An expert system for cold forging process design based on a depth-first search

Abstract

Due to the non-deterministic nature of process sequence design for multi-stage cold forging, various process designs are available depending on the initial billet geometry and the order of basic processes such as forward/backward extrusion, upsetting and trimming process. Therefore, various process sequences should be determined and compared to obtain an optimal solution. For this purpose, a depth-first search, a searching technique used in artificial intelligence, has been introduced in developing an expert system for multi-stage cold forging process design. As a result, process designers can select the optimal process sequence from the searched feasible solutions by estimating the values of evaluation functions that are introduced to represent the important design characteristics. In the present investigation, the distributions of the global effective strains in the final product and the forming loads required at each forging stage were selected to be controlled. In general, a more realistic process sequence should be determined by taking into account manufacturing conditions such as the number of forming stages, the forming loads, the shearing diameter of the coil, the open upsetting diameter, and the knock-out lengths of the die and punch. In this paper, a methodology of applying the searching technique for process sequence design is discussed, and the flexibility of the introduced searching technique is evaluated by generating design examples of a shaft part, a wrench and hexagonal bolts of AISI 1045.