Accelerated random search method for dynamic FE model updating

Purpose – This paper seeks to present a new solution algorithm for updating of finite element models in structural dynamics. A random search method is applied to improving the correlation between the numerical simulation and the measured experimental data. Design/methodology/approach – Dynamic finite element model updating may be considered as an optimization process. It is solved using modified accelerated random search (MARS) algorithm. The effectiveness of the approach is first tested on benchmark problems. Next, several objective function formulations for dynamic model updating in modal and frequency domains are investigated for numerically simulated vibrating beam. Finally, the algorithm is applied to a real beam-like structure using measured modal data. Findings – The MARS algorithm is able to provide very good results in a reduced time even for hard optimization problems. It behaves very well also for the FE dynamic model updating, highly coupled problems. The efficient updating criterion has been proposed and the approach has been validated experimentally. Research limitations/implications – The method is supposed to be time consuming for large size or complicated objective function problems but the choice of optimization parameters can accelerate the convergence. Practical implications – The MARS algorithm can be applied to model updating in civil and mechanical engineering. Originality/value – This paper is the first to apply the MARS algorithm to the problem of FE model updating in dynamics and enables one to obtain very good results. Efficient criteria for model updating have been propose