Industrial manufacturing processes involve a series of operations that transform a workpiece into a useful finished product characterized by acceptable shape and mechanical properties. A typical sequence of thermomechanical operations consists of multiple-stage hot and/or cold deformation processes interspersed with suitable heat treatment, material removal, and inspection processes. The shape of the product is achieved through the deformation and material removal processes, while the properties of the product are, in general, dependent upon the entire thermomechanical processing history. The challenge in the design of a manufacturing process is to optimize the entire processing sequence in order to achieve the best balance of manufacturing and material costs, delivery schedules, and shape and mechanical properties of the final product.
Trial and error methods have long been used to select process parameters, including processing temperatures, machine speeds, and die geometries. These methods generally result in, at best, a working design, with no attempt at optimization. Recently, research has moved to the application of conventional optimization techniques to individual manufacturing processes. A major concern from the systems engineering point of view is that individual manufacturing operations is that a single operation could be over-optimized at the expense of the cost or performance of the entire enterprise.
Global optimization of an entire manufacturing enterprise involves searching an extremely large candidate design space. Even with extremely efficient searching algorithms, several thousand candidate designs must be evaluated must be evaluated during an optimization search. Thus, traditional forging process models such as finite element analysis, which require 30 minutes or more for a single design evaluation, are infeasible for global optimization.
In the present research, a series of thermomechanical process models based on fast, simplified techniques such as upper bound and slab analysis have been merge with new volume constancy techniques to give extremely fast objective function evaluation. Evaluations are performed in less than a second, and the results compare reasonably well with finite element calculations.