“We cannot do reliability analysis. We do not have the data.” We hear this constantly, especially from greenfield projects: new plant, novel process, nothing has failed yet because nothing has run yet. The premise is wrong, and the projects that accept it forfeit their best planning window.
Failure records are one source of reliability information. They are not the only one:
- Engineering judgment is data. A senior fitter who has watched a particular bearing type die a hundred times is carrying a Weibull distribution around in his head. He just does not call it that. Structured elicitation, asking when the earliest credible failure is, when half would be gone, whether failures cluster early or late, converts that experience into parameters.
- Vendor and industry information is data. Test results, generic failure rate handbooks, and curated reliability parts libraries provide defensible starting estimates for common equipment classes.
- The physics is data. Knowing the dominant failure mechanism constrains the model before any statistics arrive. Our white paper Cataloguing Failure Mechanisms explains why the universal catalogue of acceleration models is elusive, and what disciplined engineers do instead.
The Shape Parameter Encodes the Mechanism
The Weibull distribution is the workhorse of this approach because its shape parameter encodes the failure regime itself. A shape parameter below one means infant mortality, with manufacturing defects and installation errors dominating early life. Equal to one means random failure, external events striking without regard to age, which is the only regime where a constant failure rate and the much-abused MTBF actually apply; the podcast episode Why the MTBF unpacks the damage done when that assumption is applied everywhere. Above one means wear-out, with fatigue, corrosion and erosion accumulating over time.
Even without a single site failure record, an experienced engineer can usually say which regime applies to a component and roughly when failures become credible. That is enough for a defensible starting model. Then, as operating data arrives, the estimates tighten. This is not guessing. It is structured elicitation followed by disciplined refinement, and it answers the question our white paper How Much Do You Need to Know? poses directly: enough to make the decision in front of you, which is often far less than perfection.
How the Tools Help You Discover the Benefit
Availability Workbench and Reliability Workbench accept failure models from any source on equal terms: fitted from data in the Weibull module where records exist, entered from judgment or vendor information where they do not, with exponential, Weibull and other distributions available throughout. The same simulation and RCMCost optimisation machinery runs either way. A greenfield project can therefore evaluate maintenance strategies, spares holdings and design alternatives well before commissioning, and refine everything once the plant starts writing its own history.
The deeper benefit: reliability modelling is not a reward for having good data. It is the framework that tells you which data to start collecting on day one. Our white paper Backward Blueprinting formalises this: design the reliability plan by working backward from the decision you will need to make.
Go Deeper
- Software: AWB Weibull Module, Reliability Parts Libraries, Reliability Prediction
- Services: Weibull Analysis / Survival Analysis, Reliability Engineering
- White papers: How Much Do You Need to Know?, Cataloguing Failure Mechanisms, Backward Blueprinting
- Podcasts: Why the MTBF, Beyond the Curve: Remaining Useful Life
- Next in the series: Suspended Data and Asset Criticality
How can we help? Talk to us at contact@mantua.group.