CONTROL OF A COPPER LASER USING NEURAL NETWORKS

Urenco (Capenhurst) Ltd. has been running high-power copper lasers for several years owing to an interest in selective photo-ionisation for commercial enrichment. The main advantages of these lasers as pump sou rces for this process are their efficiency at high power and their hig h pulse repetition frequency However, certain load discharge condition s have been found to greatly reduce the lifetimes of expensive laser m odulator components and therefore increase the running costs and downt ime of the lasers. This article describes the use of multiple coupled neural networks which? trained by humans with the required experience, are first able to identify load conditions likely to result in a redu ction in the lifetime of the modulator components and then alter the l oad conditions to minimise the stress placed on the modulator.