"Fault Diagnosis Method Using Feature Fusion of Geometric Properties in Compressor Indicator Diagrams"

"Fault diagnosis of reciprocating compressors is crucial for ensuring their reliable and long-term operation. To address the accuracy limitations of conventional diagnostic methods that rely on single feature sets, this paper proposes a novel fault diagnosis method based on the fusion of geometric features from indicator diagrams. This method simultaneously extracts and integrates statistical features and computational geometry features from the indicator diagram, enabling information complementarity between the two types of features. To validate its effectiveness, the proposed approach was tested using multiple supervised learning classifiers. The experimental results demonstrate that the proposed feature fusion strategy achieved 100% classification accuracy on the collected laboratory dataset. Furthermore, robustness tests confirm that the method maintains over 96% accuracy even under simulated industrial noise conditions (SNR=20 dB), demonstrating significant advantages in diagnostic precision and stability compared to single-feature methods. This study demonstrates that the proposed method provides an efficient and highly effective solution for the accurate fault diagnosis of reciprocating compressors."