At present, the detection thresholds of most on- and off-sensors of most mining equipment are fixed, and the thresholds for on- and off-state determination need to be adjusted manually. The adjustment process is complicated and unintelligent. Aiming at this situation, a working method of start-stop sensor self-learning was proposed. The method was based on the magnetic field intensity actually detected on the spot, and performs self-learning based on statistical principles to obtain the equipment start-stop threshold. Practice showed that, compared with the existing methods of manually adjusting the potentiometer or the remote controller to adjust the register, the method greatly improved accuracy, flexibility, intelligent programs, etc., and was easier to use on site.