基于压缩感知的空间场传感网络优化方法研究

Study on optimization method of spatial field sensing network based on compressed sensing

为解决现有空间场传感网络中的传感节点密集、数据冗余和监测响应慢等问题,实现对空间场信息的实时监测,将压缩感知技术引入空间场传感网络优化中。本文提出了对历史数据进行学习的自适应稀疏正交字典的建立方法,通过对稀疏字典的分析改进了传统观测矩阵的建立方法,实现了空间场传感节点布局的最优化;通过聚类方法分析观测数据在字典上的投影系数,获得了空间场正常工作情况下的分布状态曲线,实现了对异常故障点的精准识别。最后,对二源温度场数据进行实验验证,结果表明本文所提出的压缩感知技术可以在完整保留温度场分布特征的前提下,有效降低感知数据的密集程度,并利用稀疏采样的数据,实现对温度场的实时监测和故障识别。该研究对温度场中传感节点高效布置给出了科学指导,避免了空间密集采样的资源浪费,同时也为如何在空间场采样数据欠缺的情况下,精确获取完整的温度场分布提供了新的技术思路。

In order to solve the problems of dense sensing nodes,data redundancy and slow monitoring response in the existing space field sensing network,and to realize real-time monitoring of space field information,compressed sensing technology is introduced into the optimization of space field sensing network.An adaptive sparse orthogonal dictionary learning method for historical data is proposed,and the traditional observation matrix establishment method is improved by analyzing the sparse dictionary to realize the optimization of the space field sensing layout.The projection coefficients of the observed data on the dictionary are analyzed by the clustering method to obtain the distribution state curve under the normal operation of the space field,and the accurate identification of abnormal fault points is realized.Finally,experimental validation of the two-source temperature field data is conducted,and the results show that the compressed sensing technique proposed in this paper can effectively reduce the density of the sensed data while retaining the distribution characteristics of the temperature field intact,and use the sparsely sampled data to achieve real-time monitoring and fault identification of the temperature field.The study gives scientific guidance on the efficient arrangement of sensing nodes in the temperature field,avoiding the waste of resources from dense spatial sampling,and also provides a new technical idea on how to accurately obtain the complete temperature field distribution in the absence of spatial field sampling data.