Pretreatment of Data Fusion in FBG Temperature Sensing Network

Abstract

After analyzing noise and disturbance in FBG temperature sensing network, three different kinds of statistical methods are processed in experiment as pretreatment to deal with inaccuracy of data fusion results caused by outliers. After replacing outliers with statistical estimated value of three methods respectively, weight minimum mean square error (WMMSE) algorithm is operated to fuse majorized data. Pretreatment makes full use of sensors’ data, meanwhile, accuracy and robustness of measurement system are enhanced. Variances of the fusion results indicate that pretreatment can improve fusion results significantly. Fusion variance without pretreatment is 2.9762, larger than mean value method’ variance 1.9525, median method’s 1.3341 and trimmed mean method’s 1.6016.