Improving the performance of the partitioned QN-ILS procedure for fluid-structure interaction problems

AbstractPerformance of QN-ILS can be improved with data from previous time steps.When this is done, filtering might be necessary.Filtering is a compromise, as discarding data might impair the convergence speed.We introduce two new ways of filtering. Better results are obtained. The Quasi-Newton Inverse Least Squares method has become a popular method to solve partitioned interaction problems. Its performance can be enhanced by using information from previous time-steps if care is taken of the possible ill-conditioning that results. To enhance the stability, filtering has been used. In this paper we show that a relatively minor modification to the filtering technique can substantially reduce the required number of iterations.