A hybrid estimation of distribution algorithm for the semiconductor final testing scheduling problem

As the last process of the semiconductor fabrication, the final testing is crucial to guarantee the quality of the integrated circuit products. The semiconductor final testing scheduling problem (SFTSP) is of great significance to the efficiency of the semiconductor companies. To find satisfactory solutions within reasonable computational time, the intelligent manufacturing scheduling based on the meta-heuristic methods has become a common approach. In this paper, a hybrid estimation of distribution algorithm (HEDA) is proposed to solve the SFTSP. First, novel encoding and decoding methods are proposed to map from the solution space to the schedule space effectively. Second, a probability model that describes the distribution of the solution space is built to generate the new individuals of the population. Third, a mechanism is used to update the parameters of the probability model with the superior solutions at every generation. Furthermore, to enhance the exploitation ability of the algorithm, a local search procedure is hybridized to find neighbor solutions of the promising individuals obtained by sampling the probability model. In addition, the influence of parameters is investigated based on Taguchi method of design-of-experiment, and a set of suitable parameters is suggested. Finally, numerical simulation based on some benchmark instances is carried out. The comparisons between the HEDA and some existing algorithms demonstrate the effectiveness of the proposed HEDA in solving the SFTSP.