Detecting abnormal behavior of automatic test equipment using autoencoder with event log data

•Detecting abnormal behavior of automatic test equipment using event log data.•Monitoring the behavior of the wafer moving/testing within the equipment.•Conducting an empirical study to validate the effectiveness of a new approach.•Identifying repetitive temperature control/check as abnormal behavior in practice.•Collaborative efforts of data analysts and test engineers resulted in the approach. The utilization of automatic test equipment (ATE) in a wafer test is crucial because it helps maintain optimal temperatures and simultaneously provides electrical power for multiple chips. Ensuring the proper functioning of ATE is essential to avoid productivity losses, such as extended wafer testing times and potential malfunctions. Although previous research has investigated the detection of ATE malfunctions using yields, probe cards, and electrical characteristic data, these approaches exhibit limitations in data characteristics and are not suitable for continuous monitoring. This study introduces a novel approach for detecting abnormal behavior within ATE by using event log data and an autoencoder. Autoencoders have demonstrated efficacy in identifying abnormal behavior, making them apt for detecting such behavior within ATE using event log data. To develop this approach, data analysts and wafer test engineers collaborated to establish a practical methodology that encapsulates the operational characteristics of ATE. This methodology comprises four distinct steps. The proposed approach was assessed using simulated event log data and was proven effective in detecting abnormal behavior within ATE. Furthermore, the approach was applied to real-world ATE data from a semiconductor company, identifying abnormal temperature control and monitoring behavior. This detection can lead to reduced wafer test times and contribute to environmental protection efforts.