Machine Learning Framework for Industrial Machine Sound Classification in Predictive Maintenance

The "Machine Learning Framework for Industrial Machine Sound Classification in Predictive Maintenance" project aims to develop an advanced system for the classification of industrial machine sounds. The system analyzes audio data from industrial pumps to detect anomalies that may indicate mechanical issues or failures. By leveraging machine learning algorithms such as CNN+LSTM, AST, CRNN, Inception and AE, the framework efficiently processes and classifies sound data into normal and faulty categories. The framework supports predictive maintenance by identifying potential problems before they become critical, thus reducing downtime and improving the longevity of industrial machinery. The project employs a dataset of audio recordings representing different machine states and uses various feature extraction and classification techniques to ensure accurate predictions. The backend of the system is built using Python and Flask, while the frontend is developed using HTML, CSS, and JavaScript. The system also includes modules for user registration, login, sound classification, and model comparison.

Keywords:

 Industrial machines, sound classification, machine learning, predictive maintenance, CNN+LSTM, AST, CRNN, AE, Inception feature extraction, Flask.