基于动态神经网络的喷雾耙供气温度数据重构

Data Reconstruction of Gas Temperature in Spay Bar Based on Dynamic Neural Network

结冰风洞试验过程中,喷雾耙内气体温度传感器出现故障会导致数据缺失。结冰风洞喷雾供气系统是高耦合度的热力系统,喷雾耙供气温度受多种因素影响,很难用准确的物理模型描述。文中采用主成分分析(PCA)方法消除喷雾耙供气温度相关影响因素变量间的数据冗余,建立以喷雾耙入口温度、喷雾耙供气压力、风洞内部环境温度及喷雾耙供气温度延迟量为输入和喷雾耙供气温度为输出的动态神经网络异常数据重构模型,提出了序列排序与并行排序的数据训练方法,应用训练后的模型对喷雾耙供气温度单点异常与长期异常数据进行重构,并与真实值作对比。结果表明,采用并行排序的动态神经网络能够准确地实现喷雾耙供气温度数据重构,具有较好的泛化能力。

During the ice wind tunnel test,the failure of the gas temperature sensor in spray bar can lead to data loss.The gas supply system of ice wind tunnel is a thermal system with high coupling degree,and the temperature of gas in the spray bar is affected by many factors,so it is difficult to describe it accurately by physical model.In this paper,principal component analysis(PCA)was used to eliminate data redundancy between variables related to the gas temperature in spray bar,a dynamic neural network abnormal data reconstruction model was established and a data training method with sequence sort and parallel sort was proposed.The input of the model was the spray bar inlet temperature,the gas pressure,the wind tunnel ambient temperature and the delay of gas temperature in spray bar,and the output of the model was the gas temperature in spray bar.The trained model was applied to reconstruct the gas temperature in spray bar with single abnormal data and long-term abnormal data,and compare it with the real value.The results show that the dynamic neural network with parallel sort can reconstruct the gas temperature accurately and has a good generalization ability.