基于BP神经网络的温湿度环境下蒸发量补偿研究

Research on Evaporation Compensation under Temperature and Humidity Conditions Based on BP Neural Network

摘要: 温湿度环境下微流量液体称重的准确性受到液体蒸发的极大影响。本文通过搭建称重实验装置，分析了环境温度、相对湿度与蒸发速率的关系。利用反向传播(BP)神经网络构建了蒸发速率预测模型，并与改进的物理蒸发模型进行比较。结果表明，BP神经网络模型和改进的物理蒸发模型的决定系数分别为0.768和0.7405，预测平均相对误差分别为9.08%和13.19%，说明BP神经网络的拟合效果和预测能力都要优于改进的物理蒸发模型。应用BP神经网络构建蒸发模型能够实现对蒸发速率的预测。

Abstract: The accuracy of micro-flow liquid weighing in a temperature and humidity environment is greatly affected by liquid evaporation. In this paper, a weighing experiment device was built to analyze the relationship between environmental temperature, relative humidity and evaporation rate. A backpropagation (BP) neural network was used to construct an evaporation rate prediction model and compared with an improved physical evaporation model. The results show that the determination coefficients of the BP neural network model and the improved physical evaporation model are 0.768 and 0.7405, respectively, and the average relative prediction errors are 9.08% and 13.19% respectively. This indicates that the fitting effect and prediction ability of the BP neural network are superior to those of the improved physical evaporation model. The application of BP neural network to construct an evaporation model can achieve the prediction of evaporation rate.