Substrate processing system, substrate processing apparatus, anomaly detection method, and anomaly detection model construction program

Abstract

The present invention relates to a substrate processing apparatus that can suppress false alarms and detect minor anomalies. A learning unit is provided in the substrate processing system, which performs unsupervised learning by applying multiple time-series data acquired by sensors during normal operation of the substrate processing unit to a machine learning model 3. The machine learning model 3 is composed of, for example, three autoencoders 30(1) to 30(3). The learning unit applies multiple sensor values ​​as input data to the first autoencoder 30(1), and applies the difference between the input data and output data of the previous autoencoder as input data to the second and subsequent autoencoders, thereby updating the parameters of the first autoencoder 30(1) in a manner that minimizes the loss function based on the difference between the input data and output data of the third autoencoder 30(3).

Images