摘要 |
PROBLEM TO BE SOLVED: To provide an efficient method for searching an optimal value by reducing the possibility of selecting a local optimal point. SOLUTION: This optimal parameter search method calculates an error between a model output and an output of a real machine (S7), and if the error is not within a threshold (S8), Taguchi method is applied to calculate a gradient vector from a factor effect (for example, error gradient at two search points) of each calculated parameter (S9). A correction amount of the parameter is calculated on the basis of the calculated gradient vector (S10). Since the correction amount of the parameter is determined from a gradient of the error between the two search points, the possibility of converging on a local optimal point is lowered to search an optimal value. COPYRIGHT: (C)2008,JPO&INPIT
|