Temperature prediction with FBG sensor based on relevance vector machine

With the wide application areas and increase industrial benefit of fiber bragg grating (FBG) sensors, the main concerns of how to build the realtime, efficient and stable monitoring system based on fiber sensing have been raised, which can provide early warning and decrease damage loss. Furthermore, the mass data include noises and uncertain information from the monitoring system. Evaluation of monitoring system dealing with noisy and uncertain data becomes the hot spot and difficulty. When FBG sensor network is employed to evaluate different structure health condition, the structural distortion and variation of stress are heavily associated with temperature, so it is essential to get higher temperature prediction accuracy using FBG sensors. Relevance vector machine is applied to regression calculation between wavelength of FBG and temperature in this paper, and the simulation results illustrate effectiveness of the presented method compared with least square method.