Most people probably remember the TV show CSI: Crime Scene Investigation. One of the more popular “police procedural” television franchises in the recent past, CSI: (and its similarly named spin-off shows) differentiated itself by giving the crime scene’s evidence the starring role. Much of the drama in the show centered on how it painstakingly built the investigation process – understanding the “how” and the “why” – instead of focusing on the “what”. Rather than relying on intuition, investigators pieced together the physical evidence throughout the program, leading to witnesses, suspects, motives, locations, and timelines. Often, high-tech instruments would be used in the crime lab to analyze things like blood traces or bullet casings. Reports would be written which would then be used in the arrest and conviction of the perps, or conversely, to exonerate the wrongly-accused.

While real life crime scene forensic analysis may not always be as glamorous/gory as it’s depicted in Hollywood, distinct parallels can be drawn between the procedures involved in dissecting a crime scene and those used in component failure analysis. In fact, the term Forensic Engineering is used to describe the investigation of physical failures. And Forensic Materials Engineering categorizes the methods used and processes followed to study why structures and materials fail and who (or what) could have been at fault. As they say, it’s usually not good enough to determine when something broke (i.e. “number of fatigue cycles to failure”); the most interesting information comes as answers to the questions of how it broke (failure mechanisms) and why it broke (external conditions). A thorough failure analysis can answer all of these questions and more.

When manufacturing companies receive broken components from their end users (known as “field returns” or “field failures”), those components will usually go through a level of laboratory process very similar to a crime scene investigation. All available evidence will be collected, documented and analyzed, using both high-tech and low-tech instruments (tabletop light microscopes for metallography, scanning electron microscopes [SEM] for fractography, chemical analysis, etc.). Traditionally, these tests can be time-consuming, and as with any CSI time is of the essence (as well as efficiency and accuracy). Finally, reports will be produced to justify the conclusions made from the analyses. If a company is well-integrated, these failure reports will act as feedback to the design and materials teams on the front end of product lifecycle management (PLM). This would hopefully point out flaws that may have been missed during the original development activities, and bring about continuous improvement in the product. The insights can better inform the qualification testing testing procedures, so that reliability can be more realistically assessed. However, this feedback process can also take a lot of time because the field failures first have to occur, be reported, be adequately analyzed, and then results have to make their way to the proper teams in the organization. Wouldn’t it be more effective for there to be a way to perform this analysis beforehand, and help make the product better before it even has a chance to fail?

VEXTEC’s VPS-MICRO® software functions as a computational forensic engineering tool. By inputting component design and material information, you can virtually perform field tests of your parts under many different circumstances and conditions, with varying material capabilities. When the simulations complete, the output answer is not simply the predicted number of cycles to failure (the typical result of the prevalent fatigue estimation software packages around), but it also provides the entire digital trail of evidence that led to that failure. Each of the images below represents fatigue crack damage emanating from an initial defect; physical evidence on the left, and a computational simulation output of the damage from VPS-MICRO on the right. The physical evidence can show you where something failed, and after enough in-depth analysis, can give an idea of fatigue crack growth rate (da/dN). The simulation results on the right are even more informative, providing details such as crack growth (a vs. N), grain individual orientation, localized stresses, and microcracking. These simulation outputs capture the variability inherent in fatigue testing, and give quantitative evaluation of the likelihood of such events.

You can investigate the initiation mechanisms of the damage (whether they be from non-metallic defects, rogue large grains in the microstructure, surface roughness, etc.), and follow the damage right through to failure. The locations, motives, timelines, and suspects of component failure are all studied in the virtual realm, at a fraction of the time spent in physical test labs, reducing risk to end use customers as well as risk to important company metrics like PLM and warranty budgets. Just as “persons of interest” can be vindicated and set-free by the physical evidence that is found and properly processed, the predictive models used in VPS-MICRO can validated by comparison to results from a limited amount of physical testing. Once this verification and validation (V&V) activity is complete, the software can be used for a multitude of product design trade studies (geometrical effects, material second sourcing, manufacturing process changes). The evidence is still the star of the show here, and VEXTEC has added another high-tech resource to the CSI engineering toolbox!

In our last blog on the topic of corrosion, we discussed how corrosion control can be a “headache”, and how controlling corrosion implies controlling the man-made contributors such as corrosion fatigue, which is a very dominant corrosion factor. Nowadays, integrated computational material engineering (ICME) and recent advances in computational capabilities can alleviate that “headache”. VPS-MICRO^® is a probabilistic ICME tool that deals with fatigue problems. It builds a Virtual Twin^® of the microstructures of any existent component, product or a fleet of products.

VPS-MICRO is a multiscale tool that can receive data from continuum simulations, such as nodal or element solutions from Finite Element Analysis (FEA), and extends the simulation to the microstructural level. One can now accurately consider the microstructural complexities that make geometrically-identical test specimens have significantly different fatigue lives, even in well-controlled laboratory tests. The fatigue life variation caused by the microstructure is compounded by other variations in actual products in the field. VPS-MICRO overlays complex microstructures (with probabilistically-generated material properties) onto the continuum solutions, and runs simulations by transitioning cracks from nucleation to crack growth to final failure. Monte Carlo formulae are used to generate the complex microstructure: grain sizes, defects, orientations, frictional strengths, etc. The theory of VPS-MICRO is based on cyclic crystal plasticity, microstructural small flaw fracture mechanics, and long crack linear elastic fracture mechanics (LEFM).  The main output result from VPS-MICRO is the probability of failure of the tested component. It also provides more detailed information about the failed components, such as the expected statistical distributions of lives (in cycles), micro-cracks and the features on the fracture surface (i.e. grain sizes and orientations, frictional strengths and micro-stresses). All of these outputs can be visualized by using VPS-MICRO’s post-processor.

In VPS-MICRO, corrosion is considered one of the microstructural complexities. Corrosion results in geometrical microfeatures on the corroded surface. These microfeatures induce high stress concentrations (k_t) that initiate cracks more easily than a smooth surface. As an example, corrosion features in aluminum occur as big pits (macro-pits) that encompass smaller pits (micro-pits). These pits result in high stress concentration gradients that die out over a very thin layer into the depth. Stress gradients caused by micro-pits are called micro-gradients, while stress gradients caused by macro-pits are called macro-gradients. Micro-gradients can have a higher stress concentration than macro-gradients, due to the extremely rough topology on the microstructural scale, but they also die out faster.

Although the high stress occurs over a very small surface layer, corroded surfaces can reduce fatigue life by several orders of magnitude. This k_t occurs over a small scale, where the variation in the material’s microstructure greatly influences nonhomogeneous properties. In VPS-MICRO, these stress concentrations are applied concurrently to the crack growth models (nucleation, short crack, and long crack) as the crack grows. Microstructural small flaw fracture mechanics (SFFM) must be used to properly account for the thin layer of roughened surface. Continuum analysis methods do not work at this scale. To learn more about the theory behind stress and crack propagation adopted in VPS-MICRO, check out this example.

From previous projects, VPS-MICRO has provided reliable predictions of the lives of corroded components. The predicted results were within acceptable bounds of the experimental results, which highlights the benefits of using VPS-MICRO to build Virtual Twins of potentially corroded components. Employing this technology reduces the number of empirical physical tests required to qualify a component. VPS-MICRO provides time- and cost-effictive assessments of materials and designs. This decreases lead time, increases return on investment (RoI), and most definitely cures the corrosion control headache!

We previously discussed how corrosion can be “the good guy” for some applications when degradation is beneficial, such as in self-deteriorating medical implants. Also, how it can be “the bad guy” for some other applications where degradation impairs the structural integrity, like in airplane structures and oil and gas pipelines and vessels. As the bad guy, its impairing effect needs to be mitigated; and as the good guy, corrosion needs to be intentionally designed. Thus, both need to be managed and controlled. Corrosion control eats up time and money in both cases. Controlling corrosion, as the bad guy, has prompted manufacturers to develop a corrosion management system (CMS) to mitigate the effect of corrosion, since its global cost exceeded US$ 2.5 trillion in 2013 (3.4% of the global GDP). However, this CMS is not only applicable to unintended corrosion. It also can be adopted when corrosion is part of the design (the good guy). However, CMS requires advanced labs, massive amounts of data gathering and experiments, and years of experience, which in return consumes too much time and money.

With regard to CMS, the National Association of Corrosion Engineers (NACE) proposed a hierarchical pyramid that can be implemented. The pyramid consists of many elements and procedures that span from upper management down to testing practices that are tailored to control corrosion. Corrosion alone is seldom the cause of degradation and failure. The structural failure is induced by manmade—controllable—factors such as dissimilar material galvanic corrosion, stress corrosion cracking and corrosion fatigue. Controlling corrosion means controlling those manmade factors. It is the combination of corrosion and loading that has been the dominant factor driving failure for many industries such as healthcare, oil and gas, weapon systems and infrastructure and aerospace. For example, the Department of Defense (DoD) estimated that its cost of unintended corrosion, that is mainly caused by corrosion fatigue, exceeded US$ 21 billion in 2010. Assessment of corrosion fatigue requires collecting a massive amount of information and data about the microscopic structure of the materials and their reliability. That requires years of experience and running a huge number of experimental tests. Consequently, huge amounts of time and money need to be spent in order to design, make decisions and implement action items. However, in the past few years, computer modelling in material science has advanced to the point that we can start to predict corrosion damage based on the underlying physics instead of relying on large test programs. Using integrated computational materials engineering (ICME), which combines industrial knowledge with physics, we can reduce the time and money needed for the decision-making process.

VEXTEC has developed the Virtual Twin^® that predicts the probability of degradation down to the microstructural level, where damage starts. The latest version of VEXTEC’s software VPS-MICRO^® provides a probabilistic physics-based assessment of materials exposed to corrosion fatigue. VPS-MICRO is a versatile software. The software combines our historical knowledge of corrosion and fatigue with physics-based computations. It is multifunctional and can interact with the different elements of a corrosion management system depending on the type of requested results. Engaging VPS-MICRO as a Virtual Twin can provide a time and cost-efficient evaluation of new or existing materials and designs. It not only reduces the corrosion control cost, but it increases the return on investment; less amount of experimental testing, faster evaluation, and less hassle during the decision-making process. The technology adopted in VPS-MICRO will be discussed in the concluding blog in this series.

A turbocharger client of ours wanted to improve durability and reduce warranty costs on cast wheels made from a nickel superalloy with a radially-solidified (RS) microstructure. A significant portion of their previous field failures had been attributed to high cycle fatigue (HCF). Our client already had ideas about how to reduce these HCF failures by changing the wheel’s microstructure to an equiaxed (EQ) morphology. General representations of RS and EQ microstructures are shown here.

Since design changes like these would require a hefty amount of physical validation testing, they needed a way to predictively quantify the costs/benefits to the product line, should some of these proposed changes be implemented. So they turned to VPS-MICRO®. The VPS-MICRO simulation platform combines structural finite element analysis of the component (FEA, seen below) with a 3-D spatial model of the material’s microstructure to predict component durability risk. It is a probabilistic framework, accounting for variability in microstructure and strength properties, applicable damage mechanisms, and usage over time.

The primary inputs to VPS-MICRO are the design input file (stresses from the FEA and the corresponding stressed area in terms of elemental surface area) and the material input file. VEXTEC has developed plug-ins to extract the design input information from several commercial FEA software programs. The material input file contains all the relevant material properties of the component, from macro-scale to the microstructural level. These properties include those you would normally find in an FEA analysis (material modulus and Poisson’s ratio), but also microstructural properties such as grain size, population density of variously-sized inclusions/defects, and grain-level strength and energy parameters. It is the inherent variability of these microstructural properties that is a key factor of component-level fatigue life variability. The good news is that these properties can be statistically evaluated using industry standard (ASTM) tests.

Now, back to our client’s specific issue. Their RS material microstructure was originally developed to resist the onset of damage at high temperatures. The likelihood of initiating damage is low due to fewer grain boundaries. However once damage initiates, the failure probability goes up because there aren’t as many grain boundaries to arrest crack growth. Their proposed design change, using an equiaxed (EQ) microstructure instead for the turbocharger wheels, was thought to be more damage tolerant. The likelihood of initiating damage would be higher, but so would the opportunity for fatigue crack arrest (more grain boundaries). Using VPS-MICRO, our client was able to pursue a quantitative assessment of the risk of HCF failure versus grain type (radially-solidified vs. equiaxed), before any re-designed wheels were even produced or tested.

Shown below is the VPS-MICRO simulated fatigue life comparison of the current-state RS wheel, and the proposed EQ wheel (baseline average grain size = 2.7 mils). The comparison results are presented using a simulated S-N (Stress-Life) plot. The figure shows considerable variability at each stress level for both materials. Run-outs (the points on the right marked with arrows) are predicted at every stress level. A “run-out” means the simulated specimen did not fail within the number of cycles analyzed. These results indicate the RS wheel would have a lower endurance (fatigue limit) compared to the baseline EQ wheel. Generally speaking, this would seem to indicate that the EQ material is better than the RS material. These results appeared to correlate with published industry reports.

Because the wheel was a casting, there is an expected grain size variation throughout the part. Our client’s quality group thought they could maintain the EQ grain size between 1.7 and 3.8 mils, but acknowledged that sizes as high as 15 mils could occur. So they used VPS-MICRO in a different way: to evaluate the sensitivity of grain size to the risk of wheel failure. Their virtual analysis revealed that EQ wheels are not always better than RS wheels.  The figure below shows that failure probability is low for small EQ grains, but is very sensitive to grain size.  At a grain size of 15 mils, the EQ wheel is actually more likely to fail than the RS wheel (which has an average grain size of 87 mils). Probability of failure is not as sensitive to grain size for the RS wheel. Did the reversing trend make sense?

VEXTEC and our turbocharger client investigated this relationship between the grain size and HCF failure risk. After analyzing the output of the VPS-MICRO simulations, we determined that competing failure mechanisms were present:

• The area effect: it takes more small-sized grains to fill a given surface area compared to fewer, larger grains. A smaller average grain size means a statistically-higher probability of having a weaker grain in a given area. This is analogous to the “weakest link theory“, where increasing the number of links in a chain increases its probability of failure. This explains why larger grains are producing fewer failures compared to small grains (the downward trend of the figure above).
• The grain-level strength effect: as the grain size increases, an initiating fatigue crack has a larger size as well. These larger-sized starter cracks are more likely to grow (with minimal arresting) to final failure. Therefore, the local strength properties of the grains become key gatekeepers to either prevent or allow these cracks to propagate from their initial sizes.

The final conclusions reached by our client, with the assistance of VPS-MICRO, were

• Using EQ material (2.7 mils) would reduce turbocharger wheel HCF failures by at least 60%
• Not all EQ materials are equal; small changes in grain size yield large changes in durability
• The probability of wheel failure was not as sensitive to grain size for the RS material
• Replacing RS material with EQ material requires significantly-tighter production control

Our client could now make a more-informed decision about the proposed design change (producing and testing the EQ wheel). They knew they would have to cast the wheel in a production environment to capture realistic variations, and to assess their capability to hold tighter tolerance on grain size than what was previously required on the RS wheel.

We’ve said it before, and we’ll say it again: