Abstract: Aiming at the disadvantages of optical fiber SPR open-loop system, the premature convergence problem of particle swarm optimization (PSO) with global search is improved, and an improved cooperative particle swarm optimization (ICPSO) algorithm with dynamic information adjustment and controllable speed is proposed. By introducing the optimal information of subgroups into the iterative equation of the flight-state controlling of particle, the particles′ diversity is well maintained, and the prematureconvergence of particles in optimization flight can be effectively avoided for the proposed algorithm. Furthermore, the algorithm is used as the training algorithm of BP neural network, and a more optimized ICPSO-BP neural network is established. Finally, the ICPSO-BP neural network is used to identify and compensate the internal nonlinear model of optical fiber SPR open-loop system, and the compensation models of single-input, double-input and three-input ICPSO-BP neural network are established respectively. The simulation results show that the new algorithm has good performance in speed and accuracy test, thus ensuring the good linearity test effect of SPR, and laying a foundation for the further application of optical fiber SPR sensor.

Key words: fiber optics, optical-fiber sensor, surface plasmon resonance effect, particle swarm optimization; neural network, open-loop system