Spindle bearings fault diagnosis in manufacturing equipment using a one-dimensional convolutional neural network

This paper presents a simple and reproducible method for the early detection of spindle bearing degradation in manufacturing equipment using a one-dimensional convolutional autoencoder. As baseline comparators, the crest factor and mel-frequency cepstral coefficients were evaluated on a test set of 375 examples: 75 normal signals and 300 anomalous signals (100 examples for each of three fault types). To validate the proposed approach, vibration signals were synthesized to represent the normal condition and three anomaly classes — general wear, a spall on the rolling race, and looseness. We examined the effect of two threshold levels, the 95th and 99th percentiles, determined on validation data, and evaluated performance with respect to standard binary metrics such as recall, precision, and F1 score, the area under the Receiver Operating Characteristic curve, and the area under the Precision–Recall curve, as well as metrics specific to each anomaly type (Impulse, Drift, Noise). The experiments show that the autoencoder substantially outperforms the baselines in overall binary detection, achieving an area under the Receiver Operating Characteristic curve of 0.9985 and an area under the Precision–Recall curve of 0.9996, with recall up to 0.987 and precision up to 1.000 at the 99th percentile threshold, demonstrating a strong ability to distinguish normal from anomalous conditions.