Abstract: In view of the practical problem that the swing speed of cutting arm cannot be adjusted rapidly to adapt to the current cutting load when the hardness of coal and rock changes sharply in cutting process of roadheader, a control strategy of the swing speed of cutting arm based on BP neural network was proposed to ensure a constant power output of the cutting motor.In this strategy, a judging method of the cutting load based on multi-parameter criterion was proposed, which can accurately judge the change of hardness of coal and rock by the change of voltage and current of the cutting motor, the driving cylinders' pressure of the cutting arm and the vibration acceleration of the cutting arm.In order to realize constant power cutting, a BP neural network was designed to ensure that the output power of the cutting motor was equal to its rated power.The training sample of the BP neural network was obtained based on underground measured data.According to the output stability problem, the designed BP network was optimized by genetic algorithm for better performance, and the training time and iterations were decreased greatly.The mathematical model of the cutting arm system was established, and the control system of swing speed of the cutting arm was established by Simulink method.Aiming at horizontal and vertical cutting condition, the control strategy of swing speed of the cutting arm was simulated.The results show that the swing speed of the cutting arm can be adaptively adjusted according to the input signals of different cutting loads, and the swing speed of the cutting arm is in a constant state after 0.2 s with high control accuracy and fast response speed.The self-adaptive control strategy of swing speed of the cutting arm proposed in this paper can provide theoretical guidance and practical reference for ensuring constant power cutting, extending service life of roadheader and improving the efficiency of the roadway drivage.