Clustering Parameter Optimization of Predictive Maintenance Algorithm for Semiconductor Equipment Using One-way Factorial Design

In this paper, the predictive maintenance(PdM) technology was applied to the semiconductor manufacturing equipment and the DBSCAN algorithm parameters were optimized. The data obtained from the semiconductor equipment was obtained using acceleration sensors and was used to determine the conditions of the equipment. To simulate the malfunction condition of the equipment, the virtual error data was artificially generated. The DBSCAN clustering algorithm is applied to both acceleration sensor data and virtual error data to trace the moving of the center point of each cluster. By tracing the center points, the appropriate time for equipment maintenance can be determined. Also, the parameters of the DBSCAN algorithm were optimized using one-way factorial design to improve to determine the exact time of maintenance.

1. Introduction

In the semiconductor industry, the increased reliance on wafer handling equipment has also increased interest in predictive maintenance of the equipment.