Fast self-learning fault diagnosis method for industrial pneumatic control valves based on multi-feature fusion

The complex characteristics of pneumatic control valves make it difficulty to describe valve faults by establishing a accurate mathematical model, data-driven technology thus attracts widespread attention in the filed of its fault diagnosis. The existing control systems of commercial regulating valves, however, are always equipped with limited hardware equipment, which puts forward higher requirements for the fault diagnosis model and learning efficiency. Therefore, this paper presents a fast self-learning fault diagnosis method for pneumatic control valves based on multi-feature fusion. Firstly, by integratding the cloud model(CM) and dynamic-inner principal component analysis(DiPCA), a fault feature fusion method of pneumatic control valves is proposed to improve the quality of input information for the diagnosis model. Then, a low discrepancy stochastic configuration network is established to construct the diagnosis model quickly and autonomously in a supervised incremental manner according to the low discrepancy sequence, effectively improving the learning efficiency and compactness of the model. Finally, experimental data from the DAMADICS platform are employed to verify the rapidity and accuracy of the proposed method.