Study on Different Crossover Mechanisms of Genetic Algorithm for Test Interval Optimization for Nuclear Power Plants

Surveillance tests are performed periodically on standby systems of a Nuclear Power Plant (NPP), as they improve the systems‘ availability on demand. High availability of safety crit ical systems is very essential to NPP safety, hence, carefu l analysis is required to schedule the surveillance activ ities for such systems in a cost effective way without compromising the plant safety. This forms an optimization problem wherein, two different cases can be formulated for deciding the value of Surveillance Test Interval. In one case, cost is the objective function to be minimized while unavailab ility is constrained to be at a given level and in another case, unavailability is minimized for a g iven cost level. Here, optimizat ion is done using Genetic Algorithm (GA) and real encoding has been employed as it caters well to the requirements of this problem. A detailed procedure for GA formulat ion is described in this paper. Two different crossover methods, arithmetical crossover and blend crossover are exp lored and compared in this study to arrive at the most suitable crossover method for such type of problems.