An Explainable AI Framework for Data-Driven Quality Prediction in Injection Molding Processes: A Review

Injection molding is one of the most widely used manufacturing processes for producing plastic products in industries such as automotive, healthcare, consumer goods, and electronics. The quality of molded parts depends on several process parameters, including temperature, pressure, and injection speed, which interact in complex ways. With the rapid adoption of Industry 4.0 technologies, researchers have increasingly explored machine learning techniques to predict product quality, detect defects, monitor processes, and optimize manufacturing performance. While these models often achieve high prediction accuracy, their lack of transparency makes it difficult for engineers to understand how decisions are made and to confidently apply them in industrial settings.

To address this issue, Explainable Artificial Intelligence (XAI) is emerging as a promising solution to enhance the interpretability of machine learning models. XAI methods help identify the factors influencing model predictions and provide insights into the decision-making process. Recent works published in the period 2023-2025 show that SHAP is the most popular explanation technique due to its efficiency in estimating feature contributions. The results show that SHAP is the most used XAI method, present in more than 60% of the reviewed studies, and that real-time explainability in injection molding is still an unexplored area.

 However, the application of real-time explainability in injection molding remains limited. This review analyzes the integration of XAI and machine learning in injection molding, summarizes the most frequently used predictive and explanation methods, discusses existing challenges, and highlights future research opportunities for developing trustworthy and practical intelligent manufacturing systems.