Anomaly detection method and system for industrial machines, and non-statutory storage media

Abstract

An anomaly detection method for industrial machines applied in an industrial environment is disclosed. The method comprises the steps of: receiving machine data from an industrial machine; processing the machine data through an AI model to obtain reconstructed data, wherein the AI model comprises at least a Convolutional Autoencoder (CAE); and performing a dissimilarity calculation between the machine data and the reconstructed data to determine whether the machine data corresponds to normal or abnormal conditions. In particular, the Convolutional Autoencoder includes a self-attention layer and a squeeze-and-excitation network layer.

Images