Abstract
A genetic algorithm (GA) is used to optimize the hot isostatic pressing (HIPing) process for beryllium powder. The GA evaluates a HIPing model with different processing schedules in an effort to minimize temperature, pressure, processing time, ramp rates, grain growth, and distance to target relative density. It is shown that this is a constrained, multiobjective, noisy, optimization problem to which the GA is able to evolve a large number of viable solutions. However, for the GA to work in such a large multidimensional search space, it is suggested that the constraints be treated as objectives and then penalize the Pareto ranking for each constraint violated. This approach is necessary because a large-dimensional objective space naturally results in most members being Pareto rank 1.