Machine Learning Prediction and Experimental Validation of Externally Solidified Crystal Formation in Aluminum Die Casting Process

Abstract Externally solidified crystals (ESCs) can easily form during high pressure die casting (HPDC) of aluminum alloys if the process is not properly controlled, leading to significant reduction of casting quality and performance. This study explores seven machine learning (ML) models, including decision tree, random forest, logistic regression, neural network, K-nearest neighbors (KNN), support vector machine (SVM), and naïve Bayes classifier. The random forest and classification tree models showed the highest accuracy of 95% in predicting ESCs formation. The neural network, KNN, SVM, and logistic regression models all presented a good accuracy over 89%, while the naïve Bayes classifier model provided the lowest accuracy of 85%. Moreover, the importance of HPDC process parameters in ESCs formation was ranked, including superheat, melt temperature, die and shot sleeve temperatures, vacuum pressure level, intensification pressure level, and shot profile including both slow and fast speeds. Based on the ML prediction results, top four factors determining the ESCs formation are ranked as follows: Melt temperature > Superheat > Shot sleeve temperature > Fast shot speed. Prediction maps were also generated to predict the ESCs formation under different processing parameters. Two die casting trials using an Al-9Si-Mg alloy with various melt temperatures (720 °C and 680 °C) successfully validated the prediction that a lower melt temperature leads to a high probability of ESCs formation. These findings demonstrate how machine learning may be used to predict defects in HPDC products, offering a path to optimize the processing parameters and reduce casting defects.