面向电子束熔炼炉熔池温度智能控制的LSTM-PID复合策略OA

Electron beam melting(EBM)is a core technology for producing high-purity refractory metals,where precise control of the melt pool temperature directly determines the metallurgical quality of the product.Addressing challenges such as strong nonlinearity,significant inertial lag,and multi-source disturbances in the EBM process,traditional PID control exhibits limitations including large overshoot,slow regulation speed,and weak disturbance rejection.To overcome these issues,an intelligent composite strategy integrating long short-term memory(LSTM)networks and PID control is proposed to achieve precise feedforward compensation and feedback correction for electron gun power.This strategy leverages the sequential modeling and multi-step prediction capabilities of LSTM networks to predict melt pool temperature trends in real time and generate feedforward control signals,actively compensating for system lag and nonlinear characteristics.Simultaneously,PID feedback control ensures steady-state accuracy and robustness.Simulation validation based on a dynamic model of the EBM melt pool temperature demonstrates that,compared with traditional PID and fuzzy PID control,the proposed LSTM-PID composite controller reduces overshoot to 1.8%and shortens regulation time to 11.2 s during setpoint step tracking.Under feed rate step disturbances,it suppresses maximum dynamic deviation within 8℃,with a recovery time of only 15 s.This research significantly enhances the dynamic performance and disturbance rejection capability of temperature control,offering an effective paradigm that integrates data-driven and model-driven approaches for intelligent control of complex industrial processes.