Research on the roadheader cutting control system based on convolutional neural network and fuzzy PID

In response to the insufficient adaptability and low system stability of cantilever roadheader when facing changes in coal and rock hardness during tunneling, a roadheader cutting control system based on convolutional neural networks (CNN) and fuzzy PID is proposed. This includes two parts: the cross-section forming characteristics of the tunnel and the intelligent cutting control strategy. The intelligent roadheader cutting control strategy consists of a CNN coal rock hardness dynamic perception module and a cutting arm swing speed fuzzy PID control module. An effective cutting path is proposed to make the cutting head cut coal and rock top to bottom along the planned path, aiming to improve the integrity of the cross-section and reduce the error in the tunneling direction. The CNN coal and hardness dynamic perception module is used to analyze the collected cutting motor current, cutting arm vibration acceleration, and rotary oil cylinder pressure data information to perceive the characteristics of coal and; the cutting arm swing speed fuzzy PID control module is used to process the perceived data for fuzzification and defuzzification, and to output the corresponding control parameter signals the electro-hydraulic proportional valve controls the flow and pressure of hydraulic oil according to the received signals, and then the valve-controlled hydraulic cylinder controls the swing speed of cutting arm, achieving the adaptive control of the cutting arm swing speed. The experimental results in the field show that when the roadheader cuts softer media and coal, the arm works at a high swing speed; when cutting complex rock strata, the swing speed decreases as the cutting signal increases, and the cutting signal varies between 0-1; when the roadheader cuts harder rock strata, the cutting load signal is close to 1, and the swing speed of the cutting arm is reduced 0.