Temperature-Independent Fault Detection of Solenoid-Actuated Proportional Valve

Most electrically actuated hydraulic valves utilize solenoids as the actuating element due to their robustness and simplicity. This goes for both
on–off
and proportional-type valves. However, despite their robustness, solenoid coil malfunction is the largest single failure mode in solenoid actuated valves. An outspoken fault is here solenoid winding short-circuit, i.e., two windings short-circuiting, which may ultimately lead to solenoid failure if more windings short-circuit. Research has therefore also focused on detecting winding short-circuits. Common for the approaches are that they, directly or indirectly, depend on the coil winding temperature, as this directly influences the coil resistance. Alternatively, the approaches are based on injection of high-frequency signals, which is typically a costly solution, which is not a feasible approach for use in hydraulic valves, with the limitations imposed by the control electronics. Therefore, this article focuses on a temperature-independent algorithm to detect coil winding short-circuit, which is easy to implement and only relies on existing position and current sensors. The proposed algorithm is based on an extended Kalman filter, which estimates the coil resistance. As this resistance estimate is indirectly dependent on the coil temperature, a window-based cumulative sum fault detection method is included to detect transient changes in the coil resistance while compensating for the temperature variations. The algorithm is developed based on an experimentally validated model of the valve, and has been tested for several different situations through both simulations and experimentally. Based on the presented results, it is found that the algorithm can consistently detect resistance changes down to 0.11
$\Omega$
for constant input signals and down to 0.17
$\Omega$
for sinusoidal-varying input signals. This while stillbeing robust to parameter variations, such as increased valve friction, spring coefficients, and sensor signal deviations.