An efficient method to determine strain profiles on FBGs by using differential evolution and GPU

This paper shows the application of the Differential Evolution algorithm to the recovery of the deformation profile applied to a fiber Bragg grating sensor (FBG). The method uses only the sensor reflectivity, without the need of phase information, and has been shown to be highly parallelizable. The method was specially implemented to run the complex computations required by the fitness function using a Graphical Processing Unit (GPU). This provided an enhancement of 3 orders of magnitude in the computation time when compared to similar methods shown in literature. Three experiments were performed to evaluate the computational performance of the DE algorithm, and its convergence of the solution to the target deformation profile. Overall, the method is very promising and it can be applied to problems that require a fast response, such as the online monitoring of FBG sensors.