大容积电烤箱内传热过程的反向传播神经网络控制算法

Serious problem on the heating uniformity exist in the large-volume electric oven, which limits its extensive application in commercial and household fields. The commonly applied Proportion Integration Differentiation (PID) algorithm has concerns of long relaxation time and poor temperature control accuracy. This study involves a self-coding backpropagation neural network (BPNN) control strategy, aiming to improve the heating efficiency, temperature control accuracy and uniformity. The local velocity and temperature measurement as well as the egg-tart visualization methods are utilized to assess the control sensitivity of fan speed, power of airflow heating rods, power of radiation heating rods, and exhaust flowrate. Experimental result shows that, following effective data training and robustness enhancement, the BPNN control strategy can significantly reduce the prediction errors. In comparison to the PID strategy, the overheating is reduced by up to 6 °C. Meanwhile, the maximum temperature difference decreases from 54% to 36%. Accordingly, the velocity difference drops from 71. 4% to 39%, and the relaxation time shorts from 230 seconds to 100 seconds. It is indicated that the BPNN strategy can provide much quicker, more precise and uniform temperature control in the large-volume electric oven.