A hybrid approach for due date assignment in a dynamic job shop

Abstract

Due date assignment (DDA) is a crucial issue in the job shop scheduling. Orthogonal Kernel Least Squares Algorithm (OKLSA) has been demonstrated to be a useful tool for DDA in the dynamic manufacturing environment by utilizing the aggregate job- and shop-information. This paper proposes a hybrid approach named OKLSAopr by combining OKLSA with the idea of job-split-into-operations and applies it to DDA in a dynamic job shop dispatched by the Shorting Processing Time rule. OKLSAopr estimates the flow-time of a job by adding up the flow-times of the corresponding split operations, which are predicted by OKLSA with the selected operation-level information as the input attributes. A discrete-event simulation model is built to intimate the production process of a dynamic job shop and to gather experimental datasets. The results of the numerical experiments and paired t-tests indicate that the performance of the proposed OKLSAopr is statistically superior to those of five conventional rules, Back-Propagation Neural Networks, and OKLSA in terms of either mean absolute lateness or root mean squares lateness criterion.