Application of ANN for Fault Detection in Overhead Transport Systems for Semiconductor Fab

Abstract

In order to ensure safe and fast transportation of wafers in 300 mm semiconductor factories, overhead transport systems (OHT) are primarily used. These systems consist of a rail network and vehicles. To avoid congestion and delays in production, high availability of individual rail sections is essential. In order to ensure this extensive preventive maintenance is required. In this paper, we focus on automatic checks for faults of the rail network by capturing the rail with optical sensors. Our objective is the identification of faults in real time. We considered the identification using artificial neural networks (ANN). Due to the lack of fixed rules designing an ANN we tested different topologies for our application and covered adaptation of ANN to the real conditions in the fab. As a result, our ANN provides accurate real time fault detection which allows a needs-based, resource-saving and efficient maintenance procedure for a reliable OHT and hence 24/7 semiconductor manufacturing.