A modified differential evolution algorithm for damage identification in submerged shell structures

Obtaining good estimates of structural parameters from observed data is a particularly challenging task owing to the complex (often multi-modal) likelihood functions that often accompany such problems. As a result, sophisticated optimization routines are typically required to produce maximum likelihood estimates of the desired parameters. Evolutionary algorithms comprise one such approach, whereby nature-inspired mutation and crossover operations allow the sensible exploration of even multi-modal functions, in search of a global maximum. The challenge, of course, is to balance broad coverage in parameter space with the speed required to obtain such estimates. This work focuses directly on this problem by proposing a modified version of the Differential Evolution algorithm. The idea is to adjust both mutation and cross-over rates, during the optimization, in a manner that increases the convergence rate to the desired solution. Performance is demonstrated on the challenging problem of identifying imperfections in submerged shell structures.