Fast forward 18+ years from Waddington’s research with Coastal Command to where Mr.Nowlan and Mr. Heap used years of United Airlines’ maintenance data to figure out how to reduce maintenance costs in their 500+ page paper, “Reliability Centered Maintenance” (which can be downloaded from the National Technical Reports Library).
In chapter 2, “The Nature of Failure,” they described how equipment failure has six profiles based on their graphic representation.
- Bathtub. A significant proportion of parts will fail early, most will ‘go the distance’ and then they too will fail.
- Lazy J. No early mortality, parts will fail after they reach as certain age.
- Slope. No concentration of failure, just a gradual increase with age.
- Face plant J. Unlikely to fail early in life, followed by a concentration in rate, followed by a long period of reliability.
- Flat line. Entirely even probability of failure across entire lifespan.
- SIDS. Most prone to failure early on, but survivors enjoy a long lifespan.
When Age Doesn’t Matter
The paper contains an embarrassment of insight and this is just one of their many findings:
Some 89 percent of the items analyzed had no wearout zone; therefore their performance could not be improved by the imposition of an age limit… The presence of a well-defined wearout region is far from universal; indeed, of the six curves in Exhibit 2.13, only A and B show wearout characteristics. It happens, however, that these two curves are associated with a great many single-celled or simple items – in the case of aircraft, such items as tires, reciprocating-engine cylinders, brake pads, turbine-engine compressor blades, and all parts of the airplane structure… Moreover, most complex items had conditional-probability curves represented by curves C to F- that is, they showed no concentration of failures directly related to operating age.
This means that 89% of the items would not benefit from a scheduled maintenance program based on age, cycle counts, or operating hours. When critical equipment failure can’t be forecasted and the stakes are high, your design must contain fail-over, fail-safe and redundancy or alternate provision of whatever function the equipment is providing. But when the consequences are only detrimental to value generation and not loss of life or reputation, then recurring monitoring could be the answer.
Continuous Asset Monitoring
Predicting equipment failure when age is not a factor now relies on monitoring to provide early warning. The consequences of monitoring is to create a stream of information that if sampled appropriately can be combined with statistical models to detect the onset of failure. This is the new economy of predictive maintenance where low cost sensors, ubiquitous telemetry and commoditized machine learning create new opportunities for predictive maintenance and Microsoft’s IoT services has much to offer the predictive maintenance economy.