| Summary: | 碩士 === 朝陽科技大學 === 資訊工程系 === 105 === This thesis proposes a discrete backtracking search optimization algorithm (DBSA) to solve the stochastic job shop scheduling problem (SJSSP). The proposed DBSA utilizes the advantage of multi-directional search in backtracking search optimization algorithm (BSA) and quickly and efficiently selecting candidate solution in optimal computing budget allocation (OCBA). The goal is to find a good enough and high reliable solution in a reasonable computation time. The SJSSP is a practical and popular job shop scheduling problem. The stochastic characteristic indicates that the process time of a job is fluctuant randomly. The SJSSP is an NP-hard problem. To solve this NP-hard problem efficiently, the SJSSP is first formulated as a constraint stochastic simulation optimization problem with the objective function considering the sum of tardiness and earliness. Two examples include 6 jobs on 6 machines and 10 jobs on 10 machines are used to test the proposed DBSA. The random processing time in the experiment has three distributions: truncated normal, uniform, and exponential. The proposed DBSA is compared with Genetic algorithm, memetic algorithm, and existing dispatching rules. Result shows that the proposed DBSA can obtain a good enough solution in a reasonable computation time.
|
|---|
