A methodological approach for model validation in faster than real-time simulation

When successfully dealing with time constraints, faster than real-time simulation (FRTS) improves process control capabilities through providing short-term predictions. Validation of a simulation model is accomplished through comparing system observations and model data corresponding to the same time points. A methodological approach is necessary for the reatization of this comparison, considering the time-dynamic system behaviour and the potential deviations between the model and the system. A seven-step method is proposed, emphasizing the following issues: determining when predictions should be considered as valid, accomplishing validation on the basis of the available model and system data, considering that not all deviations between the model and the system are of equal significance, and indicating potential system changes, so that the model can be modified in real time. Computer networks are used as an example domain, due to their multi-entity structure and time-dynamic behaviour, offering excellent test cases for evaluating the proposed method. Experimental FRTS results from the application of the proposed method to the network domain are also presented. (C) 2002 Elsevier Science B.V. All rights reserved.