Solenoid valves are enabling components for flow and motion control in fluid power systems. It is important for reduction of a machinery’s “off” time if quick fault diagnosis of solenoid valves can be performed on site. Traditional fault diagnosis usually involves signal convolution, machine learning, or fuzzy logics, which can achieve over 90% accuracy but are all computationally intensive, and may be difficult to achieve rapid analysis on site with portable electronics. Among various kinds of faults of solenoid valves, the most common ones are the leakage caused by poor sealing between the poppet and the sleeve, stiction caused by foreign matters in the valve body, or solenoid failure. This paper proposed to simplify the fault diagnosis process by derivation of analytical models of fault characteristics of solenoid on-off valve based on electromagnetics and lumped mass fluid mechanics. Firstly, according to the structure of the solenoid valve, the electromagnetic model of the armature, solenoid, and the air gap is deduced. Secondly, five features from two sensor data are constructed. Fault isolation matrix, tested features are also defined. Thirdly, experimental system was setup to acquire the key threshold values for membership function definition. Finally, two validation tests were conducted and the initial results showed that the proposed method is capable of detecting solenoid valve stiction of various degrees.