Engine fault detection has been recognized as a critical component in enhancing reliability, minimizing failures, and reducing maintenance costs. To improve operational safety, fault detection is advanced through the integration of signal processing and machine learning techniques. An effective platform for engine fault detection has been proposed using a Dingo Optimization Algorithm-tuned Malleable Support Vector Machine (DOA-MSVM). Engine sensor data collected under both fault-free and faulty conditions—including key variables such as vibration, temperature, acoustics, and pressure—has been utilized. In the preprocessing stage, a bandpass filter is applied to attenuate noise and preserve fault-relevant frequency components. Short-Time Fourier Transform (STFT) is employed to perform time-frequency analysis, enabling the identification of dynamic signal characteristics associated with faults. The hyperparameters of the MSVM are optimized using the DOA to enhance detection accuracy across varying engine operational scenarios. The proposed framework is implemented in Python and evaluated using precision, recall, accuracy, and F1-score metrics. Experimental findings indicate that DOA-MSVM outperforms conventional classifiers—including Random Forest (RF), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Gradient Boosting (GB)—by achieving superior classification accuracy and robustness, even under diverse acoustic noise conditions. Faults are detected successfully in real time, demonstrating the model's suitability for predictive maintenance applications. Nonetheless, performance degradation is observed under high noise levels, highlighting the need for further investigation. Future work should explore multimodal sensor fusion and adaptive noise filtering strategies to enhance fault detection in complex environments. By combining machine learning with advanced signal processing, an efficient and scalable approach for real-time industrial engine fault diagnosis and predictive maintenance is presented.
Gandhi et al. (Thu,) studied this question.