Your cart is currently empty!
DFSS Instructor Prep · Module 2
Layer B — Engineering Substance
Reliability Physics & Field Failure Modes
The “why does it fail?” foundation that converts DFMEA from vocabulary into engineering, makes ALT (Accelerated Life Testing) plans defensible, and lets you diagnose any participant’s failure-mode statement in real time.
What’s in this module
- Reliability mental model — the bathtub curve & what β tells you
- The five stress drivers behind almost every failure
- The 12 failure mechanisms to know by name
- Acceleration models — Arrhenius, Coffin-Manson, Norris-Landzberg, Peck
- Mapping mechanisms to Yazaki products
- Accelerated Life Testing — design & pitfalls
- DFSS linkage — where PoF meets DMADV
- Instructor facilitation pattern
- Self-check (10 questions)
1. Reliability mental model — the bathtub & what β tells you
Before talking about failure mechanisms individually, it helps to have a single picture in mind: the bathtub curve. Every reliability conversation eventually reduces to “where on this curve does our concern live?”
1.1 What β (Weibull shape parameter) tells you instantly
The Weibull distribution is the workhorse of reliability data analysis because its shape parameter β (sometimes called “slope”) directly indicates which region of the bathtub the failure mode is in:
| β value | Implication | What it suggests |
|---|---|---|
| β < 1 | Decreasing failure rate | Infant mortality — manufacturing defects, process variation. Burn-in would help. |
| β ≈ 1 | Constant failure rate | Random or extrinsic failures. Not age-related. Burn-in won’t help. |
| β = 2–3 | Early wear-out, weak slope | Fatigue/wear mechanisms beginning to dominate. Vibration fatigue often shows β ≈ 2. |
| β > 3 | Steep wear-out | Strong wear-out — corrosion, thermal cycling damage maturing, polymer end-of-life. β can reach 5–8 for late wear-out. |
Reading Weibull data like a grown-up
When a participant presents a Weibull plot from a reliability test, glance at β first. If they claim a manufacturing defect but β is 4, something doesn’t add up. If they claim a wear-out mechanism but β is 0.8, again, mismatch. Mismatched β values almost always reveal mixed failure modes in the data — a sign the failure analysis is incomplete.
2. The five stress drivers behind almost every failure
Every wear-out mechanism is driven by one or more of just five fundamental stresses. Memorising this short list lets you decompose any field failure into its driving physics quickly.
| Stress | What it does | Yazaki examples |
|---|---|---|
| Steady-state temperature (T) | Drives chemical reaction rates, diffusion, polymer ageing, plating intermetallic growth | Under-bonnet harness, HUD on dash, BMS in pack |
| Temperature cycling (ΔT, dT/dt) | Drives thermo-mechanical fatigue — solder joints, wire-bond, plastic-metal interfaces | SMT joints in cluster/HUD/BMS, terminal-to-PCB joints in PDU |
| Mechanical (vibration, shock, flex) | Drives fatigue cracks, fretting, conductor strand breakage, connector unmating | Engine-bay harness, door-flex harness, underbody runs |
| Humidity / moisture | Drives corrosion, electrochemical migration, dendrite growth, dielectric breakdown | Failed grommet → wet harness, sealed connector seal compression set |
| Electrical (V, I, transients) | Drives electromigration, dielectric breakdown, arc damage, contact welding | HV cables, contactors in PDU, fuse box bus-bars |
Instructor pattern
When a participant describes a field failure, immediately classify it by asking: “Which of the five stresses drove this? Was it dominantly thermal, mechanical, electrochemical, or electrical?” This converts a vague complaint into a physics statement and points directly at the relevant DFMEA cells and ALT design.
In real life, stresses combine
Most automotive failures are combined-stress failures. Fretting corrosion needs vibration and humidity and temperature. Solder fatigue needs ΔT and mismatched CTE materials. This is why single-stress ALT often under-predicts field life — and why combined-stress test design (HALT, HASS) is increasingly common.
3. The 12 failure mechanisms to know by name
If you can speak fluently about these twelve mechanisms, you can engage almost any failure-mode conversation at Yazaki. Each card below gives the mechanism, the driver, where it appears, and the acceleration model.
3.1 The metallic / interconnect family
1. Fretting corrosion (terminal contacts) — THE dominant connector failure
The mechanism: small relative motion between mated terminals exposes fresh metal to air, which oxidises. Tin oxide is electrically insulating and hard. Each fretting cycle adds another oxide layer. Contact resistance climbs from milliohms to ohms over time, eventually causing intermittent or open circuit.
Why tin is the worst plating for this: Tin has soft mechanical properties (deforms easily under load) but oxidises rapidly. The oxide is harder than the substrate, builds up at the contact zone, and is repeatedly broken and re-formed by motion. Gold plating largely eliminates fretting because gold doesn’t oxidise; silver is in between.
What makes it worse: higher temperatures (oxidation rate ↑), higher fretting frequency, lower normal force, lower current (no self-heating to break through oxide).
Instructor-level nuance
Fretting corrosion of tin-plated contacts is widely cited as the single most common failure mechanism in automotive connectors. The temperature dependence is non-monotonic: tin softens around 100 °C (helps re-establish contact); above 125 °C, oxidation dominates and contact resistance rises rapidly. Cu-Sn intermetallic compound growth is also a factor at elevated temperatures.
2. Solder joint thermo-mechanical fatigue (SMT components)
The mechanism: every thermal cycle imposes a plastic shear strain on the solder joint because the component and the PCB expand by different amounts. Plastic strain accumulates micro-cracks; cracks propagate; joint opens.
Where this hits Yazaki: any product with SMT components — cluster, HUD, BMS, PDU, smart fuse box, sensors. BGAs and large ceramic packages are the worst because they have larger LD (distance from neutral point) and worse CTE mismatch.
Mitigation levers: reduce ΔT range, use compliant lead frames, match CTEs, underfill BGAs, smaller package sizes, control PCB warpage.
3. Vibration fatigue of conductors
Conductor strands work-harden and crack at points of repeated flex. The classic field signature is intermittent open circuit — wire conducts when in one position, opens when moved. Often misdiagnosed as connector problem.
Yazaki context: engine-bay harness near flexible mounts, door harnesses through grommet, hood/tailgate harness, e-axle harness on EV.
4. Wire bond & lead fatigue (semiconductors)
Distinct from solder fatigue because the cycle is driven by the device’s own self-heating (power cycling), not by ambient. Critical in power semiconductors — IGBTs/SiC MOSFETs in inverters and contactors, and in high-current PDU paths.
5. Electromigration
Electrons physically push metal atoms along high-current traces. Voids form at the cathode end; hillocks at the anode end. Aluminum traces are most susceptible; copper much less so. In Yazaki products: rare except at very-high-current bus bars and IC interconnects.
3.2 The corrosion / electrochemical family
6. Electrochemical migration (ECM) / dendrite growth
Under DC bias in the presence of moisture, metal ions migrate from one trace to another. Dendrites (metal whiskers) grow from cathode to anode. Eventually shorts the gap. Classic mechanism between adjacent PCB traces and between connector pins at narrow spacing.
Made worse by: flux residues, fingerprints, road salt ingress, no-clean process residue.
7. Galvanic / atmospheric corrosion
Two different metals in contact + an electrolyte = battery → one metal corrodes. Aluminum-copper joints are notorious. Mitigated by: matching metals, intermediate buffer layers, sealing the joint from electrolyte, or accepting a corrosion allowance.
8. Crevice / pitting corrosion of terminals
Wherever water can wick in and not dry out — under seal lips, between un-mated/partially-mated terminal blades, inside un-vented housings. Causes pitting; eventually conductor break.
9. Tin whiskering
Pure-tin plating (post-RoHS replacement of Sn-Pb) can grow filamentary “whiskers” that bridge adjacent contacts. Particularly dangerous in fine-pitch parts. Mitigation: lead-bearing matte tin (where allowed), satin tin, nickel underlayer, annealing.
3.3 The polymer / insulator family
10. Polymer / insulation ageing
PVC plasticiser migration → embrittlement. XLPE oxidative degradation. Silicone shrinkage at very high T. Rubber seal compression set → IP rating loss. All these proceed by Arrhenius kinetics — every 10 °C roughly doubles the reaction rate.
Yazaki impact: under-bonnet harness insulation cracks (classic 8–12 year warranty issue), HV cable jacket degradation, sealed-connector seal hardening leading to moisture ingress.
11. Conductive Anodic Filament (CAF) in PCBs
Internal short between adjacent vias inside a PCB. Hard to detect, catastrophic when it occurs. Mitigated by PCB material choice (CAF-resistant epoxies), via spacing rules, and conformal coating.
12. Partial discharge & dielectric breakdown (HV)
In HV cables/connectors, any insulation defect (void, contamination, sharp edge) causes local field concentration. At sufficient voltage, partial discharge starts — small ionising events that don’t fully short, but progressively erode insulation. Eventually leads to full breakdown.
This is one of the reasons HV products require 100% hi-pot test in production — a passing hi-pot result rules out gross insulation defects that would otherwise nucleate partial discharge in service.
Instructor recall framework
A useful mnemonic for tying these twelve back to the five stresses:
- Mechanical-driven: #1 fretting, #3 vibration fatigue, #4 wire bond fatigue (partly)
- Thermal-cycle-driven: #2 solder fatigue, #4 wire bond fatigue
- Steady-T-driven: #5 electromigration, #10 polymer ageing
- Humidity/electrochem-driven: #6 ECM, #7 galvanic, #8 crevice, #9 whiskers, #11 CAF
- Electrical-stress-driven: #12 partial discharge / dielectric breakdown
4. Acceleration models — Arrhenius, Coffin-Manson, Norris-Landzberg, Peck
Four models cover the vast majority of automotive reliability conversations. You don’t need to memorise every coefficient — just know which model applies to which stress, what its variables are, and roughly what acceleration factor you’d expect.
4.1 Arrhenius — for chemical / diffusion-rate-limited failures
Arrhenius acceleration factor
AF = exp[ (Ea / k) × (1/Tuse − 1/Ttest) ]
Where Ea = activation energy (eV), k = Boltzmann’s constant (8.617 × 10⁻⁵ eV/K), T in Kelvin.
Applies to: polymer ageing, intermetallic growth, electromigration, dielectric ageing, corrosion (with humidity term added). Anything reaction-rate-limited.
| Mechanism | Typical Ea (eV) |
|---|---|
| Aluminum electromigration | 0.5–0.7 |
| Dielectric breakdown (TDDB) | 0.6–1.0 |
| PVC plasticiser loss / embrittlement | 0.9–1.1 |
| Solder intermetallic growth | 0.7–1.0 |
| Connector contact oxidation | 0.6–0.9 |
The rule-of-thumb every reliability engineer uses
For most thermally-activated chemical mechanisms with Ea ≈ 0.7 eV, every 10 °C of temperature increase roughly doubles the reaction rate (the so-called “10 °C rule”). This is a back-of-envelope sanity check, not a precise number. You’ll hear it everywhere; you should be able to challenge it (“for which Ea, in which temperature range?”) when it’s misapplied.
4.2 Coffin-Manson — for low-cycle plastic-strain fatigue
Coffin-Manson (simplified for ΔT cycles)
Nf = C × (ΔT)−n
Where Nf = cycles to failure, ΔT = thermal cycle range, n = Coffin-Manson exponent. For solder alloys, n ≈ 1.9 (SnPb eutectic), n ≈ 2.5–3 (SAC alloys, lead-free). C is material/geometry dependent.
Applies to: any low-cycle fatigue driven by plastic strain — solder joints, PTH (plated-through-hole) cracking, wire-bond lift-off.
Why the exponent matters
A higher n means the model is more sensitive to ΔT — so doubling the test ΔT gives you a much bigger acceleration. This is why thermal cycle tests at −40 to +125 °C (ΔT = 165 °C) accelerate field conditions of, say, 0 to +70 °C (ΔT = 70 °C) by a large factor. Get the exponent wrong and your test plan over- or under-promises.
4.3 Norris-Landzberg — Coffin-Manson with frequency & temperature terms
Norris-Landzberg acceleration factor
AF = (ΔTtest/ΔTuse)n × (fuse/ftest)m × exp[(Ea/k)(1/Tmax,use − 1/Tmax,test)]
Norris-Landzberg is the refined version of Coffin-Manson that adds: (i) a cycle frequency term (m ≈ 1/3 typically), and (ii) an Arrhenius term using the maximum cycle temperature. Used most commonly for solder fatigue prediction.
Why this matters for ALT planning: if your test runs cycles at 4 per hour but the field sees 4 per day, the frequency correction is a factor of 24 to the m. Without it, you’d think your test was more aggressive than it really is in field-equivalent time.
4.4 Peck — humidity-temperature combined
Peck model
AF = (RHtest/RHuse)m × exp[(Ea/k)(1/Tuse − 1/Ttest)]
m typically ≈ 3 for moisture-driven corrosion. Used for damp-heat testing — 85 °C / 85% RH (“85/85”) is the classic stress test, with Ea ≈ 0.7–0.9 eV depending on the mechanism (corrosion, ECM, popcorning).
Applies to: connector seal degradation, PCB ECM/CAF, plating corrosion, moisture-driven dielectric ageing.
DFSS framing for the four models
In a DMADV Verify phase, the choice of acceleration model is what links your test plan to your field life claim. If a project claims “we tested for 1000 hours therefore 15-year life”, you must be able to ask: which model? Which Ea or n? What’s the acceleration factor calculation? Is the claim defensible? This is one of the most common quiet failure points in DFSS projects — and where instructor authority earns its keep.
5. Mapping mechanisms to Yazaki products
This is the table you’ll refer to most often when grading participant DFMEAs. For each product family, here are the failure mechanisms that dominate.
| Product | Dominant failure mechanisms | Key acceleration tests |
|---|---|---|
| LV harness | #1 fretting at connectors; #3 vibration fatigue; #7 galvanic corrosion at grounds; #10 PVC ageing; chafe-through (not mechanism per se) | Thermal cycle, vibration random, salt spray, 85/85, accelerated aging at elevated T |
| HV harness | All of LV plus #12 partial discharge; #10 XLPE/silicone ageing; shield termination fretting | HV cycling, partial discharge inception test, hi-pot at 2× rated, ISO 6469-3 isolation |
| Connectors / terminals | #1 fretting (primary), #8 crevice corrosion, #9 tin whiskering (fine-pitch), seal compression set | Vibration + thermal cycle combined, USCAR-2 environmental, mate-unmate cycling |
| Fuse box / JB / relay box | #1 fretting at terminals; #2 solder fatigue on PCBA; relay contact welding/erosion (electrical wear); housing creep | Power cycling, thermal cycle, mechanical shock, relay endurance |
| PDU / BMS | #2 solder fatigue (large packages); #4 power cycling of contactor/IGBT; #6 ECM at high-density PCB; partial discharge in HV section; sensor drift | Power cycle, thermal cycle, 85/85 biased, hi-pot, ISO 26262 fault-injection |
| Cluster / HUD | #2 solder fatigue; LED/laser lumen depreciation; LCD dead-pixel growth; optical adhesive UV degradation; capacitor electrolyte loss | Thermal cycle, humidity-bias, UV chamber, vibration, lumen maintenance test |
| Sensors (current, fuel, temp) | Calibration drift; #1 fretting at internal connector; corrosion of exposed elements; magnetic-material ageing | Long-soak at temperature, thermal cycle with bias, salt spray, vibration |
6. Accelerated Life Testing — design & pitfalls
ALT is the practical bridge from physics-of-failure to field-life claims. Most DMADV Verify phases include an ALT. This section equips you to facilitate ALT design conversations credibly.
6.1 The ALT design sequence
- Define the failure-mode of interest. “We want to detect fretting at the X-terminal” is a real ALT; “we want to test reliability” is not.
- Identify the dominant stress driver. (Refer back to the five stress drivers and the 12 mechanisms.)
- Select the acceleration model. Arrhenius? Coffin-Manson? Peck? Combined?
- Set the test stress level. Hot enough to accelerate, but not hot enough to introduce new failure modes that don’t exist in the field. Crossing the glass transition temperature of a polymer, or melting solder, would invalidate results.
- Calculate the acceleration factor (AF). From the model.
- Set sample size and test duration. Using statistical reliability demonstration (binomial / Weibull-based), e.g. Reliability R = 0.97 at C = 90% confidence with zero failures requires n ≈ 75 samples tested to one lifetime.
- Define failure criteria precisely. “Open circuit”? “Contact resistance > 100 mΩ”? “Loss of seal IP67”? Vague criteria invalidate tests.
- Run, record, and Weibull-analyse. Check β — does it match the expected mechanism?
The four common ALT pitfalls — each a “gotcha” for participants
- Over-stress changes the mechanism. Heating a connector to 200 °C may melt solder or soften plastic — no longer accelerating the field-relevant fretting mechanism.
- Single-stress test misses combined-stress failures. Pure thermal cycling won’t show fretting (no vibration component).
- Wrong acceleration model applied. Using Arrhenius for solder fatigue (which is plastic-strain-driven, not chemical) — common error.
- Censored data misinterpreted. Test ends with no failures = “passed”, but with too few samples and too short a test, this proves nothing statistically.
Instructor pattern for grading an ALT plan
When a participant presents an ALT plan in Verify, ask in this order:
- “Which failure mechanism is this test designed to provoke?”
- “What’s your acceleration model and AF calculation?”
- “What new mechanisms might emerge at this stress level?”
- “What sample size at what confidence? Show the math.”
- “What’s your failure criterion, in measurable terms?”
7. DFSS linkage — where PoF meets DMADV
| DMADV Phase | How PoF literacy makes you a better facilitator |
|---|---|
| Define | Help participants identify warranty-relevant CTQs by anchoring them in real field-failure mechanisms (“if your product fails by fretting, then the CTQ is contact resistance stability over 15 years, not just initial resistance”). |
| Measure | CTQ measurability — many reliability CTQs are predicted, not directly measured during a project. PoF-based prediction is the bridge. Ensure the “Y” being measured can actually be tied to a failure mechanism. |
| Analyze | This is the big one for DFMEA. The “Failure Mode” cell of every DFMEA row should map to one of the 12 mechanisms (or a known combination). Generic language (“part fails”) gets rewritten as “tin terminal fails by fretting corrosion under combined thermal cycling and vibration”. This single shift transforms DFMEA quality. |
| Design | Robust-design noise factors come straight from the five stress drivers. P-diagram outer-array noise levels can be derived from field-condition data. Tolerance design must consider which parameter variation drives which mechanism (e.g., normal force variation → fretting rate). |
| Verify | ALT design (above). Reliability demonstration math. Linking field-life claims to test acceleration models with defensible numbers. |
The “physics statement” rule for DFMEA
Encourage participants to write every DFMEA failure-mode entry as a physics statement: “[component] fails by [mechanism] under [stress driver]”. Examples:
- Bad: “Connector fails.” → not testable, not preventable.
- Better: “Connector loses continuity.” → describes effect, not mechanism.
- Good: “Tin-plated terminal contact resistance rises > 100 mΩ due to fretting corrosion under combined thermal cycling and vibration over 10 years.” → testable, accelerable, designable-against.
8. Instructor facilitation by function
| Function | Reliability-physics angle that lands |
|---|---|
| WH (LV/HV) | Vibration fatigue, fretting at every connector, polymer ageing, partial discharge for HV |
| Testing Center | This is their home turf — they should be your ally. Ask: “Which mechanisms do your current EOL tests actually catch? Which do they miss?” |
| EI — AR HUD / Cluster | SMT solder fatigue (Coffin-Manson), LED/laser lumen depreciation (Arrhenius), optical adhesive UV degradation, electrolytic capacitor dry-out |
| EI — Sensors | Calibration drift (Arrhenius), internal connector fretting, magnetic ageing |
| CDDC — Connectors/Terminals | Fretting (THE mechanism), tin whiskering at fine pitch, seal compression set |
| CDDC — Fuse/JB/Relay | Relay contact erosion (electrical wear), bus-bar joint fretting, solder fatigue on PCBA |
| CDDC — BMS/PDU | Power cycling fatigue of contactors/IGBTs, partial discharge in HV section, sensor drift, ECM at high-density PCBs |
| Shared Service — Thermal/EMI/CFD | They simulate the stresses; ask: “How well does your simulated ΔT correlate with measured field cycles?” |
| Shared Service — Advance Materials | Polymer ageing models, plating selection vs. mechanism (Sn vs Ag vs Au) |
| Project Mgmt / Innovation | Future-product reliability — wireless BMS, e-fuse contactor replacement, SiC vs Si in PDU |
Instructor self-check
Ten questions calibrated to the kind of statements participants will actually make.
Q1. A Weibull plot of field-return data shows β = 0.7. The participant claims the root cause is wire fatigue. What’s the most likely problem?
A. Sample size is too small
B. The plot was constructed wrong
C. β < 1 indicates infant mortality, not wear-out fatigue — the failure mechanism diagnosis is probably wrong
D. β = 0.7 is what you’d expect for fatigue
Correct — wear-out mechanisms have β > 1, typically 2–5. β < 1 points to infant mortality (manufacturing defects). The diagnosis-data mismatch suggests the failure mechanism is misidentified.
Q2. The single most common failure mechanism in automotive connectors is:
A. Tin whiskering
B. Fretting corrosion at the contact interface
C. Electromigration
D. Conductive Anodic Filament (CAF)
Correct — fretting corrosion of tin-plated contacts is widely cited as the dominant connector failure mechanism.
Q3. A participant says “we’ll test at 125 °C for 1000 hours, that’s 15 years of field life.” What’s your most important challenge?
A. 125 °C isn’t hot enough
B. 1000 hours isn’t long enough
C. Field life should be 10 years, not 15
D. Which acceleration model and activation energy support that claim, and what new failure modes might emerge at 125 °C?
Correct — the claim is meaningless without an explicit acceleration model + Ea, and you must also check whether the test temperature introduces new mechanisms.
Q4. The Coffin-Manson model is most appropriate for:
A. Solder joint fatigue under thermal cycling
B. PVC insulation ageing
C. Electrochemical migration
D. Tin whisker growth
Correct — Coffin-Manson is a plastic-strain low-cycle fatigue model, classically applied to solder joint thermo-mechanical fatigue.
Q5. The “10 °C rule” (every 10 °C doubles the reaction rate) is most accurate when:
A. The mechanism is mechanical fatigue
B. The temperature is above 200 °C
C. The mechanism is Arrhenius-type with Ea around 0.7 eV, in the normal automotive temperature range
D. The mechanism is electromigration
Correct — the 10 °C doubling rule is an Arrhenius approximation for moderate Ea around 0.7 eV. It’s misapplied when Ea is very different or for non-Arrhenius mechanisms.
Q6. A DFMEA row says “Connector fails open.” How do you improve it?
A. Add a severity rating
B. Rewrite as a physics statement: which mechanism, under which stress, over what life
C. Combine it with another row
D. Delete it
Correct — physics-grounded language (“contact resistance rises > 100 mΩ due to fretting under combined thermal cycling and vibration”) makes the entry testable and designable-against.
Q7. Norris-Landzberg adds which terms to Coffin-Manson?
A. Humidity and voltage
B. Sample size and confidence
C. Cycle frequency and an Arrhenius (peak T) term
D. Mechanical shock
Correct — Norris-Landzberg is the standard refinement that adds cycle frequency (slow cycles damage more than fast) and a maximum-temperature term.
Q8. In a P-diagram for an HV connector, “salt spray during winter road treatment” is best classified as:
A. A signal factor
B. An external noise factor
C. A control factor
D. A response variable
Correct — salt spray is an environmental stress the design must be robust against. Sealing, drainage geometry, and plating choice are the control factors.
Q9. Why is 100% Hi-Pot testing required on HV connectors in production?
A. It screens out gross insulation defects that would otherwise nucleate partial discharge in service
B. To verify mating force
C. To check CAN bus continuity
D. To verify the contact plating thickness
Correct — Hi-Pot detects insulation weaknesses that, once in service, would slowly progress to dielectric breakdown via partial discharge.
Q10. Five participants present field-failure data. Which “diagnostic question” is most universally powerful for you to ask first?
A. “What’s the unit cost?”
B. “Who’s the customer?”
C. “Was this a Tier-1 or Tier-2 defect?”
D. “Which of the five stress drivers caused this, and which of the 12 mechanisms is consistent with the evidence?”
Correct — this single question forces physics-grounded thinking and immediately reveals whether the participant has a real diagnosis or just a symptom.