10
to deal with a large mass of results.
Design standards require careful interpretationin order to ensure that
the complex force field information is takenintoaccount to ensure Model Analysis as a Design Tool
compliance with the standard. Convener: F . K. Garas, PhD, CEng, FIStructE, MICE,
Non-linear analysis of reinforced concrete structures can beuseful but Taylor Woodrow Construction Ltd., Taywood House, 345
should be applied only when it has been possible to compare the output Ruislip Road, Southhall, Middlesex UBI 2QX
from the finite element analysis with test information. The Structural Engineer, February 1977, p63
References
1. Zienkiewicz, 0. C.: ‘Finite elements - the background story’, The Qualitative Analysis of Structural
mathematics of finite elements and applications,
London, Academic
Press 1972,p1 Behaviour
2. Zbirohowski-Koscia, K., Carlton, D.: ‘Analysis of vessel structures Convener: D. Johnson, BSc(Eng), PhD, CEng, FIStructE, MICE,
with particular reference to Wylfa’, Prestressed Concrete Pressure Department of Civil & Structural Engineering, Nottingham
Vessels, London, ICE, 1967,p311 Trent University, Burton Street, Nottingham NGl 4BU
3. Zienkiewicz, 0. C., Cheung, Y. K.: ‘The finite element method for The Structural Engineer, November 1978, p309
analysis of elastic isotropic and orthotropic slabs’, Proc. ICE, 28,
1964,p471
4. Clough, R. W.: ‘The finite element in plane stress analysis’, Proc. The Design of Steel Portal Frames
2nd ASCE Conference on Electronic Computation, Pittsburgh, 1960 Convener: L. J. Morris, BSc(Eng), PhD, ACGI, DIC, CEng, FIStructE,
5 . Otter, J . R. H.:‘Computationsfor prestressed concretereactor Simon Engineering Laboratories, University of Manchester,
pressure vessels using dynamicrelaxation’, NuclearStructural Manchester M1 3 9PL
Engineering, vol. 1, 1965 The Structural Engineer, Part A, June 1983, p1 70
6. Tan, L. H., Bathe, K. J.: ‘Studies of finite element procedures - the
conjugate gradient and GMRES methods in ADINA and ADINA-
F’, Computers and Structures, 40, No.2, 1991, p441 Vibration Problems in Structures
7. ‘The Standard NAFEMS benchmarks’, NAFEM publications,1989 Convener: J. W. Smith, BSc(Hons), PhD, ACGI, CEng, MIStructE,
8. Zhu, J. Z.,Hinton, E., and Zienkiewicz, 0. C.:‘Adaptivefinite Department of Civil Engineering, University of Bristol, Bristol
element analysis with quadrilaterals’, Computers and Structures, 40, BS81TR
No. 5, 1991,p1097 The Structural Engineer, Part A, June 1983, p1 70
9. Zeng,L.F.,Wiberg,N.E.:‘Adaptive h-p proceduresfor high
accuracy finite element analysis of 2-dimensional linear elastic
problems’, Computers and Structures, 42, No. 6, 1992, p869 Computing in structural engineering
10. Halliday, C.: ‘Applying FEA’, Cadence, Ariel Communications Inc., Convener: A. T. Humphrey, CEng, MIStructE, MIMechE,
October 1992,p65 Analysis & Test Division, GEC Research, Marconi Research
11. Patnaik, S. N., et al.: ‘Improved accuracy for finite element structural Centre, West Haningfield Road, Gt. Baddow, Essex
analysis via an integrated force method’,Computers and Structures, The Structural Engineer, March 1987, p83
45, No. 3, 1992,p521
12. Bathe, K. J., Walczak, J . , Welch, A. K., andMistry, N.: ‘Non-linear
analysis of concrete structures’, Computers and Structures, 32, No. Advanced Composite Materials and
3/4, 1989,p563
13. Hu,H-T,Schnobrich,W.C.: ‘Non-linear analysis of cracked
Structures
Convener: P. R. Head, BSc(Eng), ACGI, CEng, MIStructE, MICE,
reinforced concrete’, ACZ Structural Journal, March-April 1990, p199 Maunsell Structural Plastics Ltd., Maunsell House, 160
14. Stevens, N. J . , Uzumeri, S . M., Collins, M. P., and Will, G . T.: Croydon Road, Beckenham, Kent BR3 4DE
‘Constitutive model for reinforced concrete finite element analysis’, The Structural Engineer, Part A, June 1987, p221
ACI Structural Journal, January-February 1991, p49

Management and Maintenance of


Bridges
convener: G. Davison, BSc, CEng, MIStructE,
c/o The Institution of Structural Engineers, 11 Upper Belgrave
Street, London SWlX 8BH
Structural news, 23 January 1990, p4

T h e S t r u c t u r a l E n g i n e e r / V o l u m e 7 1 /No.4/ 1 6 F e b r u a r y 1 9 9 3 59

You might also like