Estimation of Incident and Reflected Waves Using Simulated Annealing

Abstract

A new time-domain method for separating incident and reflected waves in wave flumes using any number of wave gauges is presented. The method (named LASA) is based on a local approximation model considering linear and Stokes II nonlinear components, and a simulated annealing algorithm to calculate the parameters of the local approximation model. The LASA method is directly applicable to experiments in wave flumes with an arbitrary number of wave gauges, but the LASA concept may be extended to laboratory and field three-dimensional experiments by selecting the appropriate local approximation model. The LASA method has been compared to the two-point and Kimura's method and has proved to be very robust. The application of the LASA method to physical experiments using nonstationary regular and random waves generated estimations of incident and reflected waves with relative mean squared errors lower than 3% for regular waves and ranging from 4 to 16% depending on the spectral shape.