Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page1of27.
1
2
Analytical Bounds for Efficient Crack Growth Computation
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Abstract
24
25
26
27
28
29
30
31
32
33
34
In linear elastic fracture mechanics, the rate of crack propagation is proportional to the range of
stress intensity factors. The most popular model relating these quantities is the Paris-Erdogan law.
Crack growth computation is an initial value problem whose solution cannot be obtained in closed
form, as stress intensity factors, hence crack growth rates, depend on the accumulated growth. For
complex geometries, stress intensity factors are evaluated numerically, and crack growth
computations can become computationally intensive. This paper presents a theoretical result
establishing upper and lower bounds for the crack size function for any number of cycles. The
bounds are very narrow, hence accurate crack size approximations can be obtained from only two
stress intensity factor evaluations. This leads to a huge gain in computational effort for numerical
crack growth computations. Two examples are used herein to explore the accuracy and efficiency of
the proposed solution for the crack growth initial value problem.
NuMAT/PPGEM, Federal University of Technology of Parana
Av. Sete de Setembro, 3165, Curitiba, PR, Brazil.
2
PPGMNE/CESEC, Federal University of Parana
Centro Politcnico, Jardim das Amricas, C. P. 19011
81531-980, Curitiba, PR, Brazil.
Rodrigo Villaca Santos,
[email protected]NuMAT/PPGEM, Federal University of Technology of Parana
Av. Sete de Setembro, 3165, Curitiba, PR, Brazil.
DAMEC, Federal University of Parana
Via do Conhecimento, Km 1, Pato Branco, PR, Brazil.
Andr Tefilo Beck,
[email protected]Structural Engineering Department, EESC, University of So Paulo
Av. Trabalhador Sancarlense, 400, So Carlos, SP, Brazil.
35
36
Keywords
37
38
39
Fracture mechanics, Crack size; Paris-Erdogan law;Initial value problem;
40
41
Nomenclature
FE
H
Finite Element
Hypotheses
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page2of27.
IVP
RK4
a
a0
42
43
44
45
Initial Value Problem
Fourth-Order Runge-Kutta Method
Crack Size
Initial Crack Size
Rate of Crack Propagation
46
da
dN
47
48
49
50
51
52
53
54
55
56
f(a)
C
m
N
Geometry Function
Material Constant (Paris Law)
Material Constant (Paris Law)
Number of Load Cycles
Stress Intensity Factor
Critical Stress Intensity Factor
Stress Range
Relative Deviation Function Upper Bound
Relative Deviation Function Lower Bound
Polynomial coefficient
57
Polynomial coefficient
Polynomial coefficient
Kc
UB
LB
58
59
60
61
62
Introduction
In linear elastic fracture mechanics, the rate of crack propagation
da is assumed proportional
( dN
)
63
to the range of stress intensity factors, ( K ) . In Figure 1, Region I represents near-threshold crack
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
growth. Region II represents intermediate crack propagation, where crack propagation rate is linearly
proportional to K (on a log-log scale) and where a small plastic zone appears ahead of the crack
tip. Region II is the range of application of linear elastic fracture mechanics. Finally, Region III
accounts for the accelerated crack growth just prior to failure. Non-linear fracture mechanics
concepts are required to model crack growth in this region.The Paris-Erdogan [1]crack propagation
equation describes crack growth in the linear region (Region II in Figure 1).
Due to its simplicity (only two parameters need to be identified experimentally), the Paris
equation is widely employed in many applications. However, this equation presents some major
limitations: a. it only represents crack growth in the so-called region II (linear); b. mean stress effects
are not taken into account and c. it does not take into account the loading history and resulting load
interaction effects.Several variants of the Paris equation have been developed to address these
particular aspects. This includes the model of Elber [2], which uses an equivalent stress intensity
factor to take into account crack closure under compressive stresses; and Forman [3], which includes
mean stress and stress ratio effects. Many other models are available in the literature; however, the
very existence of a myriad of crack propagating equations is evidence that no single equation is
overwhelmingly better than the others.
In order to simulate crack growth in components or structureswith complex geometry,
numerical analysis has been used extensively in recent years. This includes computation of stress
intensity factors by the Boundary Element Method [4-6], Generalized FE methods [7-9], and
Extended Finite Element Method [10-13], including cohesive crack modeling. Numerical
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page3of27.
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
computation of stress intensity factors and complex re-meshing schemes add significantly to the
computational burden of crack growth estimation. This is especially the case when random crack
propagation and/or reliability analysis is considered [14-19]; as uncertainty propagation studies
increase the computational effort significantly.
Considering the aforementioned computational burden, this paper presents a theoretical result,
in the form of a theorem, which allows to evaluateupper and lower bounds for crack growth based on
only two evaluations of the stress intensity factor. In this first study, only the Paris-Erdogan crack
propagation equation is considered. The bounds derived herein are based on polynomial functions of
the number of load cycles. The examples presented herein consider problems with analytical
geometry functions, but results can be readily extended to more general problems where stress
intensity factors are computed numerically.
The remainder of this article is organized as follows. The Paris-Erdogan crack propagation
equation is presented in Section 2. The theorem providing the crack growth bounds and its proof are
presented in Section 3. In Section 4, two example problems are addressed. Concluding remarks are
presented in Section 5.
99
100
Paris-Erdogan Crack Propagation Equation
101
Using the Paris-Erdogan crack propagation model, the crack growth problem can be written as,
102
Find a C1 ( N 0 , N1 ) , such that :
m
da
dN ( N ) = C K ( a ( N ) ) , N ( N 0 , N1 ) ;
a ( N0 ) = a0 ;
( )
(1)
103
whereCandmare material constants and Nrepresents number of load cycles. The term K ( ) is the
104
range of stress intensity factors, defined as,
K ( a ( N ) ) = ( K max K min ) ( a ( N ) )
= a ( N ) f ( a ( N ) ) ( max min )
105
(2)
= a ( N ) f ( a ( N ) ) .
106
In Eq. (2), f ( ) is the geometry function and is the far-field stress range. The problem
107
108
stated in Eq. (1) is classified as an initial value problem (IVP). Replacing Eq. (2) in Eq. (1), one
obtains,
109
Find a C1 ( N0 , N1 ) such that :
da
dN ( N ) = C a ( N ) f a ( N )
a ( N0 ) = a0 .
( )
))
, N ( N0 , N1 ) ;
(3)
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page4of27.
110
111
112
Eq. (3) is an IVP, defined by an ordinary, first order, non-linear autonomous differential
equation. More generally, Eq. (3) can be classified as a Cauchy problem, which consists in finding
the trajectories which satisfy the IVP differential equation and the initial value, ( a ( N0 ) = a0 ) . The
114
existence and uniqueness of the solution to this Cauchy problem depend on certain regularity
conditions [20]on the functions at the right of Eq. (3). Interval ( N 0 , N1 ) corresponds to load cycle
115
( )
116
117
118
119
120
Lower and upper bounds for the crack size function are determined in this section based on the
following hypotheses about the loading and the geometric correction function:
113
numbers, which are within RegionII in Fig. (1).
Lower and Upper Bounds for Crack Size Function
H1: ( N ) = 0 , N [ N 0 , N1 ] ;
121
f C1 ([ a0 , a1 ] ; + \ {0}) ;
H2 : 0 < f ( a0 ) f ( x ) f ( y ) , x y, x, y [ a0 , a1 ] ;
f ( a0 ) f ( x ) f ( y ) , x y, x, y [ a0 , a1 ] ;
H3: m 2;
(4)
122
where a0 = a ( N 0 ) and a1 = a ( N1 ) . Hypotheses (H1) assumes constant loading. Hypothesis (H2) states
123
124
125
126
127
that the geometry function must be a monotonously non-decreasing function, and that its derivative
must also be a monotonously non-decreasing functions. These conditions are met for most common
geometry functions, since they represent an intrinsic characteristic of the crack propagation problem.
Hypotheses (H1) and (H2) form the basis for the following theorem, which defines upper and lower
bounds for the crack size function.
128
Theorem: Let f ( ) and ( ) be functions which satisfy hypotheses (H1) and (H2), respectively, and
129
a [ a0 , a1 ] , the following lower and upper bounds are valid:
K ( a ) m + ( mC ) K a 2.m
( )
0 )
2
(
a
N
C
(
)
( N N0 ) ;
0
2a + f ( a ) ( N N0 )
1 + ( mC ) ( K ( a ) )m
0
2
m
a ( N ) a0 C ( K ( a0 ) )
( N N0 ) , N [ N0 , N1 ] .
f
21a0 + f ( a0 ) ( N N0 )
130
( )
( )
131
132
(5)
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page5of27.
133
Proof:From hypothesis (H2) and a second-order Taylor expansion of crack size, around a0 , one
134
obtains,
135
136
137
da
a ( N ) = a0 + ( dN
( N0 ) ) ( N N0 ) + 12
d 2a
dN 2
( ) ) ( N N0 )
, with [ N0 , N ] .
The second-order term is called the Lagrange rest. Naturally, and from hypothesis(H2), the
following inequalities can be written:
m
138
a ( s ) a ( t ) , s t with s,t [ N0 , N ] ( a ( s ) ) ( a ( t ) ) .
139
From hypothesis (H2) one obtains,
140
( f ( s )) ( f (t ))
141
Hence, oneconcludesthat,
142
( K ) ( a ( s ) ) ( K ) ( a ( t ) ) ,
143
Now, since C > 0 oneobtains,
144
( )
a2 . f
da
dN
( s) (a
1
2
.f
(t ) , s t
with s,t [ N0 , N ] .
with s,t [ N 0 , N ] .
( s ) dNda ( t ) , s t
145
The second derivative of the crack size function is evaluated as,
146
d 2a
d da
d da
da ( a ( N ) )
a N )) =
a ( N )) =
( a ) dN
(
2 ( (
dN
dN
dN
dN
da
2m d
m
m
m
m
= C 2
a 2 ( f ( a )) a 2 ( f ( a ))
da
)(
2m
= mC 2
)
(
1 a m2 1 ( f ( a ) )m + a m2 ( f ( a ) )m1 f ( a ) a m2 ( f ( a ) )m
2
m m1
) ( a ( f ( a )) )
2m
12 f ( a ) + af ( a ) .
)
(7)
Inserting Eq. (7) in Eq. (6), the Taylor-expanded crack size function becomes,
a ( N ) a0 = C ( K ( a0 ) )
148
149
150
151
= mC 2
147
(6)
2a1( ) +
( N N0 ) + mC2
( K ( a ( ) ))
( ) ( a ( ) ) ( N N )
f
f
2m
(8)
, with [ N0 , N ] .
The lower and upper bounds are obtained from a reformulation of the Lagrange restin Eq.(8).
From the behavior of the geometric correction function in hypothesis (H2), the following inequality
is proposed,
m m1
152
( a ( f ( a )) )
153
Similarly,
154
( a ( f ( a )) )
m m1
12 f ( a ) + af ( a ) a f ( a )
2m
12 f ( a ) + af ( a ) a0 f ( a0 )
2m
1 +
2a
( ) ( a ) , a [a , a ] .
(9)
1 +
2 a0
( ) ( a ) , a [ a , a ] .
(10)
f
f
f
f
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page6of27.
155
Replacing Eq. (9) in Eq. (7), one obtains the following estimate,
156
d 2a
( ) mC 2 K ( a )
dN 2
157
The function leading to the lower bound is obtained by inserting Eq. (10) in Eq. (7),
2m
1 +
2a
2m
d 2a
mC 2 ( K ( a0 ) ) 21a0 +
2 ( )
dN
158
( ) ( a ) , [ a , a ] .
f
f
(11)
( ) ( a ) , [ a , a ] .
f
f
(12)
159
160
161
162
ReplacingEqs. (11) and (12) in Eq. (8), one obtains the upper and lower bounds stated in the
theorem.
It is noted that the lower and upper bounds for crack size, stated in the theorem, depend only
on the evaluation of the geometryfunction and its derivative at two points, a0 anda*. Obtaining the
163
coefficients of Eq. (5) only requires the evaluation of f ( a0 ) , f ( a0 ) , f a , f a
164
165
166
167
computational cost for evaluating the bounds proposed in this paper is a small fraction of the
computational cost for the traditional, cycle-by-cycle integration of Eq. 1. Clearly, the computed
bounds depend on a proper choice for the value a*. It is proposed that traditional engineering insight
be used to select this value. For instance, the critical stress intensity factor, ( K C ) , could be used to
168
obtain a*. In the numerical examples presented herein, narrow bounds are obtained for a*=2 a0.
169
170
171
172
173
174
175
176
177
178
In order to evaluate the accuracy of the upper and lower bounds proposed in this paper (Eq. 5),
the IVP in Eq. (3) is solved for two example problems. Since the differential equation is autonomous,
Eq. 3 is separable; hence the crack size for any number of cycles can be obtained by direct
integration, starting at the initial value. Unfortunately, analytical integration is possible only for a
limited number of problems, for which the geometry function is analytical. For practical problems,
numerical computation of stress intensity factors and numerical integration is often required. In the
examples below, the classical fourth-order Runge-Kutta method (RK4) is employed:
NumericalExamples
Find ak +1 + , such that :
N
ak +1 = ak + ( 6 ) .( K1 + 2 K 2 + 2 K3 + K 4 ) , k {0,1,..., n} ;
m
K1 = C ak f k ;
N
K 2 = ak + ( 2 ) .K1 ;
N
K3 = ak + ( 2 ) .K 2 ;
K = a + ( N ) .K ;
k
3
4
a0 = a ( N 0 ) ;
179
( ) ( )} . Hence, the
(13)
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page7of27.
180
where ak = a ( N k ) , f k = f ( ak ) and N = N k +1 N k , k {0,1,..., n} . The RK4 method has precision
181
of fourth order, O N 4
182
to evaluate the accuracy of the numerical RK4 solution. Figure 2 shows the relativeerror,
183
( ( N ) = 100.(
184
numerical solutions ( a RK 4 ( ) ) to the problem of propagation of a crack in an infinite plate, with the
185
parameters given in Eq. (14).
( (
) [ % ] , N
{0, 50,100, ..., 90000} ,between the exact a ( ) and approximate
(14)
The lowerand upper bound crack size functions are second-degree polynomials in the number
of cycles (N),and are defined as follows:
aLB ( N ) = N 2 + N + a0 ;
2
5
aUB ( N ) = N + N + a0 , N 0,9 10 .
189
190
191
192
a( N k )
C = 109 ; m = 2; N 0 = 0; N1 = 9 10 5cycles;
a = 0.005 in; = 20 ksi, N 0,9 10 5 .
0
186
187
188
a RK 4 ( N k ) a ( N k )
) ) . For example one below, which has exact analyticalsolution, it is possible
(15)
For the two examples, the coefficients of these polynomials are presented in Table 1, whose
definitions are given in Eq. (5). The accuracy of upper and lower bounds, proposed in Eq. (5), is
measured by the following relative deviationfunction,
UB ,LB ( N k ) = 100.
193
aUB ,LB ( N k ) a RK 4 ( N k )
a RK 4 ( N k )
) [% ] , N {0, 50,100, ..., 9 10 } .
5
(16)
194
195
a.
196
197
Figure3 shows the geometrical configuration of the finite width center cracked plate considered
in Example 1. The geometryfunction is given by
f ( a ( N ) ) = sec
198
199
200
201
202
Example1: Finite width plate with center crack
a ( N )
2b
),
N 0, 9 105 .
(17)
Results of crack size function, obtained by the RK4method, and the computed upper and lower
bounds, are shown in Figure 4. Figure 5 shows the relative deviation of upper and lower bounds
w.r.t. the exact solution.In Figures 4 and 5 one observes that the bounds proposed herein are quite
narrow, providing accurate estimates of the actual crack size function.
203
204
b.
Example2: Finite plate with corner cracks
205
206
Figure 6illustrates a finite width plate with two corner cracks, with initial crack size a0 and
widthb=1/2 in.The geometry function for this example is,
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page8of27.
( )
( )
21.72 ( ( ) ) + 30.39 ( ( ) ) , N 0,9 10 .
f ( a ( N ) ) = 1.122 0.231
207
a( N )
b
+ 10.55
a N
b
a( N )
b
a N
b
(18)
213
Figure 7 shows the numerical solution through the RK4 method for the crack sizefunction, as
well as the computed upper and lower bounds. Again, the bounds obtained with the proposed
methodology are quite narrow, providing accurate estimates of the actual crack size.
Figure 8 shows the relative deviation of upper and lower bounds w.r.t. the exact result. A
comparison of Figures 5 and 8 reveals that, for this example, the relative deviation function for the
lower bound, ( LB ( ) ) assumed larger values. On the other hand, the relative deviation for the upper
214
bound function was smaller, for N = 9.105 cycles: UB ( 9 10 5 ) = 10.1597 %
215
UB ( 9 105 ) = 6.9573 % (example 2).
216
217
c.
208
209
210
211
212
218
219
220
(example 1),
Evaluation of the results
In both examples it is observed that the lower boundleads to smaller relative deviation,hence
to more accurate yet unconservative approximations. Table 2 shows the maximum and minimum
values of relative deviation for the upper and lower bounds, ( max ( UB ) , min ( LB ) ) , for both
230
examples. In particular, it can be seen in Table 2 that the relative deviation with respect to lower
bound, for example 2,resulted in larger value (in modulus).
Since computation of the bounds only requires evaluation of the geometry function (and its
derivative) for two crack sizes, it is much faster than direct numerical integration. For example 2,
computing the bounds took only 0.80 s, while numerical integrationtook 1642.67 s, leading to a
reduction factor of 1640 times. This result boosts the potential application of the bounds proposed
herein to problems involving complex geometries and solution by numerical analysis.
The computed upper bound also depends, to some extent, on the choice of a* points where
geometry function and its derivative are evaluated. Figure 9 shows the upper bound and relative
deviation of the upper bound, for example 2,with a = 2 a0 , 3a0 , 4 a0 . It is observed that the upper
231
bound functions are sensitive to different choicesof a , since the coefficient inEqs. (5) and (15)
232
depends on a . It is proposed that engineering insight should be used to select a . The largest value
233
of a , for instance, would be that leading to f ( a ) = K c , where Kc is the critical stress intensity
234
235
236
237
factor.
238
239
240
241
242
243
In this paperthe Lagrange rest, a classical result of calculus, was employed to propose a
theorem providing lower and upper bounds for function crack size, based on the solution of an initial
value problem involving the Paris-Erdogan law.The required assumption or hypothesis about the
geometry function, i.e., that it be a monotonously non-decreasing function of crack size, is generally
met in practice. Two polynomial functions are used to construct the lower and upper bounds. To
evaluate the bounds one only needs to evaluate the geometry function and its derivative for two crack
221
222
223
224
225
226
227
228
229
CONCLUDINGREMARKS
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page9of27.
244
245
246
247
sizes. Hence, the reduction in computational effort is huge, when compared with direct numerical
integration. This is especially true when the geometry function is not analytical, and stress intensity
factors are evaluated numerically. The accuracy of the proposed bounds was evaluated for two
example problems solved herein. In general, deviations were found to be around 10 to 20% for
248
a = 2a0 , which can be acceptable for engineering purposes. This accuracy can be improved by
249
250
251
252
proper choice of the support pointa*.
253
254
The first and third authors acknowledge theBrazilian National Research Council (CNPq) for
sponsoring this research.
255
256
257
258
259
ACKNOWLEDGEMENTS
REFERENCES
1.
Paris, P. C., F. Erdogan (1963).A critical analysis of crack propagation laws, J. Basic Eng., Trans.
ASME. 85, 528-534.
260
2.
Elber W (1970). Fatigue crack closure under cyclic tension. Eng. Fract. Mech. 2, 3745.
261
262
3.
Forman R. G., Kearney V. E., Engle R. M (1967). Numerical analysis of crack propagation in cyclic
loaded structures, J. Basic Engineering. 89, 459-464.
263
264
4.
Leonel E. D., Venturini W. S (2011).Multiple random crack propagation using a boundary element
formulation. Engineering Fracture Mechanics. 78, 1077-1090.
265
266
5.
Leonel E. D., Beck A. T., Venturini W. S (2011). On the performance of response surface and direct
coupling approaches in solution of random crack propagation problems, Structural Safety. 33, 261-274.
267
268
269
6.
Leonel E. D., Chateauneuf A., Venturini W. S., Bressolette P (2010). Coupled reliability and boundary
element model for probabilistic fatigue life assessment in mixed mode crack propagation. International
Journal of Fatigue. 32, 1823-1834.
270
271
272
7.
Duarte C. A., Hamzeh O. N., Liszka T. J., Tworzydlo W. W (2001). A generalized finite element
method for the simulation of three-dimensional dynamic crack propagation, Computer Methods in
Applied Mechanics and Engineering. 190, 1517, 2227-2262.
273
274
8.
Duarte C. A., Kim D. J., Quaresma D. M(2007). Arbitrarily smooth generalized finite element
approximations, Computer Methods in Applied Mechanics and Engineering. 196, 13, 33-56.
275
276
277
9.
Duarte C. A., Kim D. J (2008).Analysis and applications of a generalized finite element method with
globallocal enrichment functions, Computer Methods in Applied Mechanics and Engineering. 197, 6
8, 487-504.
278
279
10.
Rthor J., Roux S., Hild F (2010). Mixed-mode crack propagation using a hybrid analytical and
extended finite element method, ComptesRendusMcanique. 338, 121-126.
280
281
282
11.
Shen Y., Lew A (2010). Stability and convergence proofs for a discontinuous-Galerkin-based extended
finite element method for fracture mechanics, Computer Methods in Applied Mechanics and
Engineering. 199, 3740, 2360-2382.
283
284
285
12.
Elguedj T., Gravouil A., Combescure A (2006).Appropriate extended functions for X-FEM simulation
of plastic fracture mechanics, Computer Methods in Applied Mechanics and Engineering. 195, 78,
501-515.
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page10of27.
286
287
288
13.
Gupta P., Pereira J. P., Kim D. J., Duarte C. A., Eason T (2012).Analysis of three-dimensional fracture
mechanics problems: a non-intrusive approach using a generalized finite element method, Engineering
Fracture Mechanics. 90, 41-64.
289
290
14.
Virkler D. A., Hillberry B. M., GoelP. K (1979). The statistical nature of fatigue crack propagation,
Journal of Engineering Materials and Technology (ASME). 101, 148-153.
291
292
15.
Ghonem H., Dore S (1987). Experimental study of the constant-probability crack growth curves under
constant amplitude loading. Engineering Fracture Mechanics. 27, 1-25.
293
16.
Provan J (1987). Probabilistic Fracture Mechanics and Reliability, Dordrecht, MartinusNijhoff.
294
17.
Sobczyk K., Spencer B. F (1992). Random Fatigue: From Data to Theory, Academic Press, London.
295
296
18.
Beck A. T., Melchers R. E. (2004). Overload failure of structural components under random crack
propagation and loading - A Random Process Approach. Structural Safety 26, 471-488.
297
298
19.
Beck A. T., Gomes W. J. S (2013). Stochastic fracture mechanics using polynomial chaos. Probabilistic
Engineering Mechanics. 34, 26-39.
299
20.
Coddington E. A (1989). An Introduction to Ordinary Differential Equations, Dover.
300
301
21.
Ballantine J. A., Conner J. J., Handrock J. L (1989).Fundamentals of Metal Fatigue Analysis, Prentice
Hall, New Jersey.
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page11of27.
329
330
331
332
333
334
335
336
( )
da
log dN
II
III
da
= C K
dN
( )
337
338
339
log ( K)
da
Figure 1: Diagram log ( dN
) log ( K ) .
340
341
342
Figure 2: Relative error ( ( N k ) N k ) for RK4 solution of
343
344
345
346
example 1.
347
Table 1: Polynomial coefficients of Eq. (15) for the example problems.
Example
348
349
350
351
7.958310
1.317310
15
3.955610
14
6.236110
15
15
6.286310
7.873610
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page12of27.
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
a0
Figure 3: Finite width center-cracked
plate, subject to tensile loading.
368
369
370
371
Figure 4: Crack size function and lower and upper
bounds for example 1.
Figure 5: Relative deviationof lower and upper bounds,
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page13of27.
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
UB ,LB
( N ) N ) , for example 1.
a0
a0
2b
Figure 6: Finite width plate with two corner
cracks subject to tensile loading.
Figure 7: Crack size function and lower and upper
bounds for example 2.
Avila et al., Analytical Bounds for Efficient Crack Growth Computation. Submitted to AMM, March 2014. Page14of27.
Figure 8: Relative deviation of lower and upper bounds,
395
396
UB , LB
( N ) N ) , for example2.
397
398
Table 2: Maximum and minimum values of relative deviation ( max ( UB ) , min ( LB ) ) ,
399
for upper and lower bounds, for examples 1 and 2.
Example
N1
N1
min { LB ( N k )}k =1 , [% ]
10.4989
11.6396
-10.7216
-19.4529
1
2
max {UB ( N k )}k =1 , [% ]
400
401
402
403
404
405
406
Figure 9: a)Crack size function and upper bounds, ( a ( N ) , aUB ( N ) N ) ;
b) Relative deviation of upperbound, ( UB ( N ) N ) .
