Uncertain Fiber Bragg Grating Sensor Data Analysis Based on Sparse Bayesian Learning

In recent years, a number of indirect data collection methodologies have led to the proliferation of uncertain data. This paper presents a sparse Bayesian learning mechanism for uncertain fiber bragg grating sensors data. The collected sensor data are uncertain because of bad working environment and modeling methods. Research of sensor data are much more complex because of the additional challenges of representing their probabilistic information. Sparse Bayesian learning is the basis of relevance vector machine, and derives the data in Bayesian frame which is able to describes the probability characteristics of uncertainty. As a case study, the noisy strain sensor timeseries from a highway and railway bridge is used for regression and uncertainty analysis. The simulation results illustrate effectiveness of the presented sparse Bayesian learning mechanism that is superior to support vector machine and least square method. The proposed Bayesian analysis method reduces uncertainty in sensor data.