As an introduction the recursive and iterative process of problem solving by simulation is dealt with and the possibility of enhancing its effectivity by utilizing a knowledge base and an inference engine are described. The necessity of modifying both the experimental conditions and the model itself — with respect to model structure and parameters — are discussed. The principle of demon controlled simulation is introduced. The advantages of the event driven activation of sleeping procedures with respect to run time efficiency as well as the advantages achieved by the distributed control are outlined. Finally, the methodology is illustrated on two examples using the CASSANDRA simulation system where the principles described have been implemented. The first example is a producer–consumer model where the optimal number of consumers is determined by structural changes of the model. In the second example the demons determine the throughput of two production systems by parametrical changes in the experiment control, while by means of one demon — used as an inherent part of the model — an adaptive model is built.
