Prediction of failures in C.T.R. systems

The problem of predicting failure rates is becoming more important in C.T.R. engineering because of the large syatems under consideration. A large theta-pinch system will contain thousands of identical components, many of which are barely beyond the development stage and whose failure rates may be fairly high. A typical ''mean time between failure'' calculation assumes all the components in the system are on the flat part of their failure rate curve. This means the early failure components have been eliminated before that particular group of components is installed in the system. Unfortunately, some of the critical components in a large theta-pinch system may not be tested sufficiently to reach the flat part of the failure rate curve because of the time and expense involved. When this occurs, a standard MTBF calculation will have little or no meaning. Since usually only a few types of components determine the failure rate in CTR systems it is possible to compute an analogous quantity, the''expected time to next failure,'' provided the failure distributions of the critical component types are known. A Monte Carlo method for the computation when failures are independent is described and examples are given. (auth