Sheet-metal press line parameter tuning using a combined DIRECT and Nelder-Mead algorithm
Paper in proceedings, 2011
It is a great challenge to obtain an efficient algorithm for global optimisation of nonlinear, nonconvex and high dimensional objective functions. This paper shows how the combination of DIRECT and Nelder-Mead algorithms can improve the efficiency in the parameter tuning of a sheet-metal press line. A combined optimisation algorithm is proposed that determines and utilises all local optimal points from DIRECT algorithm as Nelder-Mead starting points. To reduce the total optimisation time, all Nelder-Mead optimisations can be executed in parallel. Additionally, a Collision Inspection Method is implemented in the simulation model to reduce the evaluation time. Altogether, this results in an industrially useful parameter tuning method. Improvements of an increased production rate of 7% and 40% smoother robot motions have been achieved.