A swarm intelligence approach for regression analysis of surface plasmon resonance curves in Otto chips

The use of the Otto configuration, as an alternative to Kretschmann’s, in the constriction of Surface Plasmon Resonance (SPR) sensors is an underdeveloped research area. Recently, a version of an Otto based device, baptized as Otto chip, was manufactured. Regression analysis procedures can be used to help the chip characterization by adjusting parameters of the model reflectance curve. However, as in any classical regression procedure, the initial guess must be close enough to the final solution to avoid convergence to a local minimum. An alternative approach to the classical regression procedure is the use of computational techniques inspired on swarms. Swarm intelligence, as this area of computer science is known, has successfully been used by engineers in various optimization problems. One prominent algorithm is the Particle Swarm optimization (PSO) that stands out for its low computational cost, simplicity of implementation and high efficiency on finding global optimum solutions. This paper describes the use of PSO in regression analysis of experimental SPR curves. It was shown that the PSO technique yields better results when compared with classical regression analysis methods.