Works for detecting cracks - but not much analysis beyond that.
I suspect they will take samples from the areas of issue.
Confirming material composition using a combination of Energy Dispersive X-Ray Spectroscopy (EDS),
X-ray Photoelectron Spectrometer (XPS), and verification with X-Ray Florescence Spectrometer (XRF).
I suspect that certified material was used - however when conducting any analysis you will need to confirm the materials first.
You will get a number of relative weight charts and EDS micrographs of the various samples, binding energy graphs from XPS - this will also confirm that the materials are within spec - and if you have some without cracking if the materials are slightly different in ones with failure and without.
As well you need to examine the substrate and surface and determine what has occurred from heat treatment (e.g., quench and tempering, induction hardening, flame hardening, laser hardening, and electron beam hardening), thermochemical treatment (e.g., carburizing, carbonitriding, nitriding, nitrocarburising, and boriding) or implantation (e.g., of nitrogen ions). Treatments are commonly used to improve properties such as hardness.
Without writing a paper on metallurgical analysis (which I am wholly unqualified to do) it's hard to explain everything that needs to get done - and we haven't even started to get in to what is done to examine metal treatment or coatings...
Then once all of that is done - one needs to compare it to the model and how the model compares in Stress Analysis software to what is being observed -- the problem with ANSYS type software is garbage in --> worse garbage out - so if the stresses on parts wasn't correctly anticipated - the models won't show you a valid representation of what occurs - and you end up with catastrophic issues - or parts severely over engineered.