基于自适应加权的多传感器实时数据特征值提取  

Real-time data eigenvalue extraction of multi-sensor based on adaptive weighting

作　　者：全恩懋 秦小平 许宏科 孙中洋 QUAN Enmao;QIN Xiaoping;XU Hongke;SUN Zhongyang(School of Municipal Engineering,Chongqing Water Resources and Electric Engineering College,Chongqing 402160,P.R.China;School of Electronics and Control Engineering,Chang’an Unversity,Xi’an 710064,P.R.China;China Merchants Chongqing Highway Engineering Testing Center Co.,LTD,Chongqing 400067,P.R.China)

机构地区：[1]重庆水利电力职业技术学院市政工程学院,重庆402160 [2]长安大学电子与控制工程学院,西安710064 [3]招商局重庆公路工程检测中心有限公司,重庆400067

出　　处：《重庆邮电大学学报（自然科学版）》2023年第2期368-376,共9页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家重点研发计划项目(2020YFF0217801);新疆维吾尔自治区自然科学基金(2019D01A36)。

摘　　要：为实现监测数据的特征值提取,对传感器数据的预处理、时间维及空间维融合方法开展了研究。建立了实时数据融合模型,提出了基于3σ-grubbs检验的异常数据预处理方法,兼顾了异常数据剔除的速度与精度,能很好地消除疏失误差;对单个传感器数据采用分批估计原理进行融合,得到了特征估计值,实现了数据在时间维上的融合;通过对多个传感器的特征估计值采用自适应加权方法进行赋权,实现了数据在空间上的融合,并提出了考虑传感器精度的算法修正。实例计算表明,数据经3σ-grubbs方法处理后方差减小了20%~54%,与传统的算术平均滤波方法相比,分批估计自适应加权融合算法的数据融合方差明显更小,考虑传感器精度后的融合结果更接近高精度传感器值,特征值提取结果更加准确、可靠。In order to achieve the extraction of eigenvalues from monitoring data,this paper investigates the pre-processing of sensor data and the fusion methods in time and space dimensions.A real-time data fusion model is established,and an anomaly pre-processing method based on the 3σ-grubbs test is proposed,which considers the speed and accuracy of anomaly rejection and can eliminate spurious errors.The data are weighted by the adaptive weighting method to achieve the spatial fusion of the data,and an algorithmic correction is proposed to consider the accuracy of the sensors.The example calculations show that the variance of the data is reduced by 20%~54%after processing by the 3σ-grubbs method.Compared with the traditional arithmetic average filtering method,the variance of the data fusion by the batch estimation adaptive weighting fusion algorithm is significantly smaller,and the fusion results are closer to the high-precision sensor values after the sensor accuracy is considered,and the feature value extraction results are more accurate and reliable.

关 键 词：监测数据 多传感器 疏失误差 特征值融合 分批估计 自适应加权 

分 类 号：TN919[电子电信—通信与信息系统] TP391.7[电子电信—信息与通信工程]