Hybrid Deep Ensemble Models for Real-Time Prognostics of Vehicular Engine Health Using Temporal Sensor Data

作者：P. Agrawal, Kamlesh Lakhwani · 发表于：2025 2nd International Conference On Multidisciplinary Research and Innovations in Engineering (MRIE) · 年份：2025 · DOI：10.1109/MRIE66930.2025.11156214

Recent years' transformation of automotive systems by the integration of Artificial Intelligence (AI) and the Internet of Things (IoT) has created a large need for accurate and fast engine health prognosis. Our research addresses this demand by introducing a hybrid deep ensemble modeling approach, designed to efficiently forecast engine health conditions using temporal sensor data from vehicles [10]. The methodology involves rigorous data preprocessing steps, including feature normalization, time-window slicing, and robust outlier handling, applied to a comprehensive dataset encompassing engine temperature, vibration, oil pressure, and speed logs. The efficacy of our trained and validated model is underscored by experimental results demonstrating its superior performance compared to individual constituent models, showcasing its potential for real-time engine prognostics in resource-constrained edge environments. The bar graph presented in the figure visually corroborates these findings by comparing the classification accuracy scores across various models. While individual models like SVM and Logistic Regression offer a foundational level of prediction, and boosting techniques like AdaBoost and XGBoost show varying degrees of success, our ensemble strategy yields a notable improvement. Crucially, the Optimized Ensemble model achieves the highest accuracy (0.6604), suggesting that a carefully orchestrated combination of diverse learning algorithms can indeed enhance predictive ...