Performance and Visual Explainability of Chromagram-based CNN-ANN as Acoustic Fault Classifier of Industrial Equipment

The early and accurate diagnosis and prognosis of the defect of industrial equipment is crucial for predictive and prescriptive maintenance. Fault detection in airborne signals using acoustic sensors is a relatively new initiative in the field of predictive maintenance. Anomalies are detected in the sound using the reconstruction error of a CNN Autoencoder (CNN-AE) prior to downtime incidents. Thi...