|M.Sc Student||Migdali Refael|
|Subject||Reliability-Oriented Periodic Inspection Interval for|
|Department||Department of Quality Assurance and Reliability||Supervisor||Dr. Yefim Haim Michlin|
In the course of a review of the maintenance policy of a helicopter operating organization ("the user") for a certain type of helicopter, it was found necessary to determine whether the frequency of one of the scheduled tests could be reduced without a significantly increased risk of safety mishaps.
A specific model was developed to predict the effect of test interval change on failures expectation.
To this end, field data had to be compiled on failures observed during the test under the current and proposed policies respectively. These data would indicate whether the change would affect the failure rates or MTBF’s. The experiment was to be based on the accepted rules of experiment planning and hypothesis checks, with a view to yielding definite data on the probabilities of 1st- and 2nd- kind errors, and the relative significant difference(RSD).
The expected failure rates and the maintenance frequency being low, and the helicopter complement relatively small, the failure samples had to be as small as possible.
Assuming that the failure rates are constant and the times between failures (TBF)distribution exponential, the sample size necessary for bringing out the difference in the MTBF, and the decision limits, were determined for the fixed-size case. A sequential-testing procedure was worked out for the experiment in question, with a view to reducing its duration. In it, two groups of helicopters, subjected to the respective maintenance systems, are observed. After each failure the probability is determined of the errors of the two kinds in bringing out the given RSD. The experiment is then discontinued and a decision taken on acceptance/rejection of the null hypothesis, with the errors reduced to the prescribed levels.
The peculiar feature of this type of sequential testing is that it brings out the difference between the characteristics of two random quantities, rather than indicates a decision regarding a single one.
A Mathlab written simulation program yielded examples of the operating characteristics of the experiment and the distributions of the number of failures up to the decision. Also constructed were the dependences of the average failure rate up to the decision on the actual relative MTBF differences, the given level of the latter, and the given error probability levels.
Prediction of the sample size, together with the operating characteristic, permits more
rational planning of the experiment, hence of the maintenance policy.
The data collected from a group of helicopters maintained according to the proposed policy were compiled and analyzed by several types of hypothesis test (such as the chi-square test and F-test for the MTBF ratio) ,demonstrated the efficiency of the sequential test for MTBF ratio method.