If a suggested design change is approved by the user, the user must specify how large a change he will allow. The expert system will then determine the optimal magnitude of the change. The system uses a sampling method, testing four models corresponding to the original (no change), maximum change (based on user input), and two models with design changes equidistant from the two extremes. The performance of each sample model is determined and compared. The design with the highest performance is accepted, and the model corresponding to this design is used in the next iteration. For example, if the system suggests translating the south windows towards east, and the user indicates that he will allow a maximum translation of 4.5 ft, the system will then create and simulate models where the south windows have been translated 0, 1.5, 3, and 4.5 ft. Figure 7 indicates the percentage in the goal range for these four sample models, and the model corresponding to a translation of 3 ft has the highest performance. For situations with multiple sensor planes, the performance is averaged at each point.