An Improved Genetic Algorithm for Pot Tending Machine Scheduling

Abstract

Pot tending machine (PTM) constitutes an essential asset for electrolytic aluminum industry. This paper proposes an improved genetic algorithm for PTM scheduling problem, the problem of scheduling a fixed number of PTMs to do some types of tasks for each electrolytic cell. There are several constraints, and several task types in the scheduling problem, for example, PTM must keep a safety distance between each other in order to prevent dangerous collisions and one electrolytic cell may have several types of task to be done with the help of PTM. Firstly, this work put forward a novel chromosome structure called "2+1" and a method called within-class to deal with the problem using genetic algorithm adaptively. In addition, the paper use between-class method and mutation of the job number to improve diversity of solutions for genetic group. To enhance efficiency of genetic algorithm and keep cooperation with adjacent PTMs, this paper also raises a heuristic initial condition. Experimentation shows the proposed algorithm has a satisfying performance for PTM scheduling.