Bearing-shaft systems are essential components in various automated manufacturing processes, primarily designed for the efficient rotation of a main shaft by a motor. Accurate fault detection is critical for operating manufacturing processes, yet challenges remain in sensor selection and optimization regarding types, locations, and positioning. Sound signals present a viable solution for fault detection, as microphones can capture mechanical sounds from remote locations and have been traditionally employed for monitoring machine health. However, recordings in real industrial environments always contain non-negligible ambient noise, which hampers effective fault detection. Utilizing a high-performance microphone for noise cancellation can be cost-prohibitive and impractical in actual manufacturing sites, therefore to address these challenges, we proposed a convolution neural network-based methodology for fault detection that analyzes the mechanical sounds generated from the bearing-shaft system in the form of Log-mel spectrograms. To mitigate the impact of environmental noise in recordings made with commercial microphones, we also developed a denoising autoencoder that operates without requiring any expert knowledge of the system. The proposed DAE-CNN model demonstrates high performance in fault detection regardless of whether environmental noise is included(98.1%) or not(100%). It indicates that the proposed methodology effectively preserves significant signal features while overcoming the negative influence of ambient noise present in the collected datasets in both fault detection and fault type classification.

Anomaly detection for each industrial machine is recognized as one of the essential techniques for machine condition monitoring and preventive maintenance. Anomaly detection of industrial machinery relies on various diagonal data from equipped sensors, such as temperature, pressure, electric current, vibration, and sound, to name a few. Among these data, sound data are easy to collect in the factory due to the relatively low installation cost of microphones to existing facilities. We develop a real time anomalous sound detection (ASD) system with the use of Autoencoder (AE) models in the industrial environments. The proposed processing pipeline makes use of the audio features extracted from the streaming audio signal captured by a single-channel microphone. The pipeline trains AE model by the collected normal sound. In real factory applications, the reconstruction error generated by the trained AE model with new input sound streaming is calculated to measure the degree of abnormality of the sound event. The sound is identified as anomalous if the reconstruction error exceeds the preset threshold. In our experiment on the CNC milling machining, the proposed system shows 0.9877 area under curve (AUC) score.