Correlation Analysis and Anomaly Detection in Gas Piston Unit Data Using Machine Learning Methods

This study explores correlation analysis and machine learning for detecting anomalies in gas piston unit (GPU) data. Pearson and Spearman correlation reveal key parameter dependencies, aiding fault diagnosis. Anomaly detection is performed using k-Nearest Neighbors (kNN) and Isolation Forest, identifying deviations in oil temperature, cooling system pressure, and fuel supply. The results demonstrate the effectiveness of predictive modeling in early fault detection, reducing maintenance costs. Future work includes real-time monitoring integration and advanced anomaly detection techniques.