Component Mangagement for Critical Parts in Gas Turbine EnginesJeff Stillinger
Rolls-Royce Corp.
Life determination management starts with an FAA regulation and mandate for establishing a life limit for flying machinery. There are really two components to this mandate. 1) Determination of expected component life and 2) management of the components during their life.
When one studies the cycle of an aircraft engine, one finds that the maximum stresses are deveoped during takeoff, with a lessor high value during reverse thrust at landing. For this reason, it is generally accepted that 1 flight is equivalent to one cycle in the low cycle fatigue scenario. Actually during the takeoff period as much as 609,000 lb or radial forces can be shown to develop on a disk. This is equivalent to about 139 ksi of stress in a fillet area.
A particular challenge is how to actually determine component life. We can and do perform cyclic spin tests, where part coupons are subjected to thousands of cycles and inspected repeatedly until some sign of a crack is detected. Based on the cycle number at which the crack initiates we develop a probablity of failure assuming that our test part is very ideal and unlikely to fail. Thus if we were to conider a bell curve, we would place our test cycle count very high in the upper tail of the curve. Because we want very low probability of failure in the field, we debit off a large number of cycles to find a count in the fractional percentages of probabliliy of the low tail of our bell curve. This is the value we assign as the "safe" life of the part.
For Component Life Management we carefully track what is flying. A disposition is required for all parts with minor damage. Life values are reviewed against experience. In-flight digital data is analyzed and reviewed, and non-destructive testing is performed a frequent mandated intervals.
Recently we have been applying the dicipline of probabalistic risk assessment. Computer models of parts are analized for stresses at all locations using various modeling techniques. The parts are then divided into zones of stress level. Where stress levels are low, zones tend to be large; and where maximum stresses are encountered, zones tend to be small. Then some uniform scattering of initiation points (likeky crack starters) are distributed on the entire model. It stands to reason that there is low probablility that one of these points will coincide with a small (max/ stress zone), and a much larger probability that they will lie in the large (low stress zones). Monte-Carlo simulation can be applied to these models to determine the liklihood (probablity) of a crack initiating. These simulations are often very large and run overnight on the computers performing them.
In summary, Life determination leads top a "safe life" value and is achieved by methods, testing, and analysis. Life Management leads to "safe life verification" and is achieved through observation, measurement, and control (NDT).