Maintenance strategy selector

Why the questions come in this order

The tool follows the RCM decision diagram in Nowlan and Heap's report for the US Department of Defense (Exhibit 4.4). Consequences come first, because they decide what "worth doing" means:

• Safety or environment: a task has to bring the risk to a level the site accepts, whatever it costs. If none does, redesign is required.
• Operational: a task has to cost less than the repairs plus the lost output it prevents.
• Non-operational: a task has to cost less than the repairs alone.
• Hidden: a task has to give the protective function the availability you need. The default is a failure-finding test.

Within each branch a condition check is tried first, because it replaces a part only when the part needs it; Nowlan and Heap call it the most desirable task whenever it applies. Environmental and legal consequences sit with safety, as in later RCM standards such as MIL-STD-3034A and the NASA RCM guide.

The six failure patterns

United Airlines plotted the chance of failure against age for its aircraft components and found six shapes. Nowlan and Heap published them in 1978 (Exhibit 2.13), with the share of the items studied in each:

• A. Bathtub, 4%High when new, then low and steady, then a wear-out zone.
• B. Wear-out, 2%Steady or slowly rising, then a pronounced wear-out zone.
• C. Slowly rising, 5%Rises gradually with age, with no clear wear-out age.
• D. Low, then steady, 7%Low when new or just repaired, then rises quickly to a steady level.
• E. Random, 14%The same chance of failure at every age.
• F. Early life, 68%High when new or just repaired, then steady or very slowly rising.

In the report's words, some 89 percent of the items had no wear-out zone, so an age limit could not improve them; 11 percent (A, B and C) might benefit from one. The NASA RCM guide (2008) adds a Swedish study (1973) and a US Navy study (1982) and puts random failures at 77 to 92 percent across the three.

These are aircraft and naval items, not plant equipment, and we found no comparable published study for plants that we could check. What carries over is the test: replacing on a schedule helps only a failure mode with a clear wear-out age that most units reach. Early-life failures (pattern F) point to installation, start-up and repair quality, and the NASA guide notes that a scheduled overhaul often adds more of them.

Axes: chance of failure (vertical) against age since new or since repair. Some later texts letter the patterns in a different order; these are Nowlan and Heap's.

The P-F interval and the check interval

Nowlan and Heap define a potential failure as "an identifiable physical condition which indicates a functional failure is imminent". A condition check works only if that condition can be detected, the time from it to the failure is reasonably consistent, and there is time to act.

• Check at no more than half the P-F interval (ABS, 4.1). With a 6-week warning, check every 3 weeks.
• Leave time to act. In the worst case the sign appears just after a check, so the time left is the P-F interval minus the check interval. If that is shorter than the time you need to plan and do the repair, shorten the interval. If the warning is not longer than the time to act, no check works.
• Shorten it further for higher-risk failure modes or when the P-F estimate is a guess (ABS, 4.1). Once a reading passes its alert limit, the NASA guide says to cut the monitoring interval to between a third and a quarter of what it was.
• Learn the real P-F interval. ABS gives the example of pumps whose weekly vibration readings caught defects that then ran 6 to 8 weeks before repair without failing: the P-F interval is at least 6 weeks, so the check can move to every 3 weeks.

The failure-finding interval

A protective device that fails at random and is tested every T is failed, on average, for about T ÷ (2 × MTBF) of the time. Set that equal to the unavailability you can accept and solve for T: T = 2 × (1 − A) × MTBF. If you know how often the device is called on (MTED) and how rarely you can accept the multiple failure (MMF), the unavailability is MTED ÷ MMF.

The unavailability you accept is your site's risk rule, not a standard. ABS lists the assumptions: random failures, failure rate × interval under 0.1, test and repair times short, and the multiple failure possible only from that one demand. The test itself must not create the hazard, and should prove the whole protective function, not one part of it.

What this tool does not do

• It does not define functions or list failure modes. An RCM analysis does that with the people who run and maintain the asset; the RCM worksheet holds the whole study.
• It does not decide what is safe. Whether a failure could hurt someone, and what risk the site accepts, are decisions for the site and its safety professionals.
• It does not override a law, code or insurer. Where one sets an inspection or test interval, as is common for pressure relief devices, that interval applies whatever this tool says.
• It does not supply failure data. Every warning time, MTBF and cost is yours; the tool only shows what follows from them.
• No claim is made that it meets SAE JA1011, the standard that sets what a process must include to be called RCM.

Sources and how this tool was checked

• F. Stanley Nowlan and Howard F. Heap, Reliability-Centered Maintenance (US Department of Defense, 1978, report AD-A066579): potential failure (section 2.1), the six patterns (section 2.8, Exhibit 2.13), the task criteria (chapter 3), the decision diagram and default answers (Exhibits 4.4 and 4.5).
• American Bureau of Shipping, Guidance Notes on Reliability-Centered Maintenance (2004, updated 2018): section 4 for the P-F and check intervals, section 5 for the failure-finding interval and Table 2.
• NASA, Reliability-Centered Maintenance Guide for Facilities and Collateral Equipment (2008): chapter 3 for the four outcomes, section 4.1.5 for the three failure-pattern studies, page 4-3 for shortening the interval after an alert.
• John Moubray, Reliability-centred Maintenance (RCM II), 2nd ed. (1997), the usual reference for the P-F interval and both interval rules, is not free to read and was not used to check this page.

Unit tests walk every combination of answers to make sure each ends in a recommendation, check the paths above against the decision diagram and the default answers, and check the formulas against hand-worked cases: ABS's 6-week pump example, its Table 2, and the RCM worksheet's relief valve (15-year MTBF, a demand every 8 years, one multiple failure in 2,000 years tolerated: 43.8 days).