Cleaning, repeated power cycles, cable flexing, opening, heat, or solvent can change the failure before it is documented.
Returned-unit investigation guide
Sensor Failure Analysis: What Returned Units Can Reveal
A returned sensor can reveal the verified failure mode, physical location, damage mechanism, and possible links to design, production, installation, environment, or use. One unit cannot prove prevalence, liability, or root cause by itself.
Direct answer
Preserve first, reproduce second, open last. A credible sensor failure analysis connects the complaint to measured behavior, physical evidence, application history, matched controls, and a tested causal chain before corrective action is selected.
Background photo: Bulat843 / Pexels.
Burning, corrosion, or fracture must be connected to the symptom, activation stress, production history, and competing explanations.
State the tested conditions, coverage limits, remaining hypotheses, and field data needed to reproduce an intermittent event.
Four rules protect both the customer and the evidence.
A fast replacement may restore the machine, but a useful investigation must also preserve the failed unit, reconstruct the real system, and prove that the final action addresses the cause rather than the visible symptom.
Photograph, identify, quarantine, and record packaging, contamination, connector, cable, damage, and handling before alteration.
Never clean first to make the sample presentable.Use the actual wiring, PLC input, target, mounting, time, temperature, motion, and environment instead of a short nominal bench check.
Measure the raw signal, not only the LED.Compare known-good, retained, same-lot, new-lot, and deliberately stressed samples where the hypothesis requires them.
Separate normal construction from a true anomaly.Connect failure mode, location, mechanism, activating stress, root cause, escape cause, corrective action, and effectiveness.
State uncertainty instead of forcing blame.A returned sensor is a physical record, not a complete verdict.
What makes a finding useful?
It explains the reported symptom, localizes to the abnormal area, differs from relevant controls, matches the known stress history, survives competing hypotheses, and predicts what happens when the proposed cause is introduced or removed.
External marks may show impact or incorrect mounting. A cracked sensing face may expose a moisture path. Corrosion can help map fluid entry. A conductor broken at the strain-relief transition can connect repeated bending to an intermittent open. Electrical measurements may isolate the supply, sensing, logic, communication, or output stage.
These are findings, not automatic root causes. The causal chain becomes stronger only when the physical evidence and electrical behavior agree with the application timeline, good-unit comparison, production trace, and a reproducible mechanism.
A return may reveal
Evidence that supports the investigation
- The confirmed failure symptom and measurable failure mode.
- The failed cable, connector, seal, sensing element, PCB, component, joint, or housing region.
- A physical mechanism such as open, short, fracture, corrosion, overheating, contamination, drift, or delamination.
- Candidate activation stresses such as moisture, chemical, bending, vibration, impact, current, transient, or temperature.
- A weak point that should trigger risk-based containment or population review.
A return cannot prove alone
Conclusions that require more evidence
- That the whole production lot or product family is defective.
- That burn damage identifies the original voltage waveform or responsible party.
- That corrosion must have entered through an original factory seal defect.
- That a current bench pass disproves a field complaint.
- That replacing the damaged part prevents recurrence.
Safety boundary: Do not energize or open an unknown, fire-damaged, chemically contaminated, pressure-exposed, explosive-area, safety-critical, medical, or hazardous-process unit without qualified authorization and the correct product-specific controls.
Method references: NASA Basics of Failure Analysis and IEC 62740:2015 Root cause analysis.
Use precise failure language
“The sensor burned” skips seven important questions.
Precise terms keep facts, interpretations, actions, and responsibility from being mixed together.
Reported symptom
What the user observed: for example, the PLC input flickers after warm-up.
Verified failure mode
The measured loss of required function: output changes despite a stable target.
Failure location
The physical or functional area where abnormal behavior is isolated.
Failure mechanism
The electrical, mechanical, chemical, thermal, or software process producing failure.
Activation stress
The event or condition that initiated or accelerated the mechanism.
Root cause
The controllable design, process, installation, or system cause whose removal prevents recurrence.
Escape cause
Why validation, inspection, monitoring, instructions, or change control did not prevent or detect it.
Corrective action
The verified action that removes the cause, updates controls, and demonstrates effectiveness.
Example chain: intermittent PLC input → intermittent sensor output → conductor behind strain relief → copper-strand fatigue → repeated localized bending → clamp geometry forced the bend → production flex test missed the field bend location → routing, clamp, validation, and instructions are changed and verified.
Before testing begins
Preserve the as-returned condition
The receiving record should show what arrived, how it was packaged, visible damage, contamination, accessories, cut cable length, connector, labels, and whether the unit appears cleaned, opened, repaired, dried, or altered.
1
Inspect the package
Record moisture, shock, loose parts, tamper condition, and possible transport damage.
2
Confirm identity and traceability
Case ID, model, serial/lot/date code, revision, cable/connector, machine position, and customer reference.
3
Photograph before alteration
All sides, sensing face, cable route, connector pins, labels, deposits, fasteners, cracks, and witness marks.
4
Control contamination
Do not wipe, wash, blow, scrape, dry, or chemically swab until the analysis plan permits it.
5
Perform safety triage
Assess shorts, unknown voltage, chemical, pressure, explosive-area, fire, and safety-function risks before power.
6
Establish custody and storage
Record handlers and timestamps; use suitable ESD, moisture, temperature, and sample-seal controls.
First rule: Do not destroy the best evidence while trying to make the failure appear. Plan and record the sequence before power cycles, cable flexing, cleaning, resetting, opening, or cutting.
Staged sensor failure analysis
Move from complaint to corrective action without jumping over evidence.
Each stage should answer a defined question and preserve the option to perform the next stage.
Contain and collect
Assess severity, protect customers, gather machine history, wiring, target, environment, timing, logs, and production trace.
Receive and preserve
Identify, photograph, quarantine, and document transport, handling, contamination, and as-returned condition.
Define hypotheses
Translate the complaint into a measurable failure mode and include sensor, system, application, transport, and test explanations.
Confirm and localize
Use controlled power, raw-signal logging, matched controls, reference targets, thermal behavior, continuity, and nondestructive imaging.
Open at a hold point
Select housing opening, de-potting, cross-section, SEM/EDS, chemical, or mechanical tests only for a defined unanswered question.
Prove and close
Validate root and escape causes, update containment, verify the fix, monitor effectiveness, and update FMEA, controls, and trend coding.
Failure confirmation
Reproduce the symptom without inventing a new failure.
Begin with the lowest-risk condition that can verify the complaint. Use a regulated supply, current limit, correct load, verified pinout, controlled target, known-good cables, and instrumentation that records output rather than relying on an indicator LED.
No output
Verify terminal voltage, pinout, current, load, output type, target, cable, and startup timing.
False switching
Log raw output, then add motor events, neighboring sensors, contamination, vibration, light, or echoes one at a time.
Range drift
Measure against a defined target, then vary real material, size, angle, temperature, voltage, mounting, and deposits.
Warm-up or motion fault
Log time, temperature, current, output, communication, cable sections, connector, and bracket behavior.
Electrical appearance can be changed by the PLC input, leakage, residual voltage, load, wiring, or transient path. Review sensor leakage current and residual voltage before blaming the returned unit.
Escalate analysis deliberately
Start with methods that preserve the largest future option set.
“Nondestructive” does not mean harmless. Power, repeated flex, X-ray dose, heat, vacuum, cleaning, or manipulation can change some evidence. Every method needs a defined question, controls, and a record of what it may alter.
Visual and stereo microscopy
Find impact, cracks, abrasion, heat, bent pins, damaged seals, residue, corrosion, cable strain, or prior repair.
Electrical baseline
Measure open/short, current, output, leakage, residual voltage, communication, switching distance, and insulation where appropriate.
Dynamic monitoring
Capture intermittent behavior during time, temperature, vibration, cable motion, target motion, supply variation, or network load.
Radiography or CT
Localize broken conductors, internal voids, dense-part cracks, solder anomalies, foreign material, or shifted construction before opening.
Thermal localization
Show where energy is dissipated or leakage occurs; a hot spot identifies location, not automatically root cause.
Cross-section or SEM/EDS
Expose interfaces and fine morphology; EDS identifies elements, not the exact compound, source, time, or responsible party.
Method context: JEOL EDS elemental-analysis overview and NASA failure-analysis training.
Destructive hold point: Before opening, de-potting, cross-sectioning, applying solvent, or performing a destructive pull test, document the purpose, approver, method, cut plane, retained material, comparison sample, and evidence that could be lost.
Common sensor failure signatures
A signature supports a hypothesis. It does not replace confirmation.
The same mechanism can create different symptoms, and different mechanisms can create the same visible damage.
| Subsystem or evidence | Possible mechanism | Confirmation path | Root-cause questions |
|---|---|---|---|
| Cable open near sensor exit | Strand fatigue, pull, cold bending, abrasion, cut, or weak termination. | Continuity under controlled bend, CT/X-ray, section, fracture location, and route history. | Was bend radius, bend axis, strain relief, cable type, routing, and flex validation adequate? |
| Connector corrosion or resistance | Liquid or chemical ingress, damaged seal, poor mating, contamination, or electrochemical reaction. | As-received imaging, contact resistance, residue analysis, seal inspection, and fluid compatibility. | Where was the entry path, and were connector, torque, orientation, cleaning, and maintenance controlled? |
| Cracked face or housing | Impact, overtorque, pressure, thermal stress, chemical embrittlement, or material/process anomaly. | Fracture origin, witness marks, dimensions, torque, chemical history, and material comparison. | Was stress within rating, and did the design, installation, and molding process provide enough margin? |
| Corrosion inside potting or PCB | Moisture path, ionic residue, incomplete sealing, face/cable damage, or chemical permeation. | Map corrosion gradient, CT/section, residue analysis, seals, potting, and production records. | Was the barrier defective, damaged, incompatible, or exposed beyond its validated condition? |
| Burned output or supply part | Reverse polarity, overload, short, surge, ESD/EOS, wiring error, defect, or secondary heat damage. | Circuit localization, morphology, current data, protection path, machine waveform, and component analysis. | Which energy path reproduces the evidence, and is the visible damage primary or secondary? |
| Solder or interconnect fracture | Vibration, thermal cycling, board flex, weak joint, contamination, or inadequate process window. | X-ray/section, fracture morphology, stress reproduction, lot data, and fixture/process history. | Did design strain, assembly, support, potting, vibration, and screening match the real use? |
| Coil or ferrite anomaly | Wire open/short, termination failure, cracked or moved ferrite, winding/position variation, or impact. | Resistance, inductance/impedance, CT/section, target response, and production comparison. | Is the anomaly from manufacture, potting/fixture shift, external impact, or normal construction? |
| Stable sensor, false system count | Target dwell, input filter, switch chatter, scan timing, edge logic, or interference. | Synchronized sensor output, PLC input, target motion, machine speed, and program timestamps. | Are sensor margin, machine geometry, controller timing, and software behavior allocated correctly? |
Electrical damage needs correlation: a shorted output transistor or melted metallization proves local damage, not the original waveform. Review what short-circuit protection can and cannot prevent, reverse-polarity protection, and sensor EMC when building the electrical hypothesis.
No fault found
NFF means the fault was not reproduced under the conditions tested.
It does not prove the field event was imaginary. Transport may cool, dry, relax, disconnect, or reposition the unit. A loose contact can re-seat. A transient disappears. The target, bracket, PLC threshold, cable route, chemical, vibration, or timing may be absent on the bench.
A good NFF report states test coverage, limits, remaining hypotheses, recommended field instrumentation, and clear reopen criteria.
Reference: NASA intermittent failures and No-Fault-Found terminology.
Supply, PLC input, target, background, bracket, neighboring metal/sensors, cable, connector, parameter set, and timing.
Capture voltage, current, output, communication errors, and temperature through warm-up and normal duty.
Monitor cable sections, connector, housing, bracket, and joints under controlled movement or thermal profile.
Log the waveform at sensor terminals and PLC input, plus motor, welder, solenoid, grounding, and suppression events.
Compare raw values with original production data and matched good units for marginal but meaningful shifts.
Document cannot-duplicate coverage, remaining hypotheses, field data request, sample disposition, and reopen triggers.
How to prove root cause
The best photograph is not automatically the best explanation.
A strong cause explains the symptom, matches the timeline, distinguishes failed from good units, survives competing hypotheses, and predicts the result of introducing or removing the cause.
Does the proposed mechanism produce the verified failure mode electrically and physically?
Is abnormal behavior tied to the same damage site through probing, imaging, sectioning, or substitution?
Is the causal feature or response abnormal relative to matched controls and the relevant population?
Can controlled stress create comparable electrical, physical, and timing signatures?
Does eliminating the cause prevent the failure under the causal stress?
Does the trend follow lot, line, site, use, age, exposure, or process window as predicted?
Do discriminating tests weaken other credible explanations, with remaining uncertainty stated?
Which validation, inspection, instruction, monitoring, or change-control gap allowed the issue through?
IEC 62740 describes RCA as an analysis of events that have occurred and notes that RCA techniques are not designed to assign responsibility or liability. See the official IEC record.
Report confidence clearly
Confidence levels are better than false certainty.
Use wording that separates measured fact from inference and makes missing evidence visible.
Confirmed finding
Directly measured or observed with adequate method and traceability. Example: an intermittent open was localized behind the strain-relief exit.
Confirmed mechanism
Physical and electrical evidence plus controls support the process. Example: conductor fatigue produced the intermittent open.
Probable cause
The best supported explanation, but one original condition or causal link remains incomplete.
Possible contributor
Evidence is compatible but does not discriminate it from other causes.
Not determined
The mechanism may be confirmed while the initiating event, source, or complete root cause remains unresolved.
Close the reliability loop
A useful analysis changes the system, not only the returned unit.
Replacement is a correction. Permanent corrective action must address the verified occurrence cause and escape cause, then demonstrate that the action works under representative production and application conditions.
- Repeat the causal stress and verify the original mechanism does not recur.
- Check worst-case material, process, voltage, target, temperature, vibration, chemical, and time conditions relevant to the cause.
- Confirm the fix does not create reduced range, slower response, heat, false switching, stiffness, or sealing stress.
- Update FMEA, drawing, work instruction, validation, control plan, service guidance, and change communication.
- Trend recurrence, NFF rate, production data, and field exposure before formal closure.
Frameworks: ASQ Eight Disciplines (8D), DLA MIL-HDBK-2155 FRACAS, and IEC 60812:2018 FMEA/FMECA.
Supplier report checklist
What a strong failure-analysis report should contain
A reviewer should be able to trace the unit, understand what each test could prove, distinguish observation from conclusion, and see why the action addresses the verified cause.
Verified failure mode, mechanism/cause status, affected risk, containment, next decision, and unresolved items.
Model, revision, serial/lot, dates, application, reported symptom, and handling or return history.
Packaging, photographs, labels, seals, contamination, damage, cable/connector, tampering, and prior repair.
Hypotheses, controls, sequence, equipment, conditions, limits, software, deviations, and destructive hold points.
Values, images with scale/location, waveforms, control comparisons, attempts, anomalies, and negative results.
Location, mode, mechanism, activation stress, root and escape causes, alternatives, and confidence.
Affected population, containment rationale, correction, occurrence action, escape action, owners, and change control.
Reproduction after fix, qualification and production evidence, field metric, observation period, and closure criteria.
Reliability-data warning: do not calculate a field failure rate from returned units alone. Some failed units are never returned, some returns are good units or duplicates, reporting delay varies, and shipped quantity may not represent operating exposure.
Information to include with an RMA
Better return data reduces diagnosis time and NFF cases.
The best laboratory cannot reconstruct information that was never recorded. Capture the failure before removing the sensor or changing the machine state.
Include photos, raw signals, machine position, and what changed after replacement. Avoid a loose sensor labeled only “bad.”
Output/system state, frequency, duration, recovery, first occurrence, and detection method.
Sensor, cable/connector, controller/input, parameters, serial/lot, machine and position.
Diagram/photo, terminal voltage, load/input, grounding, shield, junctions, and cable length.
Material, size, color, angle, gap, speed, runout, background, reflector/magnet, and bracket.
Temperature profile, liquid/chemical, cleaning, vibration, impact, dust/chips, light, or sound.
Logs, video, waveforms, alarms, maintenance, nearby motor/welder, power cycle, and replacement result.
Removal, cable cut, drying, cleaning, later power, storage, shipment, opening, or repair.
Downtime, scrap, recurrence, affected units, process consequence, and required response timing.
Illustrative return scenarios
How evidence changes the conclusion
These examples are reasoning patterns, not universal diagnoses.
Misses a target after warm-up
A two-minute room test passes. The analyst logs output and current from cold start, recreates the bracket and target, and cycles temperature within rating. Switching margin drifts near the application gap while matched controls remain stable.
Lesson: time, temperature, geometry, and population comparison were required to confirm the complaint.Random output loss on a moving cylinder
The sensor body works, but low-stress monitoring localizes an intermittent open at the same cable location seen in earlier returns. Sectioning shows strand fatigue, while field photos show bending at the clamp rather than through the intended loop.
Lesson: routing, clamp design, validation, and instructions matter; replacement alone would not prevent recurrence.Burned three-wire output stage
Visible heat damage does not measure the original transient. The investigation maps the path, checks protection components, reconstructs PLC/load wiring, captures machine waveforms, and performs controlled comparison testing.
Lesson: physical damage must be correlated with the application circuit before assigning source or responsibility.
Returned sensor review
Begin the investigation with preserved evidence.
Share the complete sensor model, serial or lot, exact symptom, wiring and PLC input, target and mounting, event timeline, environment, logs, photographs, replacement result, and handling history. xsz sensor can use that information to define the first safe diagnostic steps and evidence gaps.
Send these first
- Model, revision, serial/lot, and machine position
- Exact symptom, timestamps, recovery, and recurrence
- Terminal voltage, PLC input/load, wiring, and cable route
- Target material, dimensions, distance, angle, and speed
- Temperature, liquid, chemical, vibration, light, or noise
- Photos, video, logs, waveform, and post-failure handling
Frequently asked questions
Common questions about sensor failure analysis
What is sensor failure analysis?
Sensor failure analysis is the structured examination of complaint data, application conditions, electrical behavior, physical evidence, and comparison samples to determine how a sensor failed, where the failure is located, which mechanism produced it, and what conditions activated that mechanism. It supports the wider root-cause and corrective-action process.
Should a returned sensor be powered immediately?
Not automatically. First document the as-returned condition and assess electrical, chemical, pressure, fire, explosive-atmosphere, stored-energy, and safety-function risks. Power can enlarge damage, dry moisture, erase volatile evidence, or change an intermittent state.
Can one returned sensor prove a batch problem?
No. One return can reveal a real mechanism and justify risk-based containment, but prevalence requires similar returns, production trace, retained samples, screening, lot or line data, application exposure, and a defined population.
What does no fault found mean for a returned sensor?
No fault found means the reported fault was not detected under the conditions tested. It does not prove the field event did not occur. The report should state test coverage, limitations, remaining hypotheses, field data needs, and reopen criteria.
What is the difference between a failure mechanism and root cause?
The failure mechanism is the physical, chemical, electrical, mechanical, or software process that produced failure, such as conductor fatigue. Root cause explains why the system allowed that mechanism, such as a clamp forcing repeated bending at one stress concentration while validation missed the field route.
Does corrosion always mean the factory seal was defective?
No. Liquid may enter through an original sealing defect, damaged face, cable, connector, incorrect mating, chemical attack, pressure or temperature cycling, or later handling. The ingress path, exposure, materials, and construction must be correlated before deciding root cause.
Can burn marks prove overvoltage or ESD?
Burn marks prove local heating and damage, not the exact waveform or source. Overload, short circuit, surge, ESD/EOS, reversal, internal defect, thermal escape, and secondary damage may overlap. Map the failed path and correlate it with the machine circuit and measurements.
When are X-ray, cross-sectioning, SEM, or EDS needed?
Use advanced methods when they answer a defined hypothesis that lower-risk tests cannot. X-ray or CT can localize internal features before opening; cross-sectioning exposes interfaces; SEM shows fine morphology; and EDS identifies elements in a selected region. Each has preparation and interpretation limits.
What should be included in a sensor 8D report?
Include a quantified problem definition, affected scope, containment, returned-unit identity and evidence, verified failure mode, root and escape causes, validated corrective actions, implementation evidence, effectiveness metrics, and updates to related designs, processes, controls, instructions, or risk analyses.
How long should sensor failure analysis take?
There is no universal lead time. A visible cable break may be resolved quickly, while intermittent thermal, moisture, EMC, chemical, or semiconductor failures may require controlled reproduction, external laboratory work, and field data over time. Agree on staged outputs and decision dates.
Technical references
Standards and primary technical sources
Use the exact sensor documentation, qualified laboratory procedures, contractual requirements, and applicable safety rules for a real investigation.
- IEC 62740:2015, Root cause analysis.
- IEC 60812:2018, Failure modes and effects analysis (FMEA and FMECA).
- ASQ, Eight Disciplines (8D) problem solving.
- DLA ASSIST, MIL-HDBK-2155 Failure Reporting, Analysis and Corrective Action Taken.
- NASA, Basics of Failure Analysis.
- NASA, intermittent failures and No-Fault-Found terminology.
- ISO 9001 Auditing Practices Group, nonconformity review and closing guidance.
- OMRON proximity-sensor precautions for surrounding metal, mutual interference, surge, and installation.
- Analog Devices reliability and EOS/ESD technical resources.
- JEOL Energy Dispersive X-ray Spectrometer overview.
Scope note: This is a general engineering and supplier-qualification guide. It does not replace functional-safety validation, hazardous-material controls, explosive-atmosphere rules, legal evidence procedures, product instructions, or a qualified product-specific investigation plan.