RAMSynapse
Log inSign up

Reliability Prediction · Chapter 6

Common Pitfalls

The mistakes seen in practice and the guards against them.

Every pitfall below has been watched happening on real programmes; the guards are correspondingly practical.

PitfallWhat it looks likeThe guard
Selling the number as a field forecast"The prediction says 48,600 h, so the fleet will see it"Quote predictions as comparative estimates; track field MTBF separately
Unit mix-upsFIT and f/10⁶ h blended in one roll-up — a factor of 1,000Declare units at the top; convert at the boundary, once
One environment for a multi-phase lifeA missile electronics box predicted entirely in GFSegment the mission; time-weight per-phase results
Extrapolating the modelsStress ratios above 1.0, temperatures past the tablesTreat out-of-table parts as derating findings, not inputs
Forgetting the unglamorous contributorsNo PCB, solder-joint, or connector terms in the roll-upFollow the handbook's roll-up procedure to the letter
Optimistic quality claimsCommercial parts carried at military πQ "because the vendor is good"Evidence for the claimed screening, or the honest factor
Ignoring duty cycle and dormancyCalendar hours fed into operating-hour modelsDefine operating time explicitly; profile on/off phases
Quoting MTBF without its clockOperating-hour prediction argued against calendar-hour field dataState the clock with every MTBF; publish both views when duty cycle < 100%
Wear-out items left in the λ sumA life-limited item carried at a constant rate with no replacement planScreen against mean life; pull flagged items into conditional-reliability treatment
One-shot items on the hourly clockA squib or cartridge carrying a per-hour failure ratePer-demand probability, combined at the demand point — never in the λ sum
Silent zero from an unsupported environmentA model asked for an environment it does not define returns a perfect partTreat any zero rate as a defect in the analysis until proven deliberate
Double-counting field evidenceField-based specified rates further "corrected" by field-data methods or factorsOne evidentiary basis per rate, recorded with the rate
Chasing the number, not the designAssumptions tuned until the target is metThe handbook's own warning: those who treat the prediction as a number to exceed will find a way — without improving anything
Frozen-handbook blindness1995 models applied to parts that did not exist in 1995Say so in the assumptions; consider newer model sets or field data for modern parts

Most of these are process failures, not mathematics failures, and the last two deserve a sentence each. The gaming problem is old enough that the handbook warns about it explicitly — a prediction whose assumptions were tuned to clear a threshold has spent its credibility for one review meeting. And the frozen-handbook problem is structural: MIL-HDBK-217F Notice 2 predates three decades of component evolution, which is precisely why the serious modern practice is to treat handbook predictions as one pedigree among several — blended with field data where it exists, cross-checked against newer model sets where it matters, and always documented with their assumptions attached.


Want to see this on a live system model? Request a walkthrough.