|M.Sc Student||Levin Ilya|
|Subject||Improvement of Planning Methodology for Comparison|
|Department||Department of Quality Assurance and Reliability||Supervisors||Dr. Yefim Haim Michlin|
|Professor Dov Ingman|
|Full Thesis text - in Hebrew|
When two systems (basic and new) are compared for their mean time between failures (MTBF), it often occurs that the experiment has to be performed with only partial information available about the basic system due to a limited number observations, which makes for uncertainty about its MTBF. The compared systems can be of various types, such as an old model vs. a new or improved one, an old technology versus a new one, etc. In these cases the decision parameter is the MTBF ratio between the two systems. The test is designed to check whether the new system is superior to the basic by a certain factor, it being also assumed that the times between failures in both systems are distributed exponentially.
This work presents a new cost-effective planning approach for a comparison sequential test with partial information available about the MTBF of the basic system using non-simulated solutions, such as the Poisson pattern. This approach yields an array of solutions over a broad range of values corresponding to the practical needs of the planner.
The required accuracies in planning sequential tests via the Monte-Carlo method can only be reached in a time-consuming process. This study yielded a method which allows the user to find the required parameters without recourse to simulations. The input for calculating the parameters is the error probabilities of the first and second kind, the discrimination ratio, and the number of observations on the basic system.
The dependences between the parameters were analyzed, and on their basis explicit formulas were derived for calculating the values required for the planning process.