Fault Identification Method for Seal Wear State in Hydraulic Cylinder Based on Multi-Sensor Multi-Level Information Fusion

Internal leakage caused by seal wear is one of the most common faults in hydraulic cylinders, and the timely diagnosis of seal wear and internal leakage is crucial for ensuring the reliable operation of hydraulic systems. Therefore, a diagnostic method based on multi-sensor, multi-level information fusion is proposed in this paper. First, an improved wavelet threshold denoising method based on ensemble empirical mode decomposition (IWTD-EEMD) is used to denoise and preprocess multi-sensor signals, which makes fault features more obvious. Secondly, statistical features are extracted from the preprocessed signals to form the original feature set. Then, the coupled feature selection method combining the distance evaluation technique (DET) and max-correlation min-redundancy (mRMR) is used to optimize the original feature set, forming the optimal feature subset to achieve feature-level information fusion. Finally, decision-level fusion diagnosis is performed based on a convolutional neural network (CNN) and Dempster–Shafer (D-S) evidence theory. The experiments are conducted under five load conditions, and the proposed method achieves a diagnostic accuracy of 99.26%. Compared with SVM, BPNN, and DBN models, the proposed method demonstrates superior performance in identifying seal wear states, indicating that it can effectively determine the severity of seal wear. However, this study relies on a limited experimental dataset and does not directly measure actual leakage flow rates. Further validation on independent seals and unseen operating conditions is required before practical application.