Research On Temperature Control Of Single Crystal Czochralski Furnace Based On Neural Network PID

With the increasingly stringent performance requirements of high-quality scintillation crystals in today’s high-precision cutting-edge fields such as nuclear medicine imaging,the design of the Czochralski furnace system and control algorithm for growing scintillation crystals also faces severe challenges.In the crystal growth stage of equal diameter,the change of the solid phase and melt composition in the crucible and the change of the length of the solid crystal will cause the heat loss in the Czochralski furnace to be in a dynamic state,and it is difficult to keep the ideal state unchanged.Therefore,it is necessary to find an Intelligent control algorithm to adjust heating power to achieve real-time temperature compensation.The classic PID(Propagation,Integration,Differentiation)control algorithm has a wide range of applications and good control effects,but it is difficult to exert its advantages in the face of a large inertia and large lag system such as a Czochralski furnace,and cannot meet the stringent control requirements brought by high-performance crystals.In order to meet the temperature control requirements of crystal growth by Czochralski method,based on the classical PID algorithm,this paper integrates the artificial neural network(ANN)algorithm for improvement,selects BP(back propagation)neural network,forms a BP-PID controller by online adjusting PID control parameters and sum,and combines the advantages of stable control of incremental PID controller,so as to realize fast and stable control of the temperature field of Czochralski furnace Stable compound control.In addition,particle swarm optimization(PSO)algorithm is added to the improved neural network PID composite control algorithm to adjust the control defects caused by the randomness of the initial weight of the neural network,so as to improve the simulation stability of the improved algorithm in temperature control.In order to verify the effectiveness of the improved algorithm in this paper,the simulation models of classical PID control system,BP-PID control system and PSOBP-PID control system are established by using Simulink simulation tool,and the control algorithms under these three simulation models are tested for step response and square wave response respectively.The step response results show that the neural network PID compound control algorithm can meet the timely response requirements of the temperature control system,and the PSO-BP-PID control algorithm can reduce the adjustment time by 500 s and significantly improve the fluctuation of the control curve.The square wave test results show that the neural network PID compound control algorithm can achieve the effect of quickly stabilizing the temperature field,the PSOBP-PID control algorithm can achieve the target control value within 250 s,and has a stable follow-up response effect in the rising and falling edges of the square wave signal.Finally,the effectiveness of this algorithm is verified by comparing with the simulation results in the literature.