Industrial Fans Health Analytics with Deep Learning

Industrial fans are essential to many production operations. The fans, however, can malfunction or fail abruptly. Machine health monitoring (MHM) is required for safe and reliable operation. There is a need to develop contemporary technologies because machine failures could have significant consequences. In this project, engineers investigated the health of industrial fans using anomaly detection techniques. They utilized different convolutional neural network (CNN) architectures to predict the health of these fans. To determine the best feature extraction method, they used both Mel Frequency Cepstral Coefficients (MFCCs) and Log-Mel Spectrograms. The CNN models were trained and tested using a dataset obtained from actual experiments. Tests were conducted to evaluate the performance of the deep learning models created. The results showed that the three-layered and five-layered CNN models were effective in discriminating the data and identifying anomalies in the industrial fans' health.