In-mold condition-centered and explainable artificial intelligence-based (IMC-XAI) process optimization for injection molding

Abstract

This paper proposes a novel injection molding process optimization method based on the in-mold condition (IMC) and interpreted influence of in-mold condition features on part quality. In-mold condition is crucial for process optimization because it represents the actual process condition in the cavity where the polymer material is formed into the final part shape. Traditionally, the analysis of in-mold condition heavily depends on domain knowledge and insight, which introduces bias and inconsistency in in-mold condition-based process optimization. This study aims at developing an intelligent, objective, and robust process optimization method that concentrates on the highly influential in-mold condition features concerning part-quality, as interpreted by explainable artificial intelligence (XAI). In this in-mold condition-centered modeling approach, the input process parameters and final part quality were associated with the in-mold condition, with their corresponding relationship modeled by two machine learning (ML) models, respectively. The effect of in-mold condition on part quality was interpreted by applying XAI on the ML model that describes the relationship between in-mold condition and part quality. Features in the in-mold condition profiles that have high influence on part quality are given more weight in the search for diverse in-mold conditions that satisfy multiple part-quality objectives. A feasibility check has been implemented to identify, among those potential in-mold conditions, the optimal one that is physically feasible based on the ML model that governs the relationship between the process parameters and in-mold condition. The proposed method not only pinpoints better optimized process parameters than the conventional approach that omits the in-mold condition and only considers the direct relationship between the process parameters and part quality, but also reveals the possibility of further quality improvement. The optimization tool of the proposed method can be found on an online interactive platform https://imc-xai-injmold-optimization-4e26a3e3ef41.herokuapp.com/, which was created to facilitate further research based on the proposed approach. In addition to process optimization, this approach can effectively contribute to intelligent manufacturing management and Industry 4.0.