Dispatching rules are widely used for scheduling of wafer fabrication facilities, and simulation is a popular tool for evaluation of rules. A main weakness of simulation is the long computation time. In this paper, we aim at simplifying the models of fabrication facilities to reduce simulation time and meanwhile keep the ability of evaluating rules correctly. We propose a metric to assess how well a simplified model evaluates rules and formulate the model simplification problem as a multiobjective optimization problem. We also develop an evolutionary algorithm to solve the formulated problem. Experimental results show that simplified models generated by the proposed algorithm can save half of simulation time and select high-performance rules.