Asynchronous optimization techniques for distributed computing applications

Simulation-based optimization of manufacturing systems has emerged into a widely accepted decision support technique for industrial applications. Due to the computational complexity of these applications, simulation-based optimization is typically implemented in software systems that make use of distributed computing-units in order to reach an acceptable performance. However, even then dynamical methods, i.e. methods that are not bound to a fixed number of iterations are desired in order to further increase the performance of the optimization process. This paper addresses possibilities for employing population-based optimization techniques in an asynchronous way in order to react to computational node failures, limited computing resources and unbalanced simulation times. We demonstrate our approach by applying it to the optimization of set-up parameters for virtual tooling machines within an underlying research project of the leading edge cluster 'ITS OWL -- intelligent technical systems.