Bearing faults are the most common type of failure in induction motors, given their long operating times and mechanical loads. Because induction motors in industrial environments operate under various load conditions, effective methods for diagnosing bearing faults across these conditions have become increasingly important. Here, different load conditions were implemented with a powder clutch and a tension controller, and vibration data were acquired under both normal and faulty bearing conditions. To ensure diagnostic accuracy while improving time efficiency, a model bank-based fault diagnosis classifier is proposed, which utilizes independent classifiers trained for each load condition. For comparison, a single model-based classifier trained on all load conditions is also implemented. Both approaches are validated with three classifiers: support vector machine (SVM), multilayer neural network (MNN), and random forest (RF), with three input types: raw time-series signals, six statistical features, and three t-test–selected statistical features. Experimental results reveal that the model bank-based fault diagnosis classifier utilizing three statistical features selected by t-test maintained 98–100% accuracy while reducing operating time compared with Method 1 by 60.0, 71.2, and 60.0% for SVM, MNN, and RF, respectively. These results confirm that the proposed Method 2 utilizing time-domain analysis provides reliable and time-efficient performance for bearing fault diagnosis under variable load conditions.

Reliable fault detection in brushless DC motors is challenging owing to environmental complexity and high equipment costs. To address these challenges, we propose an effective and cost-effective approach using an optimized end-to-end one-dimensional convolutional neural network. Specifically, a real experimental platform simulating bearing and eccentricity faults was developed. Statistical t-tests indicated that three-axis accelerometer signals from a low-cost inertial measurement unit provided sufficient fault information for the present diagnosis task. Unlike traditional methods such as support vector machines, multilayer neural networks, and random forests, which rely on manual feature extraction, our model learns directly from raw waveforms and can handle signal drift. Under the present controlled experimental setting and the leave-one-day-out evaluation protocol, the model achieved 100.00% average window-level classification accuracy, considerably outperforming traditional methods, the performances of which declined to 67.95–71.37% under environmental shifts. Moreover, with an inference time of only 0.96 ms, 32 times faster than that of random forests, this approach is well suited for real-time embedded monitoring. The proposed method demonstrates strong potential for cost-efficient and robust fault diagnosis under the present experimental setting.