Reliability Engineering: From Gut Feel to Failure Data
Turn recurring equipment problems into evidence-based improvements with bad-actor analysis, life data, reliability metrics and better failure records.
Every plant has a few assets that fail repeatedly while everyone says "that's just how that machine is." Reliability engineering is the discipline of replacing that shrug with evidence: measuring how equipment fails, understanding why, and engineering the improvement.
This is the third article in our Reliability Engineering for Maintenance series, following reliability centered maintenance. RCM sets the strategy; reliability engineering supplies the data and analysis that keep that strategy honest.
What a reliability engineer actually does
The role isn't a senior planner or a firefighter. It is applying engineering methods to predict, measure, and improve how long equipment performs its function. In practice that means:
- Finding bad actors, the small number of assets or failure modes causing most of the lost production or cost
- Analyzing failure data to understand how equipment fails over its life
- Recommending design, operating, or maintenance changes and proving that they worked
- Supporting strategy decisions with life-cycle cost and risk evidence
Start with bad actors
You don't need sophisticated software to begin. Pull 12 to 24 months of work-order history, group the results by asset and failure mode, and rank by downtime, cost, or safety impact. A Pareto chart will usually show that a small share of assets drives a large share of the pain.
Those bad actors are where analysis effort pays back fastest, and they feed directly into root cause analysis and RCM.
Understanding how equipment fails
Failures don't all behave alike. Weibull analysis is a common tool for life data. Its shape parameter gives a quick read on failure behavior:
- Below 1: failure rate falls with time, typical of infant mortality from installation or manufacturing defects
- Around 1: failures are essentially random
- Above 1: failure rate rises with age, typical of wear-out
This matters because the right response differs. Infant mortality calls for better installation and commissioning. Random failures call for condition monitoring. Wear-out may justify scheduled replacement. Applying the wrong one wastes money.
The metrics, and their limits
- MTBF (mean time between failures) summarizes the average time between failures of repairable equipment. It is not a life expectancy, and an average can hide a bimodal pattern of early and late failures.
- MTTR (mean time to repair) reflects how long recovery takes, including diagnosis, parts, and planning, not just wrench time.
- Availability combines both ideas: the proportion of time the asset is able to perform when needed.
Use them to compare and track, but look at the underlying failure distribution before drawing conclusions.
Your data is the limiting factor
Reliability engineering lives or dies on data quality. Common problems include:
- Failure codes that are vague, inconsistent, or left as "other"
- Work orders that record the repair but not the failure mode
- Missing run-time or operating context
- Corrective work mixed with routine tasks and project work
A failure reporting and corrective action process (FRACAS) closes this gap by capturing what failed, how, why, and what was done, so each failure improves the next decision. If your CMMS data can't answer "how did this fail and what did we do?", fixing that is the best first project.
How KANE can help your plant
We help plants build the evidence base for better decisions, including:
- Bad-actor analysis using your work-order history
- Failure data and CMMS code reviews so your records support analysis
- Life-data, availability, and life-cycle cost analysis on key assets
- A practical reliability roadmap that connects findings to action
Training on reliability engineering
KANE also provides training on reliability engineering fundamentals for your maintenance, reliability, and operations teams, from reading failure data to using it to prioritize improvements. To talk about training for your site, contact us at [email protected] or (657) 478-8727.
Next in the series: Root Cause Analysis. Uptime® and Uptime® Elements™ are trademarks of their respective owners.