This paper develops an inverse architecture for the adaptive network fuzzy inference system (ANFIS) to preform the hole profile in the sheet square hole bore-expanding process. At first, a feasible hole profile, which will be deformed to be a nearby square hole in the bore-expanding process, is obtained by using a few trial-and-errors. Then, several offset lines, parallel to the feasible hole profile, are designed and simulated by FEM in the bore-expanding process to establish the basic database for ANFIS. Using this database, an exact preform of hole profile can be inversely predicted by hybrid-learning cycles of ANFIS. As a verification of this system, the square hole bore-expanding experiment was conducted to expand a blank having initial hole predicted by ANFIS. The deformed square hole of the experiments and FEM simulated results are compared. It is found that the simulated square hole shows good coincidence with the experiment for the desired square hole after forming. From this investigation, ANFIS is proved to be able to supply a useful optimal soft computing approach in the sheet metal forming category.
Journal of Materials Processing Technology 173(2), pp.136-144