Sequential optimization of the injection molding gate locations using parallel efficient global optimization

Injection molding (IM) is a versatile manufacturing process capable of rapid prototyping and mass-producing high-quality polymer parts. The present study mainly investigates the challenge of designing multiple molding gates on the complex arbitrary part surface in 3D. Currently, this problem is a challenge in mold design and engineering experience still plays an important role in designing the molding gates. To reduce the human intervention in the design process, the present study proposed a novel methodology with the following major steps: 1) using Poisson disk sampling (PDS) to preselect candidate gate locations automatically within the suitable gating region specified by designers; 2) using a space-filling initialization strategy and efficient global optimization to find the optimal gate locations. In the present setting, the molding gate design problem is formalized as a discrete optimization problem. The PDS is employed to construct the discrete solution space and EGO is used to efficiently search through a large solution space for the best design. To further promote optimization efficiency, a parallel implementation of EGO is also proposed. The effectiveness of the proposed methods is validated in two design cases. The results demonstrate the proposed EGO and Parallel EGO method is superior that the Genetic Algorithm (GA) and Surrogate Optimization (SO). Moreover, the proposed Parallel EGO converges faster than all other alternatives.