The Six Failure Patterns: Nowlan and Heap Explained

How We Evaluated

This guide is independent editorial analysis based on publicly available government documentation. Reliable does not sell reliability software, consulting, or certification and has no commercial interest in routing readers toward any particular method. Reliable does not accept payment for inclusion in this guide. Vendors may sponsor enhanced listings with additional detail, but editorial rankings are independent. Read our editorial policy.

The facts below come from F. Stanley Nowlan and Howard F. Heap, Reliability-Centered Maintenance, United Airlines, sponsored by the Office of the Assistant Secretary of Defense, December 29, 1978 (AD-A066579), Exhibit 2.13 and the surrounding Chapter 2 discussion of age-reliability characteristics, plus Chapter 3 on the four scheduled-task types and their applicability and effectiveness criteria, and NASA’s Reliability-Centered Maintenance Guide for Facilities and Collateral Equipment (FINAL, September 2008), sections 2.1, 2.2, and 4.1.5.

What the six patterns are

The six patterns describe how the conditional probability of failure changes with age. Conditional probability of failure, in Nowlan’s terms, is the probability that an item fails during the next interval given that it has survived up to the start of that interval. It is not the same as the total number of failures over a fleet. It is a running measure of risk for an item that is still in service.

Nowlan and Heap did not set out to publish a taxonomy of curves. United Airlines built these curves while it was extending overhaul times and trying to understand whether older equipment actually failed more often. When the analysts plotted conditional probability of failure against age since manufacture, overhaul, or repair, the curves fell into six basic patterns, shown together in Exhibit 2.13.

The percentages Nowlan and Heap attach to each pattern are the percentage of United Airlines items that fell into that pattern. NASA identifies that dataset as UAL 1968. Exhibit 2.13 says the curves came from United Airlines reliability analyses conducted over a number of years. The value of the exhibit is not the exact split. It is the demonstration that a bathtub is only one of six shapes, and not the dominant one.

The six patterns, A through F

The letters below are Nowlan’s letters from Exhibit 2.13. NASA later relabeled the same curves in a different order, so if you compare this guide to the NASA document, read the letters carefully.

Pattern A (the bathtub)

Pattern A is the curve most people picture when they think about wear. Nowlan and Heap note that it is often referred to in the reliability literature as the bathtub curve. It has three regions: an infant-mortality region with a high probability of failure right after manufacture or overhaul, a long middle region of constant and relatively low failure probability, and a wearout region at the end where the probability of failure begins to climb rapidly with age.

Because Pattern A has a genuine wearout region, an age limit can be worthwhile. Nowlan and Heap put a condition on it: an age limit may be desirable provided a large number of units survive to the age at which wearout begins. If the pattern truly fits, acting just before the item enters the wearout zone is justified, because the overall failure rate will be reduced.

The catch is how rarely the pattern showed up. Just 4 percent of the items fell into Pattern A. Nowlan and Heap note that it is often assumed that the bathtub curve is representative of most items. In the United Airlines items they analyzed, it was not.

Pattern B

Pattern B shows a constant or gradually increasing failure probability, followed by a pronounced wearout region at the end. Like Pattern A, it ends in wearout, so an age limit may again be desirable. Nowlan and Heap describe this curve as characteristic of aircraft reciprocating engines.

Patterns A and B are the two curves that show clear wearout. Nowlan and Heap report that only 6 percent of the items studied showed pronounced wearout characteristics across both curves combined.

That 6 percent matters for how you think about overhaul. These are the items where a maximum operating age has a defensible mechanical basis, because failure really does concentrate near a certain age. For everything else, the case for a fixed age limit gets much weaker.

Pattern C

Pattern C shows a gradually increasing failure probability with no identifiable wearout age. The risk keeps creeping up, but there is no point where failures suddenly cluster. Nowlan and Heap describe this curve as characteristic of aircraft turbine engines.

Because there is no well-defined wearout zone, it is usually not desirable to impose an age limit for a Pattern C item. The item is steadily more likely to fail as it ages, but there is no age at which a rework or discard clearly pays off.

Another 5 percent of items had this shape. Nowlan and Heap allow one exception: for a very few of these items an age limit might prove useful, provided it was cost-effective. That is a narrow door, not a general rule.

Pattern D

Pattern D shows a low failure probability when the item is new or just out of the shop, followed by a quick increase to a constant level. After that early rise, the risk flattens and stays flat.

The important feature is the plateau. Once the item passes its early life, its probability of failure does not keep climbing with age. There is no wearout region to aim an age limit at.

Pattern D is one of the three curves with no wearout zone. For items shaped like this, a scheduled age-based overhaul has no age to be scheduled against.

Pattern E

Pattern E shows a constant probability of failure at all ages. This is the exponential survival distribution, where a brand-new item and a long-serving item carry the same risk of failing in the next interval.

Under Pattern E, age tells you nothing about failure risk. An item that has run for years is no more likely to fail next week than one installed last month.

That makes an age limit pointless. Replacing a Pattern E item on a schedule swaps a known-good unit for a fresh one with exactly the same failure probability. Pattern E has no infant-mortality region. Pattern A also begins with infant mortality, and Pattern F is infant mortality followed by a constant or very slowly increasing failure probability. The separate Nowlan and Heap point is that intrusive overhaul can introduce infant mortality into an otherwise stable complex system.

Pattern F

Pattern F shows a high infant-mortality rate, followed by a constant or very slowly increasing failure probability for the rest of the item’s life. Nowlan and Heap note that this pattern is particularly applicable to electronic equipment.

The failures that do occur cluster at the beginning, not the end. Once an item survives its early life, its risk settles and stays low.

Patterns D, E, and F are the three curves with no wearout zone. Together they account for the bulk of the population. Some 89 percent of the items analyzed had no wearout zone, and after a certain age the conditional probability of failure continued at a constant rate across curves D, E, and F.

How the patterns change PM task selection

Nowlan and Heap describe four types of scheduled-maintenance task: inspect an item to detect a potential failure, rework an item before a maximum permissible age is exceeded, discard an item before a maximum permissible age is exceeded, and inspect an item to find failures that have already occurred but were not evident to the operating crew. The six patterns primarily screen whether an age-limit task is applicable. On-condition inspection and failure-finding have separate applicability tests in Chapter 3. They are not chosen off the curve alone.

Scheduled rework, and discard as an economic-life limit, only make sense when there is pronounced wearout to get ahead of. That points to Patterns A and B. For Pattern C, an age limit is usually not desirable, and is defensible only in the rare case where it turns out to be cost-effective. For Patterns D, E, and F, there is no wearout zone, so those age limits cannot improve reliability. Safe-life discard is a different test. Nowlan and Heap say a safe-life discard task is applicable only when the item is subject to a critical failure and test data show that no failures are expected below the specified life limit. That is not the same as rework or economic-life discard on a wearout curve.

This is where scheduled overhaul can backfire. Because most complex items follow curves C through F and show no concentration of failures tied to operating age, an age limit does little or nothing to improve their overall reliability. In many cases scheduled overhaul actually increases the overall failure rate, by introducing a high infant-mortality rate into a system that was otherwise stable. You take a running item, tear it down, and hand it back its early-life risk for no wearout benefit.

None of this rules out interval-based maintenance everywhere. NASA’s guide, in section 2.2, notes that interval-based maintenance is still appropriate where abrasive, erosive, or corrosive wear takes place, where material properties change due to fatigue or embrittlement, or where a clear correlation exists between age and functional reliability. The point is to select the task from evidence about the failure pattern, not from the assumption that age drives failure. When the evidence points to condition rather than age, see How to Start a PdM Program and Condition-Based Maintenance vs Predictive Maintenance.

A worked plant example

The following is illustrative and does not describe a real plant. Picture a mid-size facility, call it Northfield Processing, deciding how to maintain two items on the same boiler feedwater pump.

The first item is a simple consumable: a rubber coupling insert between the motor and the pump. It is a single-celled part with a straightforward wear mechanism. It flexes, it hardens, it cracks, and it tends to fail near a fairly consistent service life. A part like this can plausibly follow Pattern A or B, with a real wearout region. For an item like this, a discard-before-max-age task can be effective, because a large number of inserts survive to the age where wearout begins, and replacing them just before that point reduces the failure rate.

The second item is the pump assembly itself: bearings, seals, wear rings, impeller, and shaft, all interacting. It is a complex item, and complex items in the Nowlan and Heap study mostly followed curves C through F, with no failure concentration tied to operating age. If Northfield’s pump behaves like a D, E, or F item, a scheduled teardown at a fixed age does not target any wearout zone. Worse, the teardown reintroduces infant-mortality risk from reassembly, so the overhaul can raise the failure rate rather than lower it.

The decision, then, is not which pump to buy. The curve only screens whether age-related applicability is even on the table. For scheduled rework, Nowlan and Heap also require a large proportion of units to survive to the relevant age, and that original failure resistance can be restored. Effectiveness is a separate test. The coupling insert can support an age limit on that basis. The pump assembly does not, on this pattern. Same pump, two different items, and only one of them has a wearout region an age limit could aim at.

Honest Limitations

Reliability-Centered Maintenance by Nowlan and Heap is a 1978 report, sponsored by the U.S. Department of Defense and written by authors at United Airlines. It grew out of commercial aviation and military logistics. The 4, 6, 5, and 89 percent figures are the percentage of United Airlines items in Exhibit 2.13, identified by NASA as the UAL 1968 dataset, not a plant-wide average for any industry in 2026, and not a number you can drop into your own asset register without analysis.

Later work found similar shapes but different splits. NASA’s guide summarizes three studies, the original United Airlines work plus follow-on studies in Sweden in 1973 and by the U.S. Navy in 1982, and reports that across the three, random failures run between 77 and 92 percent of total failures, with age-related characteristics accounting for the remaining 8 to 23 percent. That range is NASA summarizing those three studies, not a universal constant.

Two more cautions. NASA letter-codes the six curves in a different order than Nowlan’s Exhibit 2.13; in NASA’s scheme the bathtub is Type E, while in Nowlan’s it is Pattern A, and this guide uses Nowlan’s letters throughout. And the six patterns are a starting point, not a decision engine. They tell you whether an age limit could help, but they do not replace the full RCM logic for choosing and justifying each task.

Frequently Asked Questions

What are the six Nowlan and Heap failure patterns?
They are the six conditional-probability-of-failure curves in Exhibit 2.13 of Nowlan and Heap’s 1978 report. Each plots the probability of failure in the next interval against age. Pattern A is the bathtub curve. Pattern B shows a constant or gradually increasing failure probability, followed by a pronounced wearout region. Pattern C shows a gradually increasing failure probability with no identifiable wearout age. Pattern D shows a low failure probability when new, then a quick increase to a constant level. Pattern E is a constant probability of failure at all ages. Pattern F shows infant mortality, followed by a constant or very slowly increasing failure probability.

Is the bathtub curve the most common failure pattern?
No. The bathtub curve is Pattern A, and in the United Airlines items Nowlan and Heap studied, just 4 percent of items followed it. The assumption that most equipment follows a bathtub shape is the specific belief their data contradicted.

Do most assets wear out with age?
Not in that study. Only 6 percent of items showed pronounced wearout across Patterns A and B combined, and some 89 percent had no wearout zone at all. For those items, performance could not be improved by imposing an age limit. Patterns D, E, and F lack a wearout zone, and after a certain age the conditional probability of failure continues at a constant rate. Pattern D does climb quickly early, then levels off.

How do the six patterns change preventive maintenance task selection?
Nowlan and Heap define four scheduled-task types: inspect for a potential failure, rework before a maximum age, discard before a maximum age, and inspect for hidden failures that already occurred. Scheduled rework and economic-life discard only help where there is real wearout, which points to Patterns A and B, and rarely C. Safe-life discard is separate: it requires a critical failure and test data showing no failures are expected below the specified life limit. For Patterns D, E, and F there is no wearout zone, so the curves rule against rework and economic-life discard. They do not, by themselves, select inspection. On-condition inspection still needs a detectable potential-failure condition and a usable interval. Depending on consequences and detectability, the RCM logic may select failure-finding, no scheduled maintenance, or redesign.

Are the Nowlan and Heap percentages industry averages?
No. The 4, 6, 5, and 89 percent figures describe the United Airlines items in Exhibit 2.13. NASA identifies that dataset as UAL 1968, from reliability analyses conducted over a number of years. NASA later summarized three studies and reported random failures between 77 and 92 percent with age-related characteristics for the remaining 8 to 23 percent, but even that is a range across three specific studies, not a plant-wide average for your equipment.

Why can scheduled overhaul make a complex item less reliable?
Most complex items showed no concentration of failures tied to operating age, so an age-based overhaul has no wearout zone to target. Tearing the item down and reassembling it reintroduces early-life, infant-mortality risk. For an otherwise stable system, that can raise the overall failure rate rather than lower it.

Related Guides

• How to Perform RCM: Reliability-Centered Maintenance Methodology
• RCM vs FMEA: What’s the Difference, and Which One Do You Need?
• How to Start a PdM Program
• Condition-Based Maintenance vs Predictive Maintenance
• How to Calculate Asset Criticality

Sources

• United Airlines / U.S. Department of Defense – Reliability-Centered Maintenance (Nowlan and Heap, 1978, AD-A066579; Exhibit 2.13 six patterns)
• NASA – Reliability-Centered Maintenance Guide for Facilities and Collateral Equipment (September 2008; age-reliability restatement)