神经网络遗传算法在粉体流量测量的应用

Neural Network Flowrate Measurement Based On Genetic Algorithms and Its Application to Powders

利用浓度和速度电容式传感器分别测量粉体的浓度和速度 ,结合阀门开度、喷吹罐压力、温度等参数构造神经网络粉体流量测量模型 ,考虑 BP算法训练神经网络测量模型时收敛速度慢、动态特性不够理想等不足 ,用遗传算法来优化神经网络测量模型的参数 ,以提高测量系统的精度。在现场与电子秤比对 ,最大满量程误差小于 4.2 %。

Capacitance sensors are used to measure volumetric concentration and velocity of powders, combined with pressure of puffing tank, valve on off, temperature parameters, a kind of neural network (NN) flowrate measurement model based on genetic Algorithms (GA) is presented. In order to overcome the defects such as slow convergent speed and unsatisfied dynamic characteristic when backpropagation (BP) is used to train the neural network measurement model, the method using GA to train parameters of NN is proposed to improve the performance of measurement model. Spot experiments show this method can reduce the errors of measurement. The whole accuracy is less than 4 2% according to the electronic-weighing device in work site. It has great engineering application value.