Gains:
- Ability to investigate the root cause of a damage with an AI-supported systematic framework (fracture surface, load, environment, microstructure)
- Ability to distinguish fatigue, brittle fracture, creep and corrosion mechanisms with AI and establish a chain of evidence
- Ability to verify AI's damage hypotheses with fractography, laboratory finding and engineering calculation
When a crane rope breaks, a turbine blade breaks, or a pipeline bursts, the question "why?" arises. The question arises. Failure analysis is the science of systematically investigating why and how a part malfunctions, finding the root cause and preventing recurrence. This job is one of the most responsible duties of the metallurgical engineer; because its outcome is sometimes the basis for a lawsuit, a recall, or a safety measure. AI is powerful in systematically listing possible damage mechanisms, organizing the chain of evidence, and producing draft reports. But fracture surface interpretation, fractography, and mechanism diagnosis require physical evidence; The definitive diagnosis given by AI from a single photograph is a hypothesis; the root cause cannot be declared without confirmation by laboratory findings.
Damage mechanisms: basis for differentiation
The first step to getting to the root cause is to discern the mechanism by which the damage occurs. AI is powerful at teaching the distinctive signatures of these mechanisms:
- Fatigue: Slow propagation of crack under repeated load. Beach lines (beach marks, with the naked eye) and striae (fatigue lines in SEM) are visible on the fractured surface. It is the cause of most machine damage.
- Brittle fracture: Sudden breakage without plastic deformation. Cleavage (separation from bright, crystal planes) and chevron (arrowhead) patterns on the fracture surface.
- Ductile fracture: Breaking after a significant deformation. Dimple morphology in SEM.
- Creep: Slow permanent deformation under constant load at high temperature; intergranular spaces and cracks.
- Corrosion/stress corrosion cracking (SCC): Environment + stress combined; branched cracks, corrosion products.
Hint: In a damage, there is usually not a single mechanism, but a chain: for example, corrosion creates a pit, the hole concentrates stress, from there a fatigue crack begins. Instead of having the AI ask “one reason,” have it ask “the chain of possible mechanisms and evidence for each link.”
Step by step: systematic root cause analysis
- Collect background: Part history, load, operating environment, temperature, service time, similar damages.
- Examine without damaging it: Photograph the broken surface without cleaning or touching it; corrosion product and protect the starting area.
- Visual + fractography: Starting point and mechanism traces with naked eye, stereo microscope, then SEM.
- Verify material: Composition (spectrometer), hardness, microstructure — did the material meet specification?
- Mechanism and root cause: Unite the evidence; Where was the beginning, what triggered it (design, material, manufacturing, service)?
- Verify and report: Test the hypothesis by calculation/experimentation; Write an evidence-based report. The responsibility lies with the authorized engineer.
AI is very helpful in steps 1, 5, and 6 (organize, hypothesize, outline); 2-4. The steps are physical laboratory work and AI cannot do them.
Mechanism discrimination table
mechanism
naked eye trace
SEM trace
Typical context
verification
fatigue
Beach lines, flat area
striatation
repetitive load
Load history + SEM
brittle fracture
Chevron, glossy surface
cleavage
Low temperature/impact
Charpy, microstructure
ductile fracture
submission, checkmate
dimple
excessive single load
tensile test
creep
Numerous cracks, deformations
intergranular space
high temperature
temperature history
SCC
branched crack
Intergranular/transgranular
Environment + stress
chemical analysis
three mini cases
Case 1 — Hasty diagnosis. A technician shows the AI a photo of a broken bolt; The AI says "definitely don't get tired." The technician writes the report accordingly. However, SEM examination shows cleavage instead of striae: the part has actually broken brittle at low temperature, and the real problem is that the transition temperature of the material is above the service temperature. Fatigue diagnosis would lead to incorrect corrective action (load reduction); The real solution is material replacement. Lesson: definitive diagnosis from photography is misleading without fractography.
Case 2 — Chain of evidence established correctly. There is a repeated fracture in a pump shaft. The team gives the AI the payload, environment and geometry and asks about the possible chain of mechanisms. AI gives the hypothesis that stress concentration at the sharp corner may facilitate fatigue onset. On SEM, the team sees that the fracture starts from that exact corner and contains striae; also measures the corner radius is smaller than the drawing. Root cause: fatigue due to insufficient corner radius. Solution: increase the radius. AI gave the correct hypothesis, SEM and measurement confirmed it.
Case 3 — Material non-conformity. A spring breaks prematurely. AI also lists “material may not meet specification” among possible causes. The team performs composition analysis; finds that the carbon content is below specification, so the target hardness is not achieved by heat treatment. The root cause is wrong material in the manufacturing/supply chain. AI reminded of a substance to be checked, gave laboratory evidence. Lesson: AI's systematic listing brings up a cause that may be overlooked; Experiment provides the proof.
Copiable prompt templates
MECHANISM CHAIN TEMPLATE"Role: You are damage analysis assistant. Part: [description]. Load: [...]. Environment/temperature: [...]. Service time: [...]. Observation: [fracture location, appearance]. Give me the POSSIBLE damage mechanism chain (not sole cause). For each link: supporting evidence + confirmed by what examination (SEM, composition, hardness). Definitive diagnosis; give investigation road map."
FRACTOGRAPHY GUIDE TEMPLATE "I will examine the fractured surface. Comparatively list WHAT I should look for in the SEM and with the naked eye to DISCRIMINATE the following mechanisms: fatigue, brittle fracture, ductile fracture, SCC. How do I find the starting point? Emphasize that a definitive diagnosis cannot be made from the photograph, physical examination is essential."
5 WHYS / ISHIKAWA TEMPLATE"Establish a systematic root cause analysis for the following damage: [summary]. Group possible causes with fishbone (Ishikawa) categories: material, design, manufacturing, assembly, service/maintenance, environment. Deepen with the '5 why' chain in each branch. Suggest verification testing for each root cause candidate. Give hypotheses to be tested with evidence, not definitive conclusions."
DAMAGE REPORT DRAFT TEMPLATE "Role: You are the report writing assistant. DRAFT a damage analysis report from the following findings: [findings, measurements, fractography]. Sections: summary, background, investigation methods, findings, discussion, root cause, corrective recommendation. Attribute each technical value to its source. Do not include any numbers of unknown origin. State that final responsibility rests with the authorized engineer."
Weak prompt / Strong prompt
WEAK PROMPT:"Why did this bolt break? (photo)"
STRONG PROMPT: "Role: You are damage analysis assistant. Part: M20 8.8 bolt, in a vibrating joint, ~2 years service, outdoor. In the broken neck area, a flat area + final rupture zone is visible. Give me the chain of possible mechanisms (fatigue, brittle fracture, overload) and tell me what to look for in the SEM to distinguish EACH. Tell me how to find the starting point. Do not give a definitive diagnosis from the photo; which laboratory State that your findings will make the decision."
The weak prompt invites only one definitive (and possibly incorrect) diagnosis. The powerful prompt gives the context, asks for the chain of mechanisms and distinctive SEM traces, questions the starting point, and leaves the definitive diagnosis to the laboratory.
Common mistakes
- Clean/tap the fracture surface and remove evidence from the initial area.
- Declaring the exact diagnosis given by AI from the photo as the root cause without fractography.
- Looking for a single mechanism; bypassing the mechanism chain (such as corrosion → fatigue).
- Failure to check material conformity to specification (composition, hardness).
- Trying to assign a mechanism without collecting load and service history.
- Putting unverified (AI generated) technical value in the report.
In summary
In damage analysis, AI is a powerful aid in systematically listing possible mechanisms, establishing an Ishikawa/5-why framework, organizing the chain of evidence, and producing a draft report. However, fracture surface interpretation, fractography, and mechanism diagnosis require physical evidence (SEM, composition, hardness, load history); The definitive diagnosis that AI gives from a single photo is a hypothesis. Protect the starting area, investigate the mechanism chain, test each hypothesis with laboratory evidence, and keep the responsible engineer in charge.
Application task
Choose a real or fictional damage scenario (e.g. a bolt breaking in a vibrating joint). Type the part history, load, and environment. With the “MECHANISM CHAIN” template, ask the AI for possible mechanisms and the verification method for each. Then extract what to look for in SEM to distinguish mechanisms with the "FRACTOGRAPHY GUIDE" template. Finally, write in a paragraph which laboratory finding will confirm the root cause and who will approve the report.
checklist
- [ ] I collected parts history, load, environment and service time.
- [ ] I preserved the broken surface without damaging it and photographed it.
- [ ] I hypothesized the chain of possible mechanisms (not the sole cause).
- [ ] I identified SEM/lab traces that distinguish each mechanism.
- [ ] I checked that the material complies with the specification (composition, hardness).
- [ ] I confirmed the root cause with laboratory findings; I subjected the report to the approval of the authorized engineer.