Enhancing Additive Manufacturing with Deep Learning: Predictive Modeling and Process Optimization Using the NIST Additive Manufacturing Material Database

Abstract

Bullying is an important issue in higher education, with heavy consequences for student's mental health, well-This study presents a predictive modeling approach to optimize additive manufacturing (AM) processes using a synthetic dataset based on the NIST Additive Manufacturing Material Database. A combination of regression and classification models were employed to evaluate key material properties and process parameters, aiming to improve AM output quality and reduce defect rates. Data preprocessing included normalization and correlation analysis to identify high-influence features, which informed feature selection for modeling. A Linear Regression model effectively predicted material behavior, achieving low Mean Squared Error (MSE) across training, validation, and test sets. A classification model was also developed to predict defect rates, yielding high accuracy, precision, and recall. Performance metrics, including a confusion matrix and ROC curve, underscored the model’s high specificity and sensitivity, indicating robustness in distinguishing between defective and non-defective outputs. Findings suggest that this approach has substantial potential for real-world applications in AM process optimization and quality control. However, further work involving complex modeling and real-world validation is recommended to enhance predictive accuracy and generalizability.