Virtual Equipment for benchmarking Predictive Maintenance algorithms

Abstract

This paper presents a comparison of three algorithm types (Bayesian Networks, Random Forest and Linear Regression) for Predictive Maintenance on an implanter system in semiconductor manufacturing. The comparison studies are executed using a Virtual Equipment which serves as a testing environment for prediction algorithms prior to their implementation in a semiconductor manufacturing plant (fab). The Virtual Equipment uses input data that is based on historical fab data collected during multiple filament failure cycles. In an automated study, the input data is altered systematically, e.g. by adding noise, drift or maintenance effects, and used for predictions utilizing the created Predictive Maintenance models. The resulting predictions are compared to the actual time-to-failure and to each other. Multiple analysis methods are applied, resulting in a performance table.