The optimal scheduling of complex FMS (Flexible Manufacturing Systems) requires simulation methods for gaining reliable process predictions of different scheduling scenarios. This prediction strongly depends on the accuracy of its input data, i.e. the FMS model and its parameters.
We already described the method of a process synchronized simulation at FAIM'95.[l] Based on an online process monitor it. continously compares the progresses of the simulated and the real process. The models process state can be synchronized deviation dependent.
This paper introduces the approach of a learnable simulation system based on this method. It differs between state data variable during the process and constant model parameters like operational times for each of the process steps. The latter describe the process characteristics of the FMS. Those model parameters are often extense uncertain determinable in advance. By learning algorithms it should be possible to conclude model parameters from the monitored operational data allowing for major improvements of the accuracy of the simulation model. This method will be presented based on the example of a real world FMS.