A fault diagnosis method for pilot-operated solenoid valves based on CRQA-Stacking integrated learning
As a key control component in fluid systems, fault diagnosis of pilot-operated solenoid valves is crucial for ensuring the stability and reliability of fluid systems. The highly overlapping fault signatures of solenoid valves, together with the industrial difficulty of mass-producing faulty valve specimens. Lead to misjudgment of diagnostic models and further system shutdowns under imbalanced sample distribution. To solve this problem, this paper proposes a fault diagnosis model for pilot-operated solenoid valves based on cross-recurrence quantification analysis (CRQA) and stacking heterogeneous ensemble learning. Firstly, the CRQA method is adopted to extract deep-seated features and expand the sample size. Secondly, classification and regression tree, K-Nearest Neighbor (KNN) and support vector machine are selected as base classifiers, while multinomial logistic regression is used as the meta-classifier to construct a Stacking heterogeneous ensemble model. Bayesian optimization is applied to adjust the hyperparameters of the classifiers, thereby improving the efficiency of ensemble model construction. Then, to tackle the imbalanced distribution of samples, a cost matrix is set up to compensate for misclassification by the model. Finally, fault sample data of solenoid valves are collected through the experimental platform. Ablation experiments and comparative analysis are conducted on the new method. The results show that the new method proposed in this paper achieves high accuracy and stability in fault diagnosis for sparse and imbalanced fault samples.