Beyond Industry 4.0: Leveraging AI-powered Anomalous Sound Detection for Smart Maintenance

The ongoing global changes, pushing the digital transformation to Industry 4.0, have been reflected in the launch of new services and process innovations tackling the existing pressure on costs and prices. In this context, AI is becoming an integral part of all future smart maintenance endeavors. The new generation of intelligent maintenance systems, driven by big data analysis and advanced diagnostics, are already guiding automated predictive innovation towards the idea of zero-failure activity. Automated detection of failures is crucial for smart maintenance, for building AI-based factory automation. In this context, the paper describes a solution for detecting failures based on sound obtained from the target machines. Abnormal sound data is difficult to collect, as it rarely occurs and is being hard to extract from a noisy environment and could have various patterns. The proposed solution detects anomalous sound after training the machine-learning model only with the normal operating sound of machines in a factory environment.