|M.Sc Student||Golan Shai|
|Subject||Reliability of Naval Electronic System and Experimental|
Method for Improvement Evaluation
|Department||Department of Quality Assurance and Reliability||Supervisor||Dr. Yefim Haim Michlin|
The system under consideration initially performed its tasks with adequate reliability. A few years ago faults were detected, and a reliability analysis was initiated. Because of the modifications introduced in the process, a method is now needed for experimental evaluation of the achieved improvement.
Failures are documented manually over relatively short periods, with the calendar times of occurrence recorded. The latter data were converted to the base of active time between failures.
The paper presents data for the overall system and at card level, matched against theoretical distributions.
The counting time up to the decision about the improvement should be as short as possible. This is realizable through sequential testing. Such procedures are described in Wald (1947), but there the MTBF's are evaluated against a prescribed value. An article by Michlin and Migdali (2004) presents a method of sequential testing which involves simultaneous treatment of a pair of systems - basic (MTBFb) and improved (MTBFn). There are however cases, such as the present one, in which data are already available, and a rough estimate of MTBFb is derived from them. A method for evaluating MTBFn against that estimate is presented in this paper. It involves algorithms which produce simulations for this type of experimentation. The output comprises the operating characteristic curves of the test and the distribution of its duration versus the MTBFn/MTBFb ratio .It was shown that when, in the proposed method, the failure count used in assessing a basic system tends to infinity - the resulting curves coincide with Wald's theoretical ones. A test for the system under consideration is being planned, using the proposed method.