Application of Genetic Algorithm with Preprocessing to Cable Installation Design in Large-Scale Plant

Abstract

This paper describes the application of genetic algorithm (GA) to a cable installation design in a large-scale plant. This application assumes that the design of plant equipment and cable trays has been completed. Here, to deal with the exceedingly large number of cable-route combinations in a large-scale plant, we propose a new preprocessing method using the Dijcstra method for constructing genes that significantly reduces the number of combinations and improves the search for the shortest route. Moreover, based on correlation analysis between evaluated values, we pay particular attention to cable grouping that skilled designers have come to emphasize in their design work, and propose a new method for installing cables that uses cable grouping as a genetic operator. This method achieves a level of optimization exceeding that of manual design and substantially reduces computational time. Comparing with actual design data from skilled designers, it was found that the above methods can produce an even more satisfactory solution in a practical period of time (within two hours) on a workstation.