# How to Assess Crack Failure

By: Greg Thorwald, Quest Integrity
As seen in MachineDesign Magazine. Download the PDF version.

Crack propagation might spell doom, but FEA software and FAD
analysis can tell users how much longer damaged equipment will operate
safely.

INTRODUCTION

Cracks found in machine and structural components raise two
questions: Will the crack cause an immediate failure, and if not, how
long until the crack grows large enough for failure?

To address the first question, two methods are used to evaluate cracked components:

- Ductile tearing analysis
- Failure Assessment Diagram (FAD)

Computing critical flaw sizes helps to ensure damage-tolerant
components and set appropriate inspection schedules when combined with
fatigue analysis. The historic Shippingport reactor vessel (see Figure 1) provides an example to compare the two analysis methods.

Figure 1. Shippingport Reactor Pressure Vessel

A crack in a pressure vessel is analyzed with quarter symmetry to
simplify the model. The top head and flange are omitted, since the
circumferential flange is much stiffer than the region around the
nozzle. A stress concentration by the base of the nozzle is a good
location to examine a postulated crack.

CRACK MESHES

3D crack meshes are generated for a range of crack sizes (FEACrack 2015). *Figure 2* shows cutaway views of the surface crack face for three crack sizes. The surface crack sizes are denoted by the crack length (2*c*) and the crack depth (*a*).

Figure 2. Cutaway View to Reveal the Surface Crack

A ductile tearing analysis requires a range of crack sizes to compute
the tearing modulus, so the crack length-to-depth aspect ratio is kept
constant at 2*c*/*a* = 3 to create a range of crack sizes.
Other crack aspect ratios can be used to develop a full trend of
critical crack sizes for shorter to longer crack shapes.

The concentric and radial mesh lines from the crack front are needed
to compute the J-integral. For elastic-plastic analysis, there is a set
of initially coincident nodes for the collapsed brick elements at each
crack front position. The coincident nodes can separate as the load
increases, helping to capture crack front blunting.

For the elastic-plastic analysis, the A508 steel material yield
strength is 36 ksi, and the tensile strength is 70 ksi (ASME 2010). A
Ramberg-Osgood equation is fit to the two tensile values to obtain a
stress-strain curve. The Young’s modulus (*E*) is 30,000 ksi, and the Poisson’s ratio is 0.29. The shell’s outside diameter is 125.75 in., and the thickness is 8.375 in.

FINITE ELEMENT ANALYSIS (FEA) RESULTS

The elastic-plastic analysis for each crack size uses 20 analysis
steps for equilibrium convergence and to provide output sets at each
pressure load. The principal stress results were presented for the
largest and smallest crack sizes.

The crack front J-integral values are shown in *Figure 3*. The
plot x-axis is the crack front node position angle, starting at the
bottom crack tip; the center of the plot is at the deepest point of the
surface crack. As the crack size increases, so do the J values, which is
typical for pressure loading causing tensile stress through the vessel
thickness.

Figure 3. Crack Front J-Integral Results for Each Crack Size

DUCTILE TEARING INSTABILITY ASSESSMENT

The ductile tearing instability analysis method (Anderson 2005, API
2007, Rowson 2011, and Thorwald 2011) requires a material resistance J-R
curve and J-integral results for several crack sizes. The A508 steel
J-R curve from API 579 Table F.10 is shown in *Figure 4*. The J-R curve is described by the power-law equation:

where Da is the crack extension measured from starting crack size *a*_{0} to the current crack size *a*; *C*_{T} is the tearing coefficient; and *n*_{T} is the tearing exponent. For A508 steel at 550^{o}F, the coefficient and exponent values are: *C*_{T} = 3.443 ksi*in, *n*_{T} = 0.329.

The 0.2 mm offset line shown in Fig. 4 intersects with the J-R curve at the *J*_{Ic} single toughness value, where crack front blunting gives way to ductile
tearing behavior. The slope of the 0.2-mm offset line is given by a
constant (default of 2 used here) multiplied by the flow stress. The
intersection of the J-R curve and offset line is found by iteration.

Figure 4. A508 Steel J-R Resistance Curve and Single Toughness Value

The curves in *Figure 5* present a comparison of the A508 J-R
curve to other materials. Most are generic carbon and stainless steels
from API 579, and the A53 curve is from material testing (Rowson 2011)
according to the ASTM E1820 standard (ASTM 2015).

Figure 5. J-R Curve Comparison

The ductile tearing instability assessment uses the crack front
J-integral values at a given load and compares them to the material
resistance J-R curve. *Figure 6* shows the crack front maximum J
values, which occur near the crack tip, for three pressure load cases
versus a range of crack sizes, as represented by the crack depth *a* (plot x-axis), since the 2*c*/*a* aspect ratio is kept constant. The label J_{app} is the "applied” J value from the analysis results.

The J_{app} result curves in Fig. 6 are obtained by plotting J
at the same load value from each crack analysis to get the trend of J
versus crack size for comparison to the J-R resistance curve.

Figure 6. Tangent Instability Point Used to Determine Critical Crack Size

A starting point for a ductile tearing assessment would use at least
three crack sizes—small, medium, and large—to obtain the J versus *a* trend and compute a tearing modulus. Adding more crack size cases is
beneficial because it adds more data points to the curves, ultimately
improving the trends and instability assessment.

Imagine the J-R curve in Fig. 6 being shifted left and right along the x-axis by varying *a*_{0} until a tangent point is found with a J_{app} curve. For the maximum pressure load case, a tangent point is found at J
= 3.7 ksi*in, corresponding to a starting critical crack size of *a*_{0} = 5.2 by 2*c*_{0} = 15.7 in.

If the starting crack size is slightly larger, the crack would be
unstable, predicting a failure at this pressure. In this case, the J_{app} and J-R curves would not intersect. If the starting crack size is
slightly smaller, the crack would have stable tearing, and the J_{app} and J-R curves would intersect. Note that the two lower-pressure load
cases intersect the J-R curve and would be considered stable for this
starting crack size.

There can be a region of overlap in the J versus *a* curves
near the tangent instability point, making visual identification of the
tangent point somewhat imprecise. To more precisely locate the tangent
instability point, the non-dimensional tearing modulus (*T*) is computed using:

where the tearing modulus is normalized by the Young’s modulus *E* and the yield strength; and *dJ/da* is the derivative of the J_{app} or J-R curve. T_{app} is the tearing modulus computed from J_{app}, and T-R is the tearing modulus computed from the J-R curve.

For the J_{app} curve, a polynomial curve-fit works well to fit the J versus a results and obtain the dJ/da derivative. The T_{app} tearing modulus is computed for the three pressure load cases at each crack size a and plotted versus the corresponding J values for those crack sizes in Figure 7.

Figure 7. Tearing Modulus Plot to Determine the Instability Point

In Fig. 7, points below the T-R curve are stable, and points above
the T-R curve are unstable and predict failure. The intersection of the
maximum pressure load T_{app} and T-R curves is where the slope
of those curves is equal (tangent point), and the intersection gives a
more definite instability point than on the J versus *a* plot. For the maximum pressure load case, the critical J = 3.70 ksi*in, which is found by iteration.

The critical J value from the tearing modulus curve intersection
gives the y-axis value of the tangent instability point in Fig. 6.
Solving the J_{app} curve-fit by iteration gives the crack size *a* at the tangent point on the J_{app} curve. Finally, Equation 1 is solved iteratively for the starting crack size *a*_{0} using crack size *a* and the critical J value at the tangent point.

COMPARISON TO THE FAILURE ASSESSMENT DIAGRAM METHOD

The FAD method (API 2007, Anderson 2005, Tipple 2012, and Thorwald
2016) evaluates a cracked structural component using two non-dimensional
ratios—Kr and Lr—to give the assessment point location compared to the
FAD curve.

The Kr ratio is the crack front stress intensity (K_{I}) computed using an elastic analysis, divided by the material fracture toughness (K_{Ic}).
The Lr ratio is the crack front reference stress computed using the
J-integral from an elastic-plastic analysis divided by the yield
strength. The reference stress calculation relies on the ratio of the
total J divided by an extrapolated elastic J to compute both the Lr
ratio and the analysis-specific FAD curve. If an Lr, Kr assessment point
is below the FAD curve, it is stable; a point above the curve is
unstable and predicts failure.

To compare the FAD method to the ductile tearing instability method,
the assessment point is computed for the critical crack size obtained
from the tearing instability (*a* = 5.2 by 2*c* = 15.7 in.). The K_{Ic} toughness used to compute the Kr ratio is taken from the J-R curve in Fig. 4 at the 0.2-mm offset line.

Figure 8 shows that the assessment point (black circle) is
outside the FAD curve, indicating that the FAD method is more
conservative than the tearing instability method. This is expected since
the J-R curve has a rising toughness trend to account for stable
ductile tearing before failure, while the FAD method uses the single
toughness value. The assessment point is near the right side of the FAD,
and is to the right of the A508 recommended Lr_{max} = 1.25 cutoff, indicating that plastic collapse is a possibility. If the default Lr_{max} cutoff is used, the assessment point is still to the left of that plastic collapse limit.

Figure 8. FAD Assessment Points Comparison

Another way to compare the assessment methods is to compute the
critical flaw size using the FAD method and compare that directly to the
critical crack size obtained using tearing instability. By varying the
crack sizes, the assessment points for critical crack sizes will be
located on the FAD curves. One choice is to determine the critical flaw
size on the analysis-specific FAD, and another choice is to determine
the critical flaw size on the default API 579 FAD. Both critical crack
sizes from the FAD are smaller and more conservative than from the
ductile tearing assessment.

The *table* below compares information needed by both methods,
such as elastic-plastic FEA and a range of crack sizes. Both methods
have advantages, depending on the material behavior, desired failure
assessment, and material data available. The ductile tearing assessment
requires the material J-R resistance curve, which may not always be
available. A single toughness value or lower-bound toughness may be
easier to obtain. The single toughness value used in the FAD method can
come from the J-R curve.

The FAD method needs high-enough crack front plasticity to complete
the reference stress calculations, which can cause FEA convergence
difficulties if the estimated maximum loading is too high. Often, the
needed maximum load is above the assessment load (a design or operating
load) and may take some trial-and-error for adequate J plasticity and
convergence. The ductile tearing method only needs convergence up to the
assessment load of interest, and does not have a crack front plasticity
requirement.

For smaller cracks or lower loads, the ductile tearing method may be
able to indicate quickly that the crack is stable from just a few
analysis cases. A separate plastic collapse check is needed for the
ductile tearing assessment, especially when the remaining ligament is
small, whereas the FAD method has a built-in plastic collapse check at
the Lr_{max} cutoff.

The FAD and ductile tearing methods from API 579 are single-parameter
assessment methods using K or J to characterize the crack front. In
some crack cases, a two-parameter approach may be needed to characterize
the crack front, in order to also compute, say, the T-stress. ASTM
2899 *(see reference 13)* uses a function of J and T-stress to
determine the instability location along the crack front, which could be
used to extend this analysis.

SUMMARY

The ductile tearing instability assessment and Failure Assessment
Diagram methods were compared by computing a critical flaw size of a
postulated crack in the Shippingport reactor pressure vessel.
Elastic-plastic FEA using 3D crack meshes computed the crack front
J-integral values needed for both methods. A range of crack sizes was
used in both methods to determine the critical flaw size.

The tearing instability assessment uses the J-R material resistance
curve, and the FAD method uses a single toughness value. In this
example, the tearing instability gave a larger critical crack size
compared to the FAD method. Both methods have advantages depending on
the material behavior, the desired failure assessment method, and the
material data available.

REFERENCES

- Abaqus/Standard version 6.14-1, Dassault Systémes Simulia Corp., Providence, RI.
- Abaqus Technology Brief, TB-060ARCAE-1, "Coupled Thermal-Structural Analysis of the Shippingport Nuclear Reactor Using Adaptive Remshing in Abaqus/CAE,” Revised April 2007, accessed Oct. 2015.
- Anderson, T. L., Fracture Mechanics, Fundamentals and
Applications, 3rd ed., 2005, CRC Press, Taylor & Francis Group,
Section 7.4, Section 9.3.3, and Equations 9.51 and 9.52.
- API 579-1/ASME FFS-1 2007 Fitness-for-Service, Table F.10, page
F-67 and Part 9, June 5, 2007, American Petroleum Institute and American
Society of Mechanical Engineers, API Publishing Services, 1220 L
Street, N.W., Washington, D.C. 20005.
- ASME Boiler and Pressure Vessel Code, Section II, Part D 2010
Table U, p. 464 and Table Y-1, p. 544, American Society of Mechanical
Engineers, Three Park Ave, New York, NY 10016.
- ASTM E1820-15, "Standard Test Method for Measurement of Fracture
Toughness,” ASTM International, 2015, 100 Barr Harbor Drive, PO Box
C700, West Conshohocken, PA 19428.
- FEACrack, Version 3.2.27, 2015, Quest Integrity USA, 1965 57th Court North Suite 100, Boulder, CO 80301, www.questintegrity.com.
- Rowson, Q., and M. Rock, "Comparison Between Ductile Tearing
Analysis and Linear Elastic Fracture Mechanics Analysis,” NAFEMS World
Congress, May 2011, Boston, MA.
- Thorwald, G., and M. Rock, "Ductile Tearing Analysis of a Custom
Pipe to Flange Nozzle Using 3D Crack Meshes,” NAFEMS World Congress, May
2011, Boston, MA.
- Thorwald, G., and L. Parietti, "Leak-Before-Break Assessment of a
Cracked Reactor Vessel Using 3D Crack Meshes,” Proceedings of the ASME
2016 Pressure Vessels and Piping Conference, PVP2016, July 17-21, 2016,
paper no. PVP2016-63910.
- Tipple, C., and G. Thorwald, "Using the Failure Assessment
Diagram Method with Fatigue Crack Growth to Determine
Leak-before-Rupture,” SIMULIA Community Conference, May 2012,
Providence, RI.
- Wikipedia, "Shippingport Atomic Power Station,” accessed Dec. 17, 2015.
- ASTM E2899-13, "Standard Test Method for Measurement of
Initiation Toughness in Surface Cracks Under Tension and Bending,” ASTM
International, 2013, 100 Barr Harbor Drive, P.O. Box C700, West
Conshohocken, PA 19428.
- Thorwald, G. and J. Wright, "Ductile Tearing Instability
Assessment of a Cracked Reactor Pressure Vessel Nozzle for Larger
Critical Crack Size Compared to the FAD Method”, SIMULIA "Science in the
Age of Experience” conference, Boston 2016.