Wohler Curve | Endurica LLC

Tension / Compression Cycles, R ratio and a Discussion of Wohler Curve for Rubber

May 24, 2022April 22, 2022 by William V. Mars, Ph.D., P.E.

Fatigue Life of Rubber with different strain constraints

In fatigue testing, R is the ratio of the minimum to the maximum occurring during one period of a cycle. If the mean value is zero (ie the cycle is centered on zero), then the minimum is equal and opposite in sign to the maximum. In this case, we say that R=-1. The cycle can be defined by various parameters. If we define R in terms of the stress or the strain, then R may take either positive or negative values. We might hear, for example, that R=-1 uniaxial loading is a symmetric tension / compression cycle.

Be careful about definitions, however! The Endurica material models define R in terms of the tearing energy (ie T = – dU/dA, where T is the tearing energy, U is the elastic potential energy and A is the crack area). The tearing energy is the driving force responsible for crack growth. It is always greater than or equal to zero. When R is defined in terms of tearing energy, its range is 0 ≤ R ≤ 1.

This leads to the following question that we often hear: how can Endurica compute fatigue in compression if it does not admit a negative R ratio?

Let’s look at a series of signals, all having the same strain amplitude, and each with a different mean strain. We will set the mean strain so that it is at most equal to the amplitude (corresponding to fully relaxing R=0 tension), and at least equal to the negative of the amplitude (corresponding to fully relaxing R=0 compression). In the middle of the range, we have fully reversed tension/compression (what some call “R=-1” loading when defining R in terms of strain). The principal engineering strains are plotted below for each case. This gives a smooth progression from a cycle that is only compression to a mixed cycle with both tension and compression and finally to a cycle that is only tension.

Series of signals with same strain amplitude but different Tension/Compression levels

Even for these simple uniaxial cases, the critical plane for simple compression is not the same as the critical plane for simple tension. In simple compression, due to crack closure, the critical plane is the plane that maximizes shearing. These planes make a 45-degree angle with the axis of loading (x direction). In simple tension, however, the critical plane is perpendicular to the load (x direction). In the figure below, the change in the critical plane as the mean strain increases from compression to tension is evident. On each sphere, the arrows indicate the perpendicular of a critical plane.

Mean strain increase as compression to tension is evident with change in the critical plane

The next figure shows the cracking energy density (units of mJ/mm³, and proportional to tearing energy) as a function of time on the critical plane for each case. The symbols on each line indicate the times at which the identified crack plane is open or closed. Now we can see clearly that fully reversed tension/compression (ie “R=-1” loading in terms of strain) is really R=0 when viewed in terms of tearing energy on the critical plane.

Graphs showing the cracking energy density as a function of time on the critical plane. Fully reversed tension/compression is really R=0 when viewed in terms of tearing energy.

The computed fatigue life is given in the last figure for each case using this material definition:

MAT=RUBBER
ELASTICITY_TYPE=ARRUDABOYCE
SHEAR_MODULUS=1 ! MPa
LIMIT_STRETCH=4
BULK_MODULUS=3000 ! MPa
FATIGUE_TYPE=THOMAS
FLAWSIZE=0.025 ! mm
FLAWCRIT=1 ! mm
TCRITICAL=10 ! kJ/m^2
RC=3.42E-2 ! mm/cyc
F0=2
X(R)=LINDLEY73

The computed fatigue life depending on different percentages of tension/compression

The moral of the story:

Fully reversed tension/compression cycles (“R=-1” in stress or strain terms) are really fully relaxing cycles (R=0 in tearing energy terms) from the perspective of the crack precursor on the critical plane.
The critical plane depends on whether you have tension or compression. Wohler curve analysis completely misses the fact that the failure plane is not always perpendicular to the loading direction!
A simple sinusoidal history that crosses through zero results in separate tension and compression events, each of which has its own peak and valley, and each of which influences the critical plane selection. Wohler curve analysis based on max principal stress or strain amplitude completely misses these physics.

Two Decades of Critical Plane Analysis

June 2, 2022April 1, 2020 by William V. Mars, Ph.D., P.E.

It has been 20 years since Critical Plane Analysis for rubber was first conceived and validated. There were early signs of its significance. It won awards wherever I presented it. At the 1999 SAE Fatigue Design and Evaluation meeting, it won the Henry Fuchs award. At the 2000 Tire Society meeting, it won the Superior Paper award. At the Fall 2000 ACS Rubber Division meeting, it won the Best Paper award. Upon completing my 2001 doctoral thesis, we applied for and received a US patent (2003) on it.

The strongest early sign was that I soon found myself in company with others pursuing similar thinking. The earliest was Dr. Nicolas Saintier. As far as I know, neither of us was aware of the other’s work until 2006. That was when he published an account similar enough to my own that when it came across my desk and I first started to read it, I felt certain he would cite my 2001 work as a source. I have to admit to initially feeling let down when I reached the end of his paper and found no mention of my work. I immediately looked for his other papers and found his 2001 doctoral thesis titled “Fatigue multiaxiale dans un élastomère de type NR chargé: mécanismes d’endommagement et critère local d’amorçage de fissure.” (Multiaxial fatigue life of a natural rubber: crack initiation mechanisms and local fatigue life criterion). There it was – the same founding principle of Critical Plane Analysis that I had worked so hard to articulate and validate – the idea that cracks develop on a material plane, specifically the most critical material plane, and that their localized experience drives their evolution. That we both articulated this beautifully simple and powerful principle in the same year with complete independence from each other, when no one else working on elastomers had yet spoken of this approach (there were precedents in the field of metal fatigue analysis), just shows that it was an idea whose time had come.

Although the foundational principle of Critical Plane Analysis was the same, there were also important differences between our accounts. We differed on 1) how the critical plane is selected, 2) what criterion is used to quantify the severity of loading experience by the critical plane, 3) how damage on the critical plane evolves under solicitation. The following table summarizes the key differences:

Table 1. Comparison of the Mars and Saintier Frameworks for Critical Plane Analysis.

	Mars 2001	Saintier 2001
Critical Plane Selection Method	Minimize the computed life after evaluation of damage on all planes	Maximize the principal stress prior to evaluation of damage
Multiaxial Criterion	Energy release rate estimated via cracking energy density on every plane	Stress traction on the assumed critical plane
Damage Evolution Law	Integration of crack growth rate law	Power law Wohler curve
Strain Crystallization Law	Treated as R ratio dependence of the crack growth rate law	Treated as a modifier of the stress experienced on the critical plane

It may be said that Saintier’s approach followed more closely the precedents for Critical Plane Analysis in metal fatigue, particularly with respect to the method used to select the critical plane. Selecting the plane is the first step in his method (identify the plane in order to compute the damage), but it is the last step in our method (compute the damage on each plane first and lastly pick the plane with the most damage). Saintier’s approach also depends on a Wohler curve style characterization of fatigue behavior, where ours is defined via a crack growth rate law. We have previously discussed the pros and cons of Wohler curves vs. fracture mechanics. In our approach, we placed a high priority on taking advantage of the very large pre-existing body of knowledge on the fracture mechanical behavior of elastomers, and on the economic and operational advantages that crack growth experiments enjoy.

Since my and Saintier’s first steps, there have now been many others who have contributed in various forms to the overall method, its validation and/or its application. It is safe to say that Critical Plane Analysis is here to stay, and set to continue expanding for many years (there are now several hundred research papers!).

For our part, Endurica is now in year 12 of delivering commercial grade fatigue analysis solutions built on this method. Today, Critical Plane Analysis is a production analysis workflow used by many engineering organizations to solve critical durability issues. It is the heart of the Endurica fatigue solver, and there are hundreds of trained users (look up the #fatigueninjas on twitter!). It is unrivaled for its reliability, speed and accuracy in computing the impacts of multiaxial loading on durability.

What do the next 20 years hold? We are going to see a transition in how fatigue analysis is used. OEM organizations that manage durability and risk across rubber component supply chains will transition away from receiving fatigue simulation results on an optional basis towards requiring fatigue simulations by default on every part at the inception of new programs. Expectations and achievement of cost-reduction, light weighting and sustainability initiatives will increase as product optimization begins to fully account for actual product use cases. Critical Plane Analysis has already laid the foundation for these things to happen. Older fatigue analysis methods that do not compete well against critical plane methods will become obsolete. On the research side, there will be further development of material models for use in the critical plane framework. Ageing, inelasticity, rate and anisotropy effects still need further development, for example. In 20 years, durability will be just one more thing that engineers do well every day, whether or not they know that Critical Plane Analysis was how they did it.

Mars, W. V, Multiaxial fatigue of rubber. Ph.D. Dissertation, University of Toledo, 2001.

Mars, W. V. “Multiaxial fatigue crack initiation in rubber.” Tire Science and Technology 29, no. 3: 171-185, 2001.

Mars, W. V. “Cracking energy density as a predictor of fatigue life under multiaxial conditions.” Rubber chemistry and technology 75, no. 1: 1-17, 2002.

Mars, W. V., “Method and article of manufacture for estimating material failure due to crack formation and growth.” U.S. Patent No. 6,634,236. 21 Oct. 2003.

Saintier, N, “Fatigue multiaxiale dans un élastomère de type NR chargé: mécanismes d’endommagement et critère local d’amorçage de fissure.” Ph. D Dissertation., Ecole des Mines de Paris, 2001.

Saintier, N, G. Cailletaud, R. Piques. “Crack initiation and propagation under multiaxial fatigue in a natural rubber.” International Journal of Fatigue 28, no. 1: 61-72, (2006).

Conservatism and Tradition in Fatigue Analysis

June 9, 2022January 19, 2020 by William V. Mars, Ph.D., P.E.

Slide Rule

Because Endurica’s Critical Plane Analysis is a relatively new approach to fatigue analysis of elastomers (introduced in 2001), new users often ask whether its predictions are conservative: i.e., does its predictions reliably lean in favor of safety? And is it more or less conservative than the traditional approaches it supplants?

Fatigue analysis for elastomers follows two distinct traditions. The earliest tradition traces to Sidney Cadwell’s work in 1940 which followed the even earlier ideas of metal fatigue pioneer August Wohler. This tradition is based on matching up empirical crack nucleation curves to corresponding in-service operating conditions via convenient parameters such as stress or strain. It is typically the first approach that engineers encounter in their undergraduate training, as it is often effective and relatively simple to apply. A later tradition, Fracture Mechanics, traces to the post-WWII work of Ronald Rivlin and Alan Thomas in 1953 which extended Griffith’s seminal 1921 work on rupture to elastomers. In this tradition, the energy requirements for growing a given crack provide the core organizing principle for analysis. Combined with empirical crack growth rate curves, this approach can make high accuracy life predictions for a very broad range of application scenarios. This approach is typically first encountered in graduate-level engineering programs, and due to somewhat more complicated mathematics, usually requires specialized calculation software to apply it.

There are a few big holes in the Wohler curve approach. For elastomers, perhaps the biggest limitation is that this approach assumes a priori that damage is associated with the maximum principal stress or strain. This is sometimes true for simple cases, but not always: 1) strain crystallization is known to produce off-axis cracking not aligned with the principal stress, 2) compression is known to produce cracks on planes of maximum shearing, and 3) out-of-phase multiaxial loading cases do not even possess a unique, well-defined principal direction – the directions vary in time. It is also well known that Wohler curves for rubber depend strongly on mode of deformation. Fatigue experiments in simple tension, biaxial tension, simple shear, and simple compression do not simply resolve to a single universal curve, as the Wohler approach takes for granted. To use this approach conservatively then requires that the most damaging mode of deformation – simple tension – be used as the baseline.

Perhaps the biggest limitation of the traditional Fracture Mechanics approach is that it typically focuses on only one crack at a time. In fatigue, structures begin with many microscopic cracks distributed randomly throughout. Most of the fatigue life of the structure is spent growing many small cracks. Only towards the very end of life do one or a few large cracks finally emerge as worst cases. True conservatism would mean tracking the growth of all of possible large cracks, and finding out which one(s) grow the fastest. But traditional fracture mechanics tools are not well adapted for this task. They require up front assumptions about the location and shape of the worst case crack. How can you find a worst case without considering many alternatives?

Critical Plane Analysis is simply the idea that a crack could occur anywhere in a structure, and it could occur in any orientation. It checks all of the possibilities, and it finds the worst ones. It looks at the specific loading experiences of each individual crack plane that might occur. It takes account of material behavior like strain crystallization. It takes account of crack closure conditions. It takes account of the fracture mechanical behavior of small cracks. It does not make unwarranted assumptions about the orientation of cracks. It correctly predicts the orientation of cracks for all modes of deformation. It is the most exhaustive and conservative fatigue analysis that you can do.

Don’t mistake traditional approaches with the conservative approach. Critical Plane Analysis is, by definition, the most conservative approach because it doesn’t make any assumptions about crack location or orientation, and because it checks all of the possible ways a crack might occur.

Wohler Curves or Fracture Mechanics?

May 4, 2022February 22, 2018 by William V. Mars, Ph.D., P.E.

Endurica uses a fracture mechanics based description of rubber’s fatigue behavior, rather than the classical Wohler curve (ie S-N curve) approach. This is why:

1) Wohler curves in rubber show the combined effects of several nonlinear processes, but they do not easily deconvolve into useful information about the individual processes. This means that Wohler curve users struggle to trace the causes of fatigue failures any deeper than the single monolithic empirical SN curve. When the customer or the boss asks why the part is failing, Wohler curve users end up falling back on the old “rubber is mysterious” defense. Meanwhile, users of critical plane analysis + fracture mechanics are hypothesis testing. They can check what events and what loading directions are most damaging, and what material parameters (crack precursor size, strain crystallization, threshold, crack growth rate law, thermal effects, etc.) can be exploited to gain leverage and solve the issue.

2) Fatigue failure in rubber is often dominated by “special effects”: dependence on strain level, dependence on R ratio, dependence on temperature, dependence on rate, dependence on ageing, etc. The Wohler curve crowd must choose between ignoring/oversimplifying these special effects, or running an experimental matrix that rapidly scales to an infeasibly huge size as more variables are added. While fracture mechanics users obtain a wealth of information from a single test specimen (one test can probe many different strain levels, temperatures, rates, etc), Wohler curve users obtain 1 data point per tested specimen. Look in the rubber technical literature and count the number of S-N-curves that are given, relative to the number of fatigue crack growth rate curves. Google/scholar returns less than 2000 results for “rubber Wohler curve”, and 78700 results for “rubber crack growth curve”. There is a reason that crack growth rate curves outnumber Wohler curves.

3) SN based methods are not conservative. Wohler curve users end up assuming that a crack will show up perpendicular to a max principal stress or strain direction. This assumption only works when you have the very simplest loading cases, no compression, and no strain crystallization. Users of fracture mechanics + critical plane analysis don’t worry about whether they have simple loading, finite straining, out-of-phase loading, compressive loading, changing principal directions, and/or strain crystallization. Critical plane analysis checks every possible way a crack might develop and is therefore assured to always find the worst case regardless of detailed mechanisms.

4) Wohler curves are messy. They depend strongly on crack precursor size, which naturally varies specimen-to-specimen, batch-to-batch, and between lab mix and factory processes. During SN curve testing, the size of the crack is neither measured nor controlled. This accounts for the extra scatter that is typical in these tests. In fracture mechanics testing, on the other hand, the crack is measured and controlled, leading to more repeatable and reliable results. Noisy data means that the Wohler curve crowd has trouble differentiating between material or design options. Users of fracture mechanics benefit from cleaner results that allow more accurate discrimination with less replication.

A Wohler curve does have one valuable use. The Wohler curve can be used to calibrate the crack precursor size for a fracture mechanics analysis. It only takes a few data points – not the entire curve, since the crack precursor size does not depend on strain level, or other “special effects” variables. Our recommended practice is to run a small number of nucleation style tests for this purpose only, then leverage fracture mechanics to characterize the special effects.

The bottom line is that, for purposes of general fatigue life prediction in rubber, the Wohler curve method loses technically and economically to the fracture mechanics + critical plane analysis based method that is used in modern fatigue solvers.