Physical Model-based Rapid Quantitative Diagnosis of Solenoid On–Off Valve Spool Stiction Faults

Abstract

Solenoid valves enable flow and motion control functions in the fluid power systems. Even today, on-line diagnosis of fluid power systems still remains a challenging task due to the computational cost and availability of machine operation data sets. For the prior, rapid fault diagnosis of the solenoid fault is of great economic values to the reduction in downtime maintenance. For the latter, currently the data for training networks are the major obstacles, as some of the rare faults are simply unavailable from the usual maintenance data. Facing the challenges, this paper presents a new way of quantifying the spool stiction severeness, a common fault in the solenoid on–off valves, using a proposed coupled physical model, where only temporal features from the solenoid coil driving current were extracted and applied for rapid diagnosis, without the need of spool displacement information. A test system was constructed in laboratory and different settings of valve spool stiction from normal to completely jammed were realized on the hardware. The developed coupled model is validated experimentally and demonstrates the capabilities in capturing the stiction effects. The quantitative diagnosis model based on temporal feature vectors was also tested and compared to the true stiction level, and the proposed sigmoid weightings have shown high prediction accuracy. The initial results have shown that the proposed model can quantify the spool stiction degree with accuracy at least 90% and with computation time less than 500 ms with a CPU at lower than 1.3 GHz.