For the fact that the current fault diagnosis methods cannot detect soft fault effectively, a variable detection method based on electric servo system model was proposed for health simulation and evaluation. Common failures of electric servo system, such as motor module, controller and transmission mechanism were analyzed, and the mathematical models were built up for common health states. By injecting different health coefficients into the system, the characteristics of state variable were obtained, so the knowledge base was built up to locate the source of faults and to evaluate the changing trend of the failure parameter by health factors. Finally, the resistance of winding and the motor torque constant are chosen as health factors, the feasibility of health factors estimation algorithm is verified by simulation.