Professional Documents
Culture Documents
Comparison of PTS analyses of RPVs based on 3D-CFD and RELAP52015NED
Comparison of PTS analyses of RPVs based on 3D-CFD and RELAP52015NED
h i g h l i g h t s
a r t i c l e i n f o a b s t r a c t
Article history: The integrity of a reactor pressure vessel related to pressurized thermal shocks is one of the most impor-
Received 24 November 2014 tant issues for the assessment of life time extension of a nuclear power plant. The most critical scenario
Received in revised form 24 April 2015 occurs during cold water injection through the cold leg due to a Loss-Of-Coolant Accident (LOCA). Due
Accepted 3 May 2015
to the difficulties associated with the crack modeling with the three-dimensional finite element method
(FEM), simple geometries and crack configurations are usually employed.
In the present study, a hypothetical medium break LOCA is assumed in one of the hot legs for an adopted
reference design of a two-loop pressurized water reactor. The boundary conditions obtained from RELAP5
calculations are used as input for the three-dimensional computational fluid dynamics simulations in
order to provide three-dimensional temperature distribution for the structural mechanics analysis in
which submodeling and eXtended FEM (XFEM) are applied. The results from these three-dimensional
computations are compared with those from simplified axisymmetric models based on reference data
temperatures.
© 2015 Elsevier B.V. All rights reserved.
1. Introduction accidents like a pipe break in the primary pressure circuit, a stuck-
open valve in the primary circuit that later re-closes, or a break in
The structural integrity of a Reactor Pressure Vessel (RPV) is the main steam line.
usually considered as a limiting factor for the operation time of a The most severe conditions for PTS take place when cold Emer-
nuclear power plant, as it contains the reactor core and the reactor gency Core Cooling (ECC) water is injected inside the cold legs filled
coolant, is crucial for the safe confinement of the radioactive fuel initially with hotter primary water and/or steam. The cold plume
inventory, and is regarded as non-replaceable. Furthermore, RPV will flow into the downcomer and cool the RPV walls causing large
of nuclear power plants are exposed to neutron irradiation, which temperature gradients. Pressurized thermal shock transients lead
causes embrittlement of the ferritic steels and makes the material to high tensile circumferential and axial stresses in the RPV walls,
susceptible to brittle fracture (Odette and Lucas, 1986). these high stress states may produce crack initiation, propagation
One potential risk for the integrity of a RPV is the brittle failure and, in the worst case, even brittle failure. As a consequence, the
due to a Pressurized Thermal Shock (PTS), which occurs in case of RPV has to be assessed against cleavage fracture (Keim et al., 2001;
an emergency cooling of the core, as a Loss-Of-Coolant Accident Qian and Niffenegger, 2013a,b, 2014; Shum et al., 1994).
(LOCA). This is typically associated with a depressurization which, Accurate prediction of the temperature distributions is neces-
in the worst case, can be followed by a re-pressurization. Pres- sary for determination of the stresses resulting from these thermal
surized thermal shocks are produced by a number of events and loads and for assessment of failure probability of the RPV due to
crack propagation at critical locations (Smith, 2010; Boyd, 2012).
However the one-dimensional models, used in thermal hydraulic
∗ Corresponding author. Tel.: +41 563104364. analysis softwares such as RELAP5 or TRACE, cannot realistically
E-mail address: vicente.gonzalez@psi.ch (V.F. González-Albuixech). represent the complex mixing phenomena in the downcomer.
http://dx.doi.org/10.1016/j.nucengdes.2015.05.025
0029-5493/© 2015 Elsevier B.V. All rights reserved.
V.F. González-Albuixech et al. / Nuclear Engineering and Design 291 (2015) 168–178 169
detected flaws with the fracture toughness, KIC , for the whole PTS.
Nomenclature Calculation of the mode I SIF, KI , is generally based on the Linear
Elastic Fracture Mechanics (LEFM) theory. Moreover, the fracture
RPV Reactor Pressure Vessel toughness testing standards use highly constrained specimens and
PTS Pressurized Thermal Shock consequently the application of the LEFM framework can yield to
LOCA Loss-Of-Coolant Accident over-conservative results. However, these results are still of great
CFD Computational Fluid Dynamics benefit for determining the RPV structural integrity for a given crack
FEM Finite Element Method geometry (Keim et al., 2001; Qian and Niffenegger, 2013b, 2014;
XFEM eXtended Finite Element Method Shum et al., 1994; Zhu and Joyce, 2012).
LEFM Linear Elastic Fracture Mechanics In the present work a fracture mechanic analysis of a realistic
SIF Stress Intensity Factor RPV for a reference Pressurized Water Reactor (PWR) is performed.
PWR Pressurized Water Reactor A Medium size LOCA (MLOCA) scenario according to a 70 cm2 leak
MLOCA Medium Loss-Of-Coolant Accident in the hot leg is chosen because it results in very rapid cooling
BPG Best Practice Guidelines in addition to high system pressure loads on the RPV and cor-
FM Fracture Mechanics responds to the worst situation for the RPV structural integrity
RCP Reactor Coolant Pumps analysis. After defining the initial and boundary conditions from
SIP Safety Injection Pumps the RELAP5 system code calculations, the CFD analysis of PTS is
FE Finite Element performed with state-of-the-art computational methods and fol-
a semielliptical crack depth [mm] lowing as close as possible the Best Practice Guidelines (BPG) for the
c semielliptical crack length [mm] application of CFD in nuclear safety analysis (Mahaffy et al., 2007).
E elastic modulus [MPa] The obtained transient results for the three-dimensional tempera-
h height of modeled RPV geometry or simplified RPV ture distributions are then extracted from CFD for the subsequent
geometry [m] Fracture Mechanics (FM) analysis. The fracture mechanic analysis
KI mode I linear elastic stress intensity factor relies on submodeling and XFEM allowing the analysis of differ-
[MPa m0.5 ] ent crack geometries and locations without complicated remeshing
Ri inner radius of the modeled RPV geometry [m] techniques. The same analysis is repeated for the same PTS tran-
t thickness of RPV wall [mm] sient using the temperature distribution obtained from reference
data on a simplified mechanical model with axisymmetric load
conditions.
Computational Fluid Dynamics (CFD), on the other hand, is able The calculated SIFs are compared to those resulting from sim-
to take into account the details of geometry and to predict multi- plified models showing that the core region simplified model with
dimensional features of the mixing process between the ECC homogeneous cooling yields to non-conservative results. Conse-
subcooled water and the primary hot water or two-phase mixture quently if only a simplified model and boundary conditions are used
present in the cold legs and the downcomer. However CFD models the results cannot be considered as over-conservative as the tran-
are usually very time consuming from the computational point of sient effect is underestimated, thus for structural integrity analysis
view. Therefore, simplified models with an axisymmetric thermal a three-dimensional model as realistic as possible should be used.
hydraulic approximation are used for the PTS temperature evolu-
tion during the transient instead of using the three-dimensional
plume information. 2. Computational fluid dynamics model
The mechanical model for RPV analysis is also very often
reduced to a three-dimensional axisymmetric model (or even to The RELAP5 (SCIENTECH Inc., 1999) system code was used to
a two-dimensional model) containing only the characteristic prop- investigate different LOCA transient scenarios assuming breaks
erties of the wall, due to the difficulties associated to a complete with varying sizes in each test for the cold and hot legs. In the cur-
realistic three-dimensional modeling. Furthermore, these simpli- rent study, the RELAP5 results for a break size of 70 cm2 in the hot
fied models are usually assumed as over-conservative (Qian and leg are used to extract the reference initial and boundary condi-
Niffenegger, 2013b, 2014; González-Albuixech et al., 2014a,b). tions for the CFD calculations. A hypothetical break in the hot leg is
However, if more detailed temperature information is used, it is not more severe for PTS than a break in cold leg, because in the latter
clear if such simplifications and symmetry assumptions are valid. case part of the ECC water would flow out of the break and will
The numerical structural integrity analysis of the RPV usually not reach the downcomer. After the break, the system pressure
relies on the modeling of the RPV and the crack within the Finite Ele- decreases quickly and the Reactor Coolant Pumps (RCP) are postu-
ment Method (FEM) framework. The necessity of a mesh that adapts lated to be tripped with the actuation of a pressure signal. The high
to the geometry of the RPV and also to the crack topology imposes pressure safety injection water is injected in each cold leg from the
some limitations, which simplified models entail. However, some Safety Injection Pumps (SIP) at a temperature of 30 ◦ C. The pressure
new techniques have been recently developed that allow more then reaches a value of 6.9 MPa at which additional safety water is
comprehensive analysis of the RPV, like the eXtended Finite Ele- injected from two accumulators connecting the cold legs at differ-
ment Method (XFEM; Duflot, 2006; Gravouil et al., 2002; Moës et al., ent lines before the system pressure continues to decrease again.
1999, 2002; Sukumar et al., 2000). Furthermore, XFEM was recently In the current study, only the accumulator connected to loop B (see
implemented in the commercial finite element code ABAQUS, Fig. 1) is assumed operational which leads to asymmetric cooling
Hibbitt et al. (2013) and already successfully applied to study a sim- conditions. This was found to create larger thermal loads than in
plified RPV model (González-Albuixech et al., 2014a,b). A submodel case of symmetric injection conditions (Sharabi et al., 2014).
analysis technique is linked to the complete three-dimensional RPV At the start of the ECC injection, the flow in the loops was almost
model study, as a complete detailed structural integrity analysis is stagnant. The ECC from SIP reaches a value of 80 kg/s in each loop
not feasible due to the mesh refinement required for studying local and it remains constant for a long period of time. The flow rate from
effects as cracks. the accumulator has an average value of 200 kg/s and the temper-
The integrity analysis of RPVs is mostly based on the compar- ature of the injected water was assumed to be 10 ◦ C. In order to
ison of the mode I stress intensity factor (SIF) of postulated or save computation time, the CFD calculations were launched at the
170 V.F. González-Albuixech et al. / Nuclear Engineering and Design 291 (2015) 168–178
conditions relevant to PTS phenomenon; i.e., with the start of the Exact representation of the inlet nozzle to the RPV is consid-
ECC injection in the cold legs and by taking the initial conditions at ered in the model. In particular, the curved connection between
that time instant from RELAP5 results. It was assumed that both the the inlet nozzle and the RPV is taken into account because it has a
SIP and the accumulator inject the emergency cooling water at the strong influence on both the flow separation, and on the mechanical
same time in order to start the CFD calculations from quasi-steady integrity of this area. The hot leg is not considered in the com-
state conditions where the flow in the loops approaches zero. The putational domain and the outlet boundary is placed at the RPV
initial temperature was set to 285 ◦ C and it is assumed uniform outlet nozzle. For the cold legs the computational domain extends
everywhere in the system. The void fraction in the loops and in the for distance of two pipe diameters (2D) upstream of the injection
downcomer was zero and the flow in the loops remains in single locations. In actual situations, part of the injected water will flow
phase during the transient. The initial and boundary conditions are to the pump volume at the beginning of the injection; however,
summarized in Table 1. as a conservative assumption, the pump volume is not modeled
A CAD file is constructed for the RPV for a reference design of and all the flow is assumed to be eventually directed towards the
the two-loop PWR nuclear power plant. Detailed description of downcomer. Structural details in the core are not included in the
the adopted computational domain is shown in Fig. 1. The model numerical simulations but a simplified porous media approach is
comprises a three-dimensional full description of the reactor pres- adopted. Because the ECC flow is much lower than the flow rate in
sure vessel with the injection lines connected to the cold legs. The normal operation, the pressure drop in the core is very small and
ECC injection lines are connected at right angle to the sides of the accordingly, it is not expected that structural details in the core
cold legs with the SIP and accumulator lines opposing each other. would play an important role on mixing in the downcomer.
The neutron shield is located in the active core area between the The CFD simulation for the case reported in Table 1 is performed
core barrel and RPV walls which provides shielding for the com- using ANSYS Fluent 15.0 CFD code (ANSYS Inc., 2013). The mesh
plete vessel 360◦ circumferential area. It represents an obstacle is built exclusively with blocked-structured hexahedral elements
to the flow in the downcomer and it has a significant influence following the BPG (Mahaffy et al., 2007) in order to have the flow
on the flow field and accordingly on the mixing characteristics of field aligned with the mesh in major parts of the domain. Proper
the cold water falling down in the downcomer. Different details of mesh refinements are needed in the zones relevant to PTS; i.e., in
wall materials for the cold legs, RPV, core barrel and neutron shield the cold legs, downcomer and the lower plenum. Structured mesh
are represented in the computation with their detailed geometrical requires that the same mesh refinement spans through the whole
information. domain which will create excessive mesh numbers by placing fine
mesh in large volume zones like the reactor core and the upper
plenum where the details of the flow field is not very important.
In order to avoid this problem, mesh interfaces are created at non-
Table 1
Initial and boundary conditions for the CFD model. critical faces with different mesh refinements on each side. A mesh
interface is created at the inner wall of the core barrel where it
Loop A Loop B
connects the solid zone of the core barrel with the fluid zone of the
Accumulator, m [kg/s] 0 200 core region for coupled heat transfer calculations as shown in Fig. 2.
SIP, m [kg/s] 80 80 Another mesh interface is placed at an interior flow surface at the
Accumulator, T [◦ C] – 10
SIP, T [◦ C] 30 30
downcomer outlet.
Cold leg, m [kg/s] 0 0 The mesh has been created using ICEM CFD 15.0 (ANSYS Inc.,
Initial pressure [MPa] 6.9 2013) and it is composed of 5.1 million hexahedral cells. Most of
Initial temperature [◦ C] 285 the mesh cells are concentrated in the cold legs, downcomer and
Void fraction [dimensionless] 0
the lower plenum. The maximum aspect ratio is about 120 where
V.F. González-Albuixech et al. / Nuclear Engineering and Design 291 (2015) 168–178 171
Table 2
Thermo-mechanical properties of the RPV material.
Fig. 5. Temperature evolution on the inner RPV wall, details of the plume.
Fig. 6. Global von Mises equivalent stresses evolution on the inner RPV wall for the first 540 s.
V.F. González-Albuixech et al. / Nuclear Engineering and Design 291 (2015) 168–178 173
Fig. 7. Local stress components at 399 s. The maximum values corresponding to the location of postulated cracks as indicated by an ‘x’.
Fig. 10. Submodel mesh and crack details: (a) Inlet nozzle, (b) Wall region.
A detailed submodel is built for the two regions of the RPV and axial cracks, the scheme of each orientation is shown in
indicated in Fig. 8 (the core region and the inlet nozzle of loop Fig. 10.
B) as can be seen in Fig. 10. The sizes of the submodel are The integrity analysis relevant parameter, KI , at the deepest
at least 5 times the crack size to minimize boundary effects. crack point, is obtained by calculating the interaction integral
The submodels have linear hexahedron mesh which is refined implemented in ABAQUS, Hibbitt et al. (2013).
where the cracks are located. Due to accuracy reasons the region
surrounding the crack is a regular mesh with an element size 5. Simplified model
smaller to 1/10 of the crack front curvature where the SIF
is extracted (González-Albuixech et al., 2013a,b, 2014a). Two In the following, a simplified model that considers only the belt
crack orientations are tested for each location, circumferential region of the RPV, which is irradiated by higher neutron fluence,
V.F. González-Albuixech et al. / Nuclear Engineering and Design 291 (2015) 168–178 175
Fig. 11. Simplified RPV model geometry for axisymmetric boundary conditions.
Fig. 12. Overview of the mesh and crack details for axisymmetric model: (a) axial crack, (b) circumferential crack.
is used in order to be compared with the full three-dimensional An overview of the mesh of linear hexahedron finite elements
calculations of the whole RPV described in Section 4. The model used in the analysis is presented in Fig. 12. The model assumes
contains the most important geometric properties, as shown in symmetries whenever possible, but the XFEM implementation in
Fig. 11. The inner side of the RPV is assumed to be subjected to a ABAQUS (Hibbitt et al., 2013) does not allow symmetry planes
thermal shock caused by the cold plume from the emergency cool- which contain crack planes. The analysis is similar to those
ing water. Although the cooling is not homogeneous but instead performed in González-Albuixech et al. (2014a). The mesh is
typically has a plume shape coming from the inlet nozzles down to surrounding the crack regularly and very refined in order to obtain
the core region, the simplified numerical model assumes rotation- accurate SIF using XFEM (González-Albuixech et al., 2013a,b).
ally symmetric and homogeneous cooling along the z-axis. The element size in this area is the same as that used in the 3D
176 V.F. González-Albuixech et al. / Nuclear Engineering and Design 291 (2015) 168–178
20
300 MLOCA
Water temperature [° C]
MLOCA
Pressure [MPa]
250 15
200
10
150
5
100
50 0
(a) (b)
MLOCA
Heat transfer coefficient of
25
water-RPV [kW/m ⋅K]
20
2
15
10
0
0 6000 12000
Time [second]
(c)
Fig. 13. MLOCA transient: (a) water temperature history, (b) pressure history, (c) water heat transfer coefficient history.
Fig. 14. SIF comparison, three-dimensional model with temperature calculated by CFD and simplified model with axisymmetric temperature distribution taken from ref.
data: (a) axial crack, (b) circumferential crack.
V.F. González-Albuixech et al. / Nuclear Engineering and Design 291 (2015) 168–178 177
Acknowledgment
Fig. 15. Comparison between SIFs of axial and circumferential cracks postulated at
The authors are grateful for the financial support of the PISA
the inlet nozzle and at the inner wall, at a distance of 0.35 h below the top.
Project provided by the Swiss Federal Nuclear Safety Inspectorate
(ENSI) (DIS-Vertrag Nr. H-100668).
submodels to allow a better comparison. The same mesh is used
in the heat transfer and in the structural mechanics calculations. References
The history of the water temperature, pressure and heat trans-
fer coefficient between water and inner wall of the RPV for the ANSYS Inc, 2013. FLUENT 15.0 theory guide.
Apanasevich, P., Coste, P., Niceno, B., Heib, C., Lucas, D., 2014. Comparison of CFD
analyzed transient using reference data corresponding to the core simulations on two-phase pressurized thermal shock scenarios. Nucl. Eng. Des.
region are shown in Fig. 13 (González-Albuixech et al., 2014a; 266, 112–128.
Qian and Niffenegger, 2013b). Note that the temperature history Boyd, C., 2012. Pressurized thermal shock. In: Proceedings of the THICKET-3.
Duflot, M., 2006. A study of the representation of cracks with level sets. Int. J. Numer.
is intended to be representative for the cooling plume.
Methods Eng. 70 (11), 1261–1302.
González-Albuixech, V.F., Giner, E., Tarancón, J.E., Fuenmayor, F.J., Gravouil, A.,
2013a. Convergence of domain integrals for stress intensity factor extraction
6. Analysis
in 2-D curved cracks problems with the extended finite element method. Int. J.
Numer. Methods Eng. 94, 740–757.
In Fig. 14 we compare the mode I SIF, evaluated with the full González-Albuixech, V.F., Giner, E., Tarancón, J.E., Fuenmayor, F.J., Gravouil, A.,
three-dimensional model in which CFD temperature results are 2013b. Domain integral formulation for 3-D curved and non-planar cracks with
the extended finite element method. Comp. Methods Appl. Mech. Eng. 264,
considered, with those from the simplified model using axisym- 129–144.
metric temperature distribution taken from reference data. The González-Albuixech, V.F., Qian, G., Niffenegger, M., 2014a. Integrity analysis of reac-
simplified model underestimates the SIF in comparison with the tor pressure vessels subjected to pressurized thermal shocks by XFEM. Nucl.
Eng. Des. 275, 336–343.
three-dimensional model for the circumferential crack and over- González-Albuixech, V.F., Qian, G., Niffenegger, M., 2014b. Integrity analysis of a
estimates it for the axial crack. It is shown that the simplified reactor pressure vessel with quasi laminar flaws subjected to pressurized ther-
model is not able to capture the behavior associated with the mal shocks. Nucl. Eng. Des. 280, 464–472.
Gravouil, A., Moës, N., Belytschko, T., 2002. Non-planar 3D crack growth by the
realistic three-dimensional model. Furthermore, as the simplified extended finite element and level sets—Part II: level set update. Int. J. Numer.
model underestimates some results in comparison with the realis- Methods Eng. 53, 2569–2586.
tic three-dimensional model based on CFD temperature results, in Hibbitt, Karlsson, Sorensen, 2013. Abaqus 6.13 manual.
Keim, E., Schmidt, C., Schöpper, A., Hertlein, R., 2001. Life management of reactor
general it cannot be seen as conservative. The reason for the dif-
pressure vessels under pressurized thermal shock loading: deterministic proce-
ference in SIF of the full 3D model and the simplified axisymmetric dure and application to western and eastern type of reactors. Int. J. Press. Vessels
model is due to different temperatures and to their axisymmetric Pip. 78, 85–98.
Mahaffy, J., et al., 2007. Best Practice Guidelines for the Use of CFD in Nuclear Reactors
distribution in the simplified FE-model.
Safety Applications. NES/CSNI/R(2007)5, Paris.
A complete three-dimensional model with the temperature field Menter, F.R., 1994. Two-equation eddy-viscosity turbulence models for engineering
obtained from realistic CFD computations allows also for compar- applications. AIAA J. 32 (8), 1598–1605.
ing SIFs of cracks placed at various locations on the RPV, as shown in Moës, N., Dolbow, J., Belytschko, T., 1999. A finite element method for crack growth
without remeshing. Int. J. Numer. Methods Eng. 46, 131–150.
Fig. 15, for different crack orientations. Not surprisingly the crack Moës, N., Gravouil, A., Belytschko, T., 2002. Non-planar 3D crack growth by the
postulated at the inlet nozzle corner shows the largest SIF for an extended finite element and level sets—Part I: mechanical model. Int. J. Numer.
axial crack. Methods Eng. 53, 2549–2568.
Odette, G.R., Lucas, G.E., 1986. In: Steele, L.E. (Ed.), Irradiation Embrittlement of
Reactor Pressure Vessel Steels: Mechanisms, Models and Data Correlations,
7. Conclusions Radiation Embrittlement of Reactor Pressure Vessel Steels—An International
Review. ASTM STP 909, Philadelphia, pp. 206–241.
Qian, G., Niffenegger, M., 2013a. Procedures methods and computer codes for the
Fracture mechanics analyses of a RPV model subjected to PTS probabilistic assessment of reactor pressure vessels subjected to pressurized
loads were performed using results from the thermal-hydraulic thermal shocks. Nucl. Eng. Des. 258, 35–50.
Qian, G., Niffenegger, M., 2013b. Integrity analysis of a reactor pressure vessel sub-
system code RELAP5, reference data and extensive CFD sim-
jected to pressurized thermal shocks by considering constraint effect. Eng. Fract.
ulations, followed by three-dimensional FE calculations. Such Mech. 112–113, 14–25.
three-dimensional calculations allow the realistic consideration Qian, G., Niffenegger, M., 2014. Deterministic and probabilistic analysis of a reactor
of local cooling effects (e.g., due to the cold water plume) and pressure vessel subjected to pressurized thermal shocks. Nucl. Eng. Des. 273,
381–395.
therefore a more detailed analysis than with simplified models SCIENTECH Inc., 1999. RELAP5/Mod3 Code Manual, Vol. I: Code Structure, System
which are often used for economic reasons. The comparison of the Models and Solution Methods. The Thermal Hydraulics Group, Idaho.
178 V.F. González-Albuixech et al. / Nuclear Engineering and Design 291 (2015) 168–178
Sharabi, M., Niceno, B., Gonzalez-Albuixech, V.F., Niffenegger, M.,2014. Computa- Smith, B.L., 2010. Assessment of CFD codes used in nuclear reactor safety simula-
tional fluid dynamics of pressurized thermal shock phenomena in the reactor tions. Nucl. Eng. Technol. 42 (4), 339–364.
pressure vessel. In: Proceedings of the CFD4NRS-5. ETH-Zurich, Switzerland, pp. Sukumar, N., Moës, N., Moran, N., Belytschko, T., 2000. Extended finite element
9–11, Sept. method for three-dimensional crack modelling. Int. J. Numer. Methods Eng. 48,
Scheuerer, M., Weis, J., 2012. Transient computational fluid dynamics analysis of 1549–1570.
emergency core cooling injection at natural circulation conditions. Nucl. Eng. Willemsen, S.M., Komen, E.M.J., 2005. Assessment of RANS CFD modelling for pres-
Des. 253, 343–350. surized thermal shock analysis. In: Proceedings of the NURETH-11, Avignon, 2–6
Shum, D.K., Theiss, T.J., Rolfe, S.T., 1994. Application of J-Q fracture methodology to Oct.
the analysis of pressurized thermal shock in reactor pressure vessels. Fracture Zhu, X.K., Joyce, J.A., 2012. Review of fracture toughness (G, K, J, CTOD, CTOA) testing
Mechanics, 24th. ASTM STP 1207, Philadelphia, pp. 152–168. and standardization. Eng. Fract. Mech. 85, 1–46.