Engineering Failure Analysis: What It Is and Why It Matters
Engineering Failure Analysis: What It Is and Why It Matters
Joshua R. Lehman
Author
Failure Analysis12 min read
Most engineering teams respond to failures reactively: something breaks, they replace it, they move on. If the same thing breaks again, they replace it again. The cycle continues until someone asks why — and the answer to why is where engineering failure analysis begins.
Failure analysis is not forensics performed after a catastrophic event. It is a systematic discipline applied to every failure that provides information about whether the design, material, manufacturing process, or operating environment is performing as intended. Done well, it converts the cost of a failure into an investment in prevention. Skipped entirely, it converts a recurring failure into a recurring expense.
This series covers the methods engineers use to identify failure modes, interpret physical evidence, predict failures before they occur, and build products that fail less often. This first post sets the foundation: what failure analysis is, what it costs to skip it, and how the process works.
What Counts as a Failure
A failure is not only a fracture or a complete stoppage. Failure is any
condition where a component or system no longer performs its intended function
within its specified operating conditions. A bearing that runs noisily at
rated load has failed. A seal that leaks below its rated pressure has failed.
A shaft that deflects beyond its allowable tolerance under design load has
failed. Recognising subtle failures early is often more valuable than
investigating spectacular ones.
Engineering failures fall along a spectrum that most teams only recognise at the visible end. The most obvious failures are fractures, seizures, and complete functional stoppages — the part breaks, the machine stops, the product does not work. These are easy to recognise and impossible to ignore.
Below that threshold is a much larger category of performance failures: components that function but not well enough. A gearbox that runs within specification at room temperature but overheats under full load at 40°C. A weld that passes visual inspection but fails under cyclic load at 60% of its rated capacity. A connector that mates reliably in the factory but corrodes in the field within two years. These failures are often harder to recognise because the part is technically "working" — just not working well enough.
Below that is the category that causes the most long-term damage: incipient failures, conditions that have not yet caused a functional problem but will. Surface pitting that has not yet propagated. Fretting corrosion that has not yet reached the fatigue threshold. Clearance that has not yet allowed contact but is trending toward it. Incipient failures rarely receive attention because nothing has gone wrong yet — but they are the most economical to address, because the fix at this stage is a drawing change rather than a warranty claim.
Failure analysis addresses all three categories, but the leverage is highest on incipient failures. The discipline of looking carefully at returned parts, failed prototypes, and inspection findings — not just at production stoppages — is what separates teams that improve from teams that manage.
When a failure occurs and no analysis is performed, three things happen. First, the immediate cost is paid: scrap, rework, warranty, customer impact. Second, the root cause remains unknown, which means the failure will recur. Third, the opportunity to improve the design, process, or specification is lost.
The cost of a single field failure is the most visible part of the equation. Warranty costs are documented. Customer complaints are counted. But the recurring cost — the same failure repeating across multiple units, across multiple years, because the underlying cause was never addressed — is often an order of magnitude larger than the first event.
A conservative model for recurring failure cost: if a failure mode occurs on 2% of units annually at a warranty cost of $400 per event, and the product ships 1,000 units per year, the annual warranty cost from that single failure mode is $8,000. Over five years: $40,000. A thorough failure analysis that identifies and corrects the root cause might cost $2,000 to $5,000 in engineering time. The return on that analysis is immediate.
The less visible cost is the reputational and contractual consequence of repeat failures. Customers who experience the same problem twice are not neutral — they escalate to quality holds, corrective action requests, and in regulated industries, supplier qualification reviews. These consequences rarely appear in the cost of the failure itself but are often the most serious long-term result of skipping analysis.
A structured failure analysis follows a consistent sequence regardless of the failure type.
Define the failure. Before examining anything, write a clear statement of what failed, under what conditions, after how much service time, and what the consequences were. Vague failure definitions lead to vague conclusions. "The bearing failed after six months of operation under load cycling between 400 N and 1,200 N at 1,400 rpm in an ambient temperature of 35°C" is a useful definition. "The bearing broke" is not.
Preserve and document the evidence. The failed part is the primary evidence. Handle it carefully. Photograph it before any cleaning or sectioning. Note the fracture surface orientation, the location of cracks, the distribution of wear, the presence of corrosion products or debris. Evidence destroyed in the handling process cannot be recovered.
Preserve the Evidence
The most common mistake in failure investigation is cleaning the failed part
before examination. Corrosion products, wear debris, and surface contamination
are evidence — they tell you what the operating environment looked like and
often identify the failure mechanism directly. Do not clean, paint, or machine
a failed part before documenting its condition. Once the evidence is gone, it
is gone.
Identify the failure mode. Failure mode is the physical mechanism by which the failure occurred: fatigue fracture, ductile overload, abrasive wear, galvanic corrosion, creep, hydrogen embrittlement. Failure mode identification uses visual examination, fractography, material analysis, and dimensional inspection. This step answers the question: how did it fail?
Identify the failure mechanism. Failure mechanism is the chain of contributing factors that allowed the failure mode to occur: stress concentration from an undocumented chamfer, surface finish that failed to retain lubricant, a material substitution that reduced fracture toughness, an operating condition that exceeded the design envelope. This step answers the question: why did it fail?
Identify the root cause. Root cause is the earliest point in the chain where a different decision would have prevented the failure. It is frequently a process or specification gap rather than a material or geometry problem. The root cause answer typically leads to a corrective action that closes a systemic gap rather than just replacing the failed part.
Document and implement corrective action. A failure analysis without corrective action is an academic exercise. The findings must translate to a specific change — a drawing revision, a material specification update, a process control requirement, a design modification — and that change must be verified as implemented. The failure analysis is not complete until the corrective action is in service.
Engineering failures organise into a manageable number of physical mechanisms, each with characteristic signatures that guide the analysis.
Fatigue is the most common failure mode in mechanical components subject to cyclic loading. Fatigue fractures begin at a stress concentration — a notch, a hole, a surface defect, a change in section — and propagate incrementally with each load cycle until the remaining cross-section can no longer support the load. The fracture surface typically shows beach marks (concentric lines that track crack progression) and a distinct final fracture zone. Fatigue failures almost always indicate that the stress at the crack initiation site exceeded the fatigue limit of the material.
Overload failures occur when the applied stress exceeds the material's yield or ultimate strength in a single event or in a small number of events. Ductile overload shows significant plastic deformation before fracture; brittle overload fractures with little deformation. Overload failures indicate either an unexpected load, a material deficiency, or a section too small for the actual operating load.
Wear removes material through mechanical contact. Abrasive wear scratches the surface with hard particles; adhesive wear transfers material between contacting surfaces; surface fatigue produces pitting and spalling from subsurface crack growth. Wear failures indicate inadequate lubrication, wrong material pairing, misalignment, or contamination.
Corrosion attacks material through chemical or electrochemical reaction with the environment. Uniform corrosion is predictable and manageable; pitting corrosion is highly localised and can initiate fatigue cracks at surprisingly small pit depths; stress corrosion cracking combines tensile stress with a specific corrosive environment to produce fractures at stresses well below the normal failure threshold.
Thermal failure occurs when temperature exceeds the material's service rating, causing creep (permanent deformation under sustained load at elevated temperature), thermal fatigue from repeated heating and cooling, or phase changes that alter material properties. Thermal failures are often misdiagnosed as mechanical failures because the evidence — dimensional change, microstructural alteration — requires metallurgical examination to identify.
Most Failures Have Multiple Contributing Factors
Experienced failure analysts rarely find a single root cause. The typical
result is a combination of factors — a stress concentration that reduced the
safety margin, combined with a material that had slightly lower toughness than
specified, in an environment that was slightly more corrosive than the design
assumed. Any one factor alone might have been tolerable; the combination was
not. Document all contributing factors, not just the most obvious one.
Not every failure warrants a full investigation. The decision to invest in analysis should be proportional to the expected return.
Always investigate when: a failure resulted in injury, property damage, or significant production loss; a failure occurred significantly earlier than the design life; a failure pattern is recurring across multiple units; a failure involves a safety-critical component; a customer is requesting a formal corrective action.
Investigate selectively when: a single failure occurs in a non-critical application without safety consequences; the failure is clearly caused by abuse or misuse outside the design envelope; the component is at or near the end of its design life and the failure is consistent with wear-out.
Use results to inform future designs even when formal investigation is not warranted. A part that failed within specification conditions, at a service life shorter than designed, is telling you something about the design margin. Recording the failure mode and service life — even without a full root cause investigation — builds a database that improves future designs.
A small equipment manufacturer was experiencing recurring seal failures in a pneumatic cylinder used in their product. The seals were replaced under warranty at a rate of roughly 40 units per year, each requiring a service visit valued at $1,300. Annual warranty cost from this single failure mode: approximately $52,000.
The failure had been attributed to seal wear — a generic conclusion that had been carried forward for two product generations without correction. A failure analysis was initiated: four failed seals were recovered from the field, examined under 10× magnification, and compared to new seals from the same lot.
The examination revealed that the seal lips showed spiral wear marks consistent with extrusion over the sealing land, rather than the abrasive wear pattern expected from normal service. This indicated that the bore surface finish was rougher than required for the seal material, causing the lip to catch during installation and create a micro-tear that propagated under pressure cycling.
Bore surface finish was measured on several production cylinders using a surface profilometer. The drawing specified Ra 1.4–1.8 µm for the bore. Actual measured values ranged from Ra 2.1 to 3.4 µm — the machining specification was not being met, and the inspection process had no control point for bore finish.
Root cause: no surface finish measurement in the production inspection plan. The drawing specification existed but was not being verified. Corrective action: added bore surface finish to the first-article and periodic inspection plan; retrained the machining operator on finish toolpath parameters; revised the drawing to add a measurement callout flag.
Field failure rate after the corrective action: less than 2 units per year. Annual warranty cost reduction: approximately $49,400. Engineering time invested in the analysis: 14 hours.
The discipline of failure analysis is not a response to crises — it is a routine practice that organisations embed in their design, testing, and quality processes. Teams that do this well share several characteristics.
They treat failed parts as assets, not waste. A failed prototype is not discarded; it is examined before being scrapped. Returned warranty parts are not remanufactured without first recording the failure mode and location. The physical evidence of a failure has more diagnostic value than any amount of simulation.
They separate blame from cause. Failure investigations that focus on who made the decision are ineffective because they incentivise concealment. Investigations that focus on what combination of conditions produced the failure — design margin, material property, process variation, operating condition — are effective because they identify the change that prevents recurrence.
They close the loop. Corrective actions are tracked from finding to implementation to verification. A corrective action that sits in an engineering memo without a drawing revision, a specification update, or a process change has not been implemented.
Failure Analysis as a Design Improvement Loop
The most valuable outcome of a failure analysis is not the repair of the
failed part — it is the improvement to the design or process that prevents the
next failure. Teams that consistently close the loop between field failures
and design updates build products that improve with every generation. Teams
that replace without analysing build products that fail the same way,
generation after generation.
This post introduced the failure analysis framework and the failure categories it addresses. The next post in this series covers visual inspection — the first and most important tool in any failure investigation. Learning to read what a failed surface is telling you, with nothing more than a loupe and good lighting, is a skill that identifies the failure mode in the majority of cases before any laboratory work is required.
Failure analysis converts the cost of a failure into an investment in prevention; skipping analysis converts a single failure into a recurring expense
Failure modes include fatigue, overload, wear, corrosion, and thermal degradation — each with physical signatures that guide diagnosis
The failure analysis process moves from failure definition through root cause identification to corrective action implementation
Most failures have multiple contributing factors; document all of them, not just the most visible one
A failure analysis culture treats failed parts as evidence, separates blame from cause, and consistently closes the loop to design or process improvements
