Hierarchical modeling and multiresolution simulation

As systems become more complex, the need to explicitly account for uncertainty during modeling and simulation grows. The interactions between assumptions made in modeling different subsystems may greatly affect system behavior. Unless these assumptions are quantified and included in the simulation, results can be misleading or even completely wrong. Piecewise linear (PL) modeling is proposed as a method for quantifying the uncertainty in a model. With PL models, sets of models with varying amounts of uncertainty are easily developed. Robust simulation is then used to account for uncertainty during analysis. Also, robust simulation allows dynamic selection of models. Through the use of PL modeling and robust simulation, unexpected model interaction can be predicted. These techniques are demonstrated on three simple illustrative examples. A model library is developed for a saturation. This saturation is then used in a feedback system, and the simulation results of various models are examined. A final example demonstrates the benefits of changing model accuracy during simulation