An Intelligent Deep Learning Framework for Industrial Machine Sound Anomaly Detection Using YAMNet Embeddings and Bidirectional GRU Networks

Abstract

Maintaining operational effectiveness, avoiding expensive equipment failures, and guaranteeing worker safety all depend on the ability to identify irregularities in industrial machine noises. However, because industrial surroundings are diverse and unpredictable, this task is difficult. incorporating shifting operational circumstances and background noise. In order to capture the crucial acoustic features of machinery sounds, this study offers a thorough method that makes use of sophisticated feature extraction techniques. Effective anomaly detection tactics are investigated using a variety of machine learning and deep learning techniques Several datasets are used in the study to assess the suggested techniques in various experimental settings. Extensive testing results show the approach's efficacy in real-world industrial settings and highlight its potential to improve sound analysis and predictive maintenance. By offering a dependable way to identify anomalies in machine operations, this work advances the capabilities of industrial monitoring systems.