Advanced Hybrid Machine Learning Framework for Fault Prediction in Rotating Equipment Leveraging Random Forest and Gradient Boosting on High-Frequency Sensor Data

Abstract

Predictive maintenance of rotating equipment has become essential for ensuring operational reliability, optimizing lifecycle costs, and minimizing unplanned downtime in industrial systems. This study presents a Hybrid Ensemble Framework that integrates Random Forest (RF) and Gradient Boosting (GB) to enhance the accuracy and robustness of fault prediction using high-frequency vibration data. The proposed framework leverages the complementary strengths of both ensemble models: RF provides resilience to noisy features and reduces variance through bootstrap aggregation, while GB improves overall predictive power by sequentially minimizing residual errors. High-frequency vibration signals are preprocessed, denoised, and transformed into meaningful time–frequency features using advanced signal-processing techniques. These features serve as inputs to the hybrid model, which employs a weighted fusion strategy to generate reliable fault-classification outputs. Experimental results demonstrate that the hybrid approach outperforms individual baseline models in terms of precision, recall, F1 score, and early fault-detection capability. The framework offers a scalable and data-driven solution for real-time monitoring, enabling industries to transition from traditional reactive maintenance to intelligent condition-based decision-making.