A BP-Smith combined temperature control method for thin film preparation processing

Thin film preparation methods are receiving increasing attention because of their wide range of applications in semiconductor devices, optoelectronics, flat panel displays, solar cells, sensors, micromechanical systems, and other fields, and they provide essential technologies for the development and application of various advanced materials. This paper proposes a method combining BP neural network control and Smith prediction to solve the nonlinear thermal regulation control problem and consider the coupling effect between temperature regions in a multi-region setting. The BP neural network is used to control the dynamic neural network to identify the dynamic model of the temperature area and realize online learning of neural network weights. The system uses Smith prediction to solve the delay problem and ensure system performance. Our controlled equipment is a tank for the boron diffusion process. The control goal of this paper is to track the target temperature stably and accurately through the proposed method. The temperature control accuracy is within ±1 °C tolerance of set point in steady state. The temperature control strategy proposed here also adds an error factor for temperature coupling in the neural network part, and the control results are better able to meet the temperature control requirements of the actual process. This method provides innovative insights into and effective solutions for temperature control during thin film preparation. It reduces the amount of overshoot, saves a lot of power and manpower for model change temperature control, and is a highly adaptive model change control method. This paper begins by theoretically analyzing the advantages of neural networks and Smith predictive control. Secondly, thermal simulations are performed to analyzes the coupling conditions between the temperature zones. Finally, experimental tests evaluate the steady-state and dynamic performances of the control strategy and verify the intended advantages of the proposed control method.