The six failure patterns are the six conditional-probability-of-failure curves identified by Stanley Nowlan and Howard Heap in their 1978 reliability study for United Airlines, showing that the odds of a given asset failing do not follow one universal "wear-out" shape. Their work, funded by the US Department of Defense and later foundational to Reliability Centered Maintenance (RCM), overturned a comfortable assumption: that most equipment ages predictably and can be protected by scheduled overhauls. The data said otherwise. For maintenance planners, this is the single most important reason to move spend away from blind calendar-based rebuilds and toward condition-based work.
Nowlan and Heap plotted, for large populations of aircraft components, how the conditional probability of failure changed as a function of operating age. Conditional probability is the chance an item fails in the next interval given that it has survived to now. They found that these curves clustered into six distinct shapes, and, critically, that only three of them show failure risk rising with age. The other three are flat or fall over time, meaning older parts are no more likely to fail than younger ones.
This finding is closely related to, but broader than, the classic bathtub curve, which is only one of the six patterns. The bathtub is intuitive and still taught everywhere, yet it described a minority of the components studied. Understanding all six requires a working grasp of failure statistics; the Weibull distribution is the usual tool for fitting these curves to real data.
Add the three age-related patterns (A, B, and C) and you reach roughly 11 percent. The remaining 89 percent (D, E, and F) are dominated by random or early-life failures where a scheduled overhaul does nothing useful and can even introduce fresh infant-mortality defects.
Imagine a fleet of 200 identical gearbox bearings that follow Pattern F. In the first 500 operating hours, infant-mortality defects (bad installs, contamination) cause 12 failures. After that break-in, the population settles to a low random rate of about 1 failure per 1,000 bearing-hours.
A planner sets a "safe" 4,000-hour scheduled replacement. Because the bearings do not wear out, every one of those swaps replaces a healthy bearing with a fresh one, and each fresh install re-enters the 500-hour infant-mortality zone. If even 6 percent of new installs carry a defect, you have deliberately manufactured a new wave of early failures on a component that was running fine. The calendar PM did not reduce risk; it recycled it. A condition-based maintenance approach that watches vibration and temperature would leave the healthy bearings alone and act only on the ones actually degrading.
The practical takeaway is not "stop doing preventive maintenance." It is match the tactic to the pattern. Time-based replacement earns its keep only on components that genuinely show a wear-out age (patterns A, B, and C, and even then only when there is a clear, dominant wear-out point). For the rest, the RCM answer is on-condition tasks, failure-finding checks, redesign, or a deliberate run-to-failure decision where consequences are tolerable.
This is also where the broader debate of reactive versus proactive maintenance gets sharper: proactive does not have to mean calendar-driven, and often should not.
The uncomfortable reason many plants default to calendar PM is that they lack the failure history to plot these curves at all. Classifying a component as random versus wear-out requires time-stamped failure records, run hours, and condition readings. Without that, planners fall back on the safest-sounding schedule. Reliable pattern analysis depends on a disciplined CMMS that captures every work order with its failure mode, and on real-time condition signals rather than manual rounds. Cleaner data also directly improves your overall equipment effectiveness, because unplanned stops shrink when maintenance targets the assets that are actually degrading.
Applying the six failure patterns is only as good as the data underneath it, and that is the layer Fabrico provides. Fabrico is a real-time OEE and production monitoring platform paired with a field-ready CMMS, giving you the time-stamped failure history, run hours, and work-order detail needed to classify each asset's real failure behavior instead of guessing. Its computer-vision monitoring reads machine state even on older equipment with no PLC, so condition signals are available on the exact legacy assets that most often follow random or infant-mortality patterns. Work orders, asset registers, preventive scheduling, and spare-parts tracking live in one place, and Fabrico is EU-built with EU data residency. Fabrico is the real-time data foundation that lets a reliability team decide, with evidence, which assets deserve a schedule and which deserve to be watched.
No. It means scheduled hard-time replacement is only effective for the roughly 11 percent of components whose failure risk genuinely rises with age. For those assets it is exactly the right tactic. The point is that applying a calendar schedule to the other 89 percent wastes labor and parts and can introduce new infant-mortality failures, so the tactic must be matched to each component's actual pattern.
The bathtub curve is just one of the six patterns (Pattern A). Nowlan and Heap showed it applied to only a small share of the components they studied. Treating the bathtub as universal leads planners to expect a wear-out zone that most equipment never reaches, which is precisely the assumption their research corrected.
You need per-asset failure history with dates and run hours, the failure mode for each event, and ideally condition data such as vibration or temperature trends. With enough records you can fit a Weibull curve per failure mode and see which of the six patterns dominates. Most plants find the gating problem is data capture discipline, not the statistics themselves.
Ready to stop scheduling overhauls on assets that never wear out? See how Fabrico turns real-time machine data and structured work-order history into the evidence base for smarter maintenance decisions. Book a Fabrico demo and put your failure data to work.
Login
Blog
