Cloud manufacturing is one of the modern manufacturing systems that aim to control and manage production processes through a cloud platform to optimize the usage of resources and achieve customer objectives. Scheduling large-scale problems with diversity in the required tasks and the available services is a significant challenge for any researcher, especially with the possibility of recurring dynamic events, such as the arrival of new tasks. Thus, our first objective is to present a mixed-integer mathematical model that demonstrates the complexity of scheduling problems in a dynamic CMfg environment. We developed a parallel distributed genetic algorithm (PDGA) for optimizing large-scale scheduling problems on Apache Spark. The PDGA adopts a resilient distributed dataset (RDD) to improve the decomposition of the population and also enhance both the performance and the speed of the evolution process. The PDGA then updates the schedule by using a greedy strategy to improve the overall performance. To verify the effectiveness of the PDGA, we evaluated the algorithm on a benchmark and generated 8 large-scale problems based on this benchmark. The experiments show that the PDGA provides better performance and lower computational time compared to the traditional genetic algorithm and particle swarm optimization algorithm. |
