Data mining–based disturbances prediction for job shop scheduling

Abstract

In real production manufacturing process, there are many disturbances (e.g. machine fault, shortage of materials, tool damage) which can greatly interfere the original scheduling. These interventions will cost production managers extra time to schedule orders, which increase much workload and cost of maintenance. On account of this phenomenon, a novel system of data mining–based disturbances prediction for job shop scheduling is proposed. It consists of three modules: data mining module, disturbances prediction module, and manufacturing process module. First, in data mining module, historical data and new data are acquired by radio frequency identification or cable from database, and a hybrid algorithm is used to build a disturbance tree which is utilized as a classifier of disturbances happened before manufacturing. Then, in the disturbances prediction module, a disturbances pattern is built and a decision making will be determined according to the similarity between testing data attributes and mined pattern. Finally, in the manufacturing process module, scheduling will be arranged in advance to avoid the disturbances according to the results of decision making. Besides, an experiment is conducted at the end of this article to show the prediction process and demonstrate the feasibility of the proposed method.