Abstract: To improve the effectiveness of the dynamic scheduling for complex manufacturing systems,a data driven dynamic scheduling method was proposed.In the proposed method,the composite rule set was embedded in the scheduling sample data,and then the scheduling sample data were optimized through the Design of Experiment (DOE).To improve the accuracy and efficiency of the dynamic scheduling,the Fuzzy C Means (FCM) clustering and the Extreme Learning Machine (ELM) approaches were successively used to cluster and learn scheduling models from the optimal sample set.As a result,the learnt model was applied to the dynamic scheduling.The proposed method was verified on a semiconductor manufacturing benchmark model,namely theMIMAC6 model.The results showed that it had greater improvements in both the long-term and short-term performance indicators of the manufacturing system compared with the single-rule scheduling approaches.Therefore,the system performance could be comprehensively optimized.

Key words: dynamic scheduling, data driven, extreme learning machine, fuzzy C means clustering, complex manufacturing system

摘要： 为提高复杂制造系统动态调度的有效性,提出一种数据驱动的动态调度方法。采用组合式调度规则作为调度策略,通过试验设计方法对调度样本数据进行优化;采用模糊C均值聚类算法和极限学习机算法对最优样本集进行聚类和学习,得到调度模型供动态调度使用,有效地提高动态调度的精度和效率。所提方法在半导体制造Benchmark模型MIMAC6上进行了验证,结果显示,所提方法较单一规则的调度在制造系统长、短期性能指标上均有较大的改善,能综合优化制造系统生产性能。

关键词: 动态调度, 数据驱动, 极限学习机, 模糊C均值聚类, 复杂制造系统