Maintainability Prediction Revisited: Diagnostic Behavior, System Complexity, and Repair Time

If repair time for complex svstems and equipment could be related to simple overriding equipment characteristics, then prediction of repair time could be used as an equipment design tool. However, available prediction techniques have beeii shown to be inadequate for new technology. Repair time data from several thousand repairs of various types of electronics equipment, when plotted on Weibull cumulative probability paper, showed remarkablv similar characteristics, including a large skew toward longer repair times for data taken under field conditions. It was hypothesized that the skewed tail of the distribution wvas related to the way in which the maintenance technician interacted with the equipment and the environment in the diagnostic process. A three-parameter interaction model wvas developed from which the tail of the distribution and its expected value could be calculated. Measured active repair time distributions for 11 ground-based and one airborne electronic equipment wvere then compared with a family of curses computed from the model; "best-visual-fit" estimates of the three parameters were found to fall within fairly narrow ranges. The were then substituted in the expected value formula to obtain estimates of mean time to repair for the 12 equipments. While these computed estimates were found to correspond closely (r=0.98) with the actual measured values in both field and laboratorv ensvironimlenits, the validit) of the interaction model and its underlying assumptions needs to be tested further.