Impact localization takes an important part in the structural health monitoring (SHM) process to ensure the structure safety, especially in the field of aerospace and civil infrastructure. Fiber Bragg Grating (FBG) sensors have bright prospect in the SHM system, thus it is essential to research the algorithms which process the FBG-acquired signal. This paper proposes a multi-class pattern recognition algorithm to process signals obtained by FBG sensors in CFRP composite SHM system, and to accomplish the goal of impact recognition and localization. Considering the limited training set in the actual monitoring system, we specifically select the SVM model, and then extend it to the multi-class classification through one-vs-one method. Meanwhile, since the relatively low sampling rate of FBG sensors, we choose energy of signal as the input features of our impact localization algorithm. Finally, we implement the algorithm to an actual impact localization monitoring system for the composite plate to verify the accuracy, and the proposed algorithm was affirmed on the cross validation set. Our research shows that the deviation of the proposed algorithm in actual application is acceptable.
doi: 10.12783/SHM2015/212