Detection of imbalance faults in industrial machines by means of frequency-based feature extraction using machine learning and deep learning approaches

This study investigated the effectiveness of machine learning and deep learning models in diagnosing imbalance faults in industrial machines, using Fast Fourier Transform (FFT) for frequency-based feature extraction. As imbalance shortens equipment life and increases maintenance costs, vibration data was analysed and frequency components were extracted using FFT for classification. Support Vector Machine, Random Forest and Multi-Layer Perceptron models were then compared using the metrics of accuracy, precision, recall and F1 score. The Multi-Layer Perceptron model performed best with 99% accuracy, capturing the patterns extracted by FFT most effectively. Random Forests made successful predictions, but had a high error rate in some classes. Support Vector Machines, on the other hand, offered lower accuracy. Combining FFT with machine learning contributes to the diagnosis of faults in rotating machines. Model performance could be improved in future using larger data sets, hyperparameter optimisation and methods such as wavelet transformation.