Failure Analysis Of Engineering Materials

Why Parts Fail And What You Actually Need To Do About It

Most people think failure analysis is about running an SEM and calling it a day. It's not. It's about knowing which surface to look at, what scale to start at, and when to stop before you've destroyed every possible clue on the sample. I've spent enough time at fracture surfaces to know that the first cut you make determines whether you're doing science or just cleaning up someone else's mess. At its core, failure analysis of engineering materials is the process of reconstructing how and why a component stopped working as intended. The "why" is the hard part. The "how" usually presents itself pretty obviously if you haven't already ground it into dust with improper prep. People jump straight into fractography because textbooks teach it that way. Start with macro inspection instead. Photograph the part in place if you can. Document the loading history, the service environment, any repairs or modifications, and the timeline from first noticed issue to final rupture. That context alone will eliminate half the possibilities before you even think about metallurgy. I once pulled apart a bracket that kept cracking at a weld toe, and the fracture surface told us nothing useful until I found the weld repair log showing three separate heat passes without interpass temperature control. The metallurgy was secondary. The process was the primary cause.

The Workflow That Actually Works

Step one is always non-destructive examination if the part is still intact. Visual inspection, dye penetrant, magnetic particle, ultrasound where applicable. You need to know where the crack initiated before you touch anything destructive. The initiation site tells you the failure mode. Everything after that is confirmation and quantification. Sampling strategy comes next and it's where most people fail. You need three distinct samples minimum: the fracture surface, a cross-section through the failure zone, and unaffected base material for comparison. Take the fracture surface first. Handle it with gloves. If it's an environmentally assisted fracture, even breathing on it can deposit moisture and alter the surface. I've seen people wipe a fatigue surface with a lap cloth and then wonder why their SEM images showed nothing but smeared debris. Don't do that. Place it in a sealed container with a desiccant pack if you're not imaging within twenty-four hours. Metallography follows, and the sectioning technique matters more than most realize. Wire EDM is the standard for hard or tempered materials. Cutting with an abrasive disk generates heat that can alter the microstructure right where you need to see it. I've seen heat-affected zones from grinding completely mask stress corrosion cracking in 4140 steel. The cracks were there. The prep just cooked the area and blunted them beyond recognition. Wire EDM cuts at roughly three thousandths of an inch per minute on most alloys. It's slower. It's also the only way to guarantee you're looking at the real structure.

Chemical analysis is straightforward unless you're dealing with coatings or surface treatments. OES for bulk composition. XRF for quick screening. But here's the counter-intuitive part: composition is almost never the root cause in service failures. It's a contributing factor, sure, but the vast majority of failures come from processing errors, design flaws, or environmental mismatch, not from a material that's off-spec by a fraction of a percent. That said, you should always verify composition because it rules out mix-ups. I once dealt with a tube that cracked under hydrostatic testing and the lab initially blamed hydrogen embrittlement until I ran a quick OES check and found the supplier had delivered 1018 instead of 4130. Wrong material, wrong everything.

Fractography: Reading The Surface

Scanning electron microscopy is the workhorse for fracture surface analysis. Fatigue shows beach marks at low magnification and striations at higher mag. Ductile overload produces dimples. Brittle fracture gives you cleavage steps or intergranular separation. Stress corrosion cracking typically presents as a mix of brittle features with corrosion products filling the cracks. The trick is recognizing that these modes frequently coexist. A single component can show fatigue initiation, progressive growth, and final overload rupture all in one fracture surface. Mapping the zones takes practice and it requires keeping your stage movements deliberate and documented. One thing beginners consistently miss: magnification selection. Start at fifty to two hundred X to identify the overall morphology, then move up. Going straight to five thousand X is like trying to read a book by staring at individual letters without knowing which page you're on. You'll waste an hour finding features that are irrelevant to the failure mechanism. Energy dispersive spectroscopy on the fracture surface is where people get creative in the wrong way. EDX spot analysis on corrosion products or inclusion sites can confirm whether chlorides, sulfates, or other aggressive species are present. But mapping entire fracture surfaces is usually overkill unless you're dealing with a coating delamination or a contamination question. The time investment doesn't pay off for routine fracture mode identification.

Common Pitfalls In Failure Analysis Of Engineering Materials

Contamination during sampling is the number one problem. Using the wrong cutting fluid, touching fracture surfaces with bare hands, storing samples in non-sealed containers near chemical vapors. I had a case where a titanium alloy implant fracture was being analyzed and the lab used a standard mineral oil as a coolant during sectioning. The oil contaminated the surface and the SEM team spent two days trying to figure out what the carbon-rich deposit was before someone remembered the cutting fluid. Five minutes of recall instead of two days of dead investigation. Another pitfall is assuming the visible fracture tells the whole story. Subsurface defects, hydrogen ingress, temper embrittlement - these won't show up on the fracture surface alone. You need microstructural examination of the cross-section. I found a case of delayed hydrogen cracking in a high-strength bolt where the fracture surface looked like clean brittle rupture, but the cross-section revealed a network of microcracks branching perpendicular to the main fracture plane, oriented along grain boundaries that had been degraded by temper embrittlement during the original heat treatment. The bolt had been sitting in a parts bin for six months before it failed. Time-dependent. Nothing about the surface told you that without the cross-section. There's also the temptation to over-analyze. Running every test in the book on every sample because you want to be thorough. It's better to run fewer tests with clear hypotheses than to generate data you can't interpret. A proper failure analysis plan should list what you expect to find and what tests will confirm or refute each hypothesis. Stick to that plan unless the results force you to pivot.

Tools And Techniques Beyond The Obvious

Hardness testing across a weld or heat-treated zone gives you a quick map of microstructural variation. A Vickers indent pattern across a HAZ will show you exactly where the softening or hardening occurred. This usually takes ten minutes and rules out or confirms heat input problems faster than any microstructural analysis will. Residual stress measurement with XRD or hole-draining is useful but often misapplied. Most people request residual stress analysis on a fractured part, which is meaningless because the fracture has already relieved stresses in the immediate vicinity. You need the residual stress data from an identical, unfailed component if you're looking for a correlation. I've seen too many reports where someone measured residual stress on the broken part and then tried to use that data to explain why it broke. The data was post-fracture. It couldn't have caused the fracture. Finite element modeling has become standard in serious failure analysis work, and for good reason. When you've identified the geometry, loading, and material properties, a quick FEA run can confirm whether a stress concentration you suspected is actually severe enough to cause failure. The numbers don't lie, but they don't tell the whole truth either. Boundary conditions are always approximations. I usually run three models: worst case load, nominal load, and a sensitivity check on material properties. If all three predict failure at the same location, you've got confidence. If they disagree, you need better input data, not a fourth simulation.

When You Shouldn't Do This Yourself

Environmental scanning electron microscopy requires equipment most shops don't have. If your facility has an SEM but not an ESEM, you're limited to conducting samples under high vacuum. Non-conducting materials will charge and distort the image. You can coat them with carbon or gold, but coating alters surface chemistry and makes EDX analysis of those surface features unreliable. If you're dealing with polymers, ceramics, or corroded metals and you don't have access to an ESEM or a good environmental chamber attachment, send the sample out. The cost of outsourcing a few days of instrument time is lower than the cost of a wrong conclusion. Similarly, transmission electron microscopy for nanoscale fracture feature analysis is rarely necessary outside of research environments. Standard SEM resolution handles the vast majority of industrial failure cases. Don't let a fancy microscope choice drive your analysis plan. Pick the tool that answers your specific question.

Documentation That Actually Matters

Your final report needs to be readable by someone who wasn't in the room. Include photos with scale bars, labeled fracture maps, micrographs with clear magnification and etchant identification, and a traceable chain of custody for all samples. The most useful sections are the hypothesis tree and the elimination log. Show what you considered, what evidence ruled each alternative in or out, and what conclusion remains. Every step should be reproducible by another analyst following the same protocol. I keep a personal reference library of fracture surfaces from known failure modes because pattern recognition is faster than first-principles analysis every time. After looking at enough intergranular stress corrosion cracks in sensitized 304 stainless, you develop a visual shorthand that lets you classify a new sample in minutes rather than hours. The shorthand gets you to the right question fast. The formal analysis still takes the full time it takes.