Comparison of Machine Learning Models for Fault Classification in Rotating Machines

This paper evaluates the performance of neural network architectures, including Multi-Layer Perceptron (MLP), Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM), for classifying faults in rotating machinery. Using data from the MAFAULDA simulator, experiments were conducted with fixed and overlapping windows, comparing models across metrics such as accuracy, precision, recall, F1-score, and computational efficiency. Results show that LSTM achieves 98.7% accuracy with overlapping windows. At the same time, MLP excels in computational speed, with an average training time of 0.762 milliseconds per window. These findings contribute to developing efficient fault classification systems using machine learning techniques.