Abstract: Gas turbine production is a typical kind of discrete manufacturing. The production characteristics of multiple varieties and small batches present challenges to workshop scheduling, resulting in low production efficiency and difficulties of meeting product delivery deadlines. The Harmony Search (HS) algorithm is often used to solve such workshop scheduling problems due to its simplicity and ease of operation. However, the convergence rate of traditional HS algorithm is relatively low, and it is easy to get trapped in local optima. Accordingly, this paper builds a mathematical model for scheduling in a gas turbine manufacturing workshop with the objective of minimizing the maximum completion time. A discrete improved multi-population hybrid HS algorithm is proposed to solve the problem. Combining the advantages of HS algorithm and the variable neighborhood search algorithm, we propose an encoding method based on operations. The Metropolis rule of simulated annealing is used in population iteration to improve population diversity. An adaptive memory retention probability and pitch adjusting rate are proposed to adjust parameters, improving the global optimization capability of the algorithm. We also incorporate variable neighborhood searching to accelerate the convergence of the proposed algorithm. Performance tests and case studies show that the proposed algorithm outperforms existing algorithms.

Key words: gas turbine manufacturing workshop scheduling, harmony search (HS) algorithm, variable neighborhood search (VNS), Metropolis rules