A learning-based methodology for dynamic scheduling in distributed manufacturing systems

Chanshing Chiu, Purdue University

Abstract

Heterarchical control manufacturing systems characterized by pursuing full local autonomy and a cooperative approach to global decision making provide attractive prospects such as reduced complexity and higher modularity. However, low productivity is often a problem with such systems. The low productivity, we believe, is mostly due to the lack of an efficient scheduling scheme. To enhance productivity in a distributed manufacturing system under heterarchical control, we develop a framework for a dynamic scheduling scheme that explores routing flexibilities and handles uncertainties. We propose a learning-based methodology to extract scheduling knowledge for machines to real-time select next coming parts. The proposed methodology includes three modules: discrete-event simulation, example generation, and incremental induction. First, a sophisticated simulation module is developed to implement the dynamic scheduling scheme and to generate training examples. Second, in an example generation module, the technique of searching good training examples for inductive learning is proposed. Finally, in an incremental induction module, the tolerance-based learning algorithm developed not only acquires scheduling knowledge from training examples, but also adapts to any newly observed examples and thus facilitates knowledge refinement. The experimental results show that the number of modifications on scheduling knowledge in terms of a decision tree is significantly reduced and the accuracy is also enhanced when given a small tolerance. The dynamic scheduling scheme outperforms static scheduling with the aid of scheduling knowledge in a distributed manufacturing system.

Degree

Ph.D.

Advisors

Yih, Purdue University.

Subject Area

Industrial engineering

COinS