利用复介电常数和BP人工神经网络进行浓度预测的研究

制备了一定含水率不同浓度的CuCl2的PVA样品,用微波矢量网络分析仪和微波传感器测量S11参数,计算得到相对复介电常数。以样品相对复介电常数的实部、虚部及对应频率作为输入,以CuCl2溶液的浓度作为输出,建立BP人工神经网络模型。用训练样本集对网络训练后,检验样本的预测结果与实际值最大误差为0.97%。结果表明,利用复介电常数和BP人工神经网络进行浓度预测是一种很好的方法,进而为环境监测提供了方法依据。

Using Complex Permittivity and BP Artificial Neural Networks for the Study of Forecasting Concentration

The polyvinyl alcohol blend samples containing CuCl2 solution with different concentration and same water content were prepared. Parameter S11 was measured with microwave sensor and vector network analyzer, which was used to calculate relative complex permittivity. BP artificial neural network model was established with frequency, real and imaginary part of relative complex permittivity as inputs and concentration of CuCl2 solution as output. The model was trained with training sample aggregation. The maximum error between the forecasted and real value was 0.97%. All of these showed that it was a good method to forecast solution concentration using relative complex permittivity and BP artificial neural network model. It provided basis and method for environment contamination detecting.