Preventive maintenance intervals deserve evidence, but a screened number is not enough evidence by itself. A Weibull age-replacement model can make assumptions visible: MTBF, beta, planned PM cost, failure cost, current interval, and the shape of the local cost curve. Those inputs still need source-backed failure data and qualified interpretation before a PM schedule changes.
This guide frames the source gaps around a PM interval screen. It covers what beta and eta mean, why censored records and failure-mode separation matter, how the age-replacement cost formula behaves, and where OEM manuals, condition monitoring, safety, fleet compliance, warranty, and reliability engineering review control the actual maintenance decision.
Why PM Interval Screens Need Boundaries
Too-frequent PM can replace useful life and introduce maintenance-induced failures. Too-infrequent PM can increase downtime, collateral damage, safety exposure, and emergency repair cost. A cost curve can help compare assumptions, but it cannot decide whether the assumptions are valid.
The age-replacement model depends on the failure distribution and the cost ratio. In a real maintenance program, those are not just numbers. They come from CMMS records, censored survivor records, failure-mode classification, cost accounting, OEM limits, duty cycle, condition monitoring, and criticality review.
Use a screened interval to identify questions: Is beta source-backed? Are failures from one mode? Are suspensions included? Does the equipment have safety, environmental, warranty, fleet, or regulatory requirements that override local economics?
PM Interval Optimizer
Optimize preventive maintenance frequency using Weibull reliability analysis and cost-risk modeling. Find the interval that minimizes total maintenance cost.
Weibull Distribution: What the Parameters Mean
The two-parameter Weibull model uses shape beta and scale eta. Beta describes the failure-rate trend and eta is the characteristic life. A screen can compute eta from MTBF and beta, but the beta value itself has to come from valid failure data or qualified engineering assumptions.
Beta less than 1 suggests decreasing hazard, beta near 1 suggests roughly constant hazard, and beta greater than 1 suggests increasing hazard. Those labels are only useful when the data population is clean. Mixed failure modes, installation defects, changed duty cycle, missing suspensions, or copied vendor assumptions can all make beta misleading.
For PM review, beta near or below 1 should trigger caution around age-based replacement. It may point toward condition monitoring, design correction, process correction, or run-to-failure for noncritical items, but the consequence review controls.
R(t) = exp(-(t / eta)^beta)
MTBF = eta x Gamma(1 + 1 / beta)
These equations do not validate the input beta, failure data, or maintenance decision.
Collecting the Right Failure Data
The quality of a Weibull review depends on the quality of the failure data. You need time-to-failure records, operating exposure, and suspension records for components replaced preventively or still in service. Ignoring suspensions can bias life estimates.
The data must be for a single failure mode. Mixing failure modes (e.g., bearing fatigue failures and seal leakage failures on the same pump) produces a meaningless Weibull fit. If your CMMS work orders describe failures generically ("pump failed," "replaced bearing"), you need to go back and classify them by root cause before running the analysis.
Data quantity alone is not enough. You still need fit method, confidence intervals, goodness-of-fit, changed operating conditions, PM history, rebuild quality, and outlier handling. For regulated fleets or safety-critical equipment, records and inspection requirements can control regardless of the cost curve.
Data sources can include CMMS work orders, maintenance logs, equipment history files, operator logs, condition-monitoring trends, inspection records, and OEM service information. ISO 14224-style data discipline can help organize the review, but exact applicability depends on the sector and asset.
1. Separate failure modes.
2. Record exposure in the unit that matches duty cycle.
3. Include suspensions and survivors.
4. Keep installation, rebuild, and operating-condition changes visible.
5. Do not let a screen replace a qualified Weibull fit.
The Age Replacement Model
The age replacement model connects an entered reliability distribution to a cost-rate prompt. It calculates expected cost per unit of operating time as a function of the PM interval. The formula used by the screen is:
C(t) = [Cp * R(t) + Cf * (1 - R(t))] / integral of R(x) from 0 to t
Where Cp is planned PM cost, Cf is unplanned failure cost, R(t) is survival probability, and the denominator is the expected cycle length. A lower cost point can be useful for screening, but it is not automatically a practical or acceptable interval.
The cost ratio drives the curve. A high failure cost can move the local cost point shorter; a high planned PM cost can move it longer. That does not account for safety, environmental consequences, production windows, spare-parts limits, warranty, legal inspection intervals, or maintenance-induced failure.
C(t) = [C_pm × R(t) + C_fail × F(t)] / E[cycle length]
Use the minimum as a review prompt only.
Practical Review After the Screen
After a screen produces a local cost point, review the inputs and consequences before any change. Check the failure mode, data source, censored records, confidence range, OEM limits, warranty, regulatory requirements, inspection obligations, safe-work controls, spare-parts availability, and labor window.
If the local point is longer than the current PM, that may indicate over-maintenance, bad beta input, missing failure cost, or a flat curve. If it is shorter, that may indicate high consequence cost, a strongly age-related failure mode, or a data problem. Both directions require review.
For safety-critical, environmentally regulated, production-critical, warranty-controlled, fleet-regulated, or insured assets, interval changes should go through reliability engineering and the site change-management process. Pilot changes need monitoring criteria and rollback triggers.