Computer Diagnostics and Forecasting of Pulse Electron Linac Failures

Most failures of linear electron accelerators (LEA) are caused by malfunctions in the systems of RF-feeding, modulators and clystrons. And a great part of malfunctions lead to the increased dispersion of amplitudes of output signals of the system. In this paper we consider the algorithms of failure prediction based on some indication functions and the use of these algorithms in the LEA control systems. I. EVALUTION OF A POSSIBILITY FOR FAILURES IN THE ACCELERATOR PULSED SYSTEM Let us consider two strategies of resource expenditures for troubleshooting the LEA systems. The first strategy, we shall call it as “de facto”, envisages the resource expenditure only after failure detection by the accelerator operator. The second or the probabilistic strategy is based on the prediction provided by the control system computer. Let n be the mean number of failures in a system for some time period, e.g. per year. Then, for the “de facto” strategy, the annual average expenditures for troubleshooting in this system are