Automated manufacturing processes such as automotive tandem press lines include time dependent complex control functions. All motions and critical interactions between moving parts must be synchronised to avoid collisions and reach high production rate. It is even for a skilled operator hard to optimise these processes on-line. Therefore, a hardware-in-the-loop simulation including real industrial control systems and its control code establish an essential tool for optimisation. Additionally, an efficient optimisation algorithm is required to reach a useful simulation-based optimisation method. This paper proposes a new optimisation algorithm starting with the Lipschitzian algorithm DIRECT as global search and then switches over to the Nelder-Mead simplex algorithm for local convergence. During the switch over, the new algorithm determines all local candidates of the set of points evaluated by DIRECT and starts multiple Nelder-Mead local searches in each of these. An optimisation study for an automotive press line shows that the proposed algorithm combines the benefits of the Lipschitzian and the simplex algorithms in an efficient way. The importance of multiple local searches from all local candidates found is also shown in the study. Based on the same number of function evaluations, it is also shown that this algorithm reaches improved press line performances compared to the stochastic differential evolution algorithm.