Abstract
Aiming at the problems of high safety risks, economic costs, and inefficiency in experimental research on boom-type roadheaders, this study proposes a virtual-real fusion experimental system for the cutting module. This system incorporates functions including digital modeling of coal-rock, simulation of mechanical properties of the cutting unit, and integration of virtual and physical experiments. To address the challenge of obtaining cutting tooth loads at coal-rock interfaces, a discretized digital coal-rock volume construction method is proposed. For rapid mechanical performance simulation of the cutting unit, a chain-type digital mapping body construction method is developed. Through deep learning, numerical simulation, and digital twin technologies, a virtual-real fusion platform was established, enabling virtual experiments to dominate the calibration of physical experiments. The system is capable of simulating pose variation and vibration trends of the entire machine during cutting. The minimum average error for stress at the cylinder base is 4.44 MPa, with a virtual-real system connection period under 100 ms. Based on this system, a reinforcement learning training environment was developed. Using Deep Deterministic Policy Gradient (DDPG), the control of the cutting unit was optimized to achieve low-stress state cutting, verifying the system's feasibility.