Neural Network Optimization of Mechanical Properties of ABS-like Photopolymer Utilizing Stereolithography (SLA) 3D Printing

The optimization of mechanical properties in acrylonitrile butadiene styrene-like (ABS-like) photopolymer utilizing neural network techniques presents a promising methodology for enhancing the performance and strength of components fabricated through stereolithography (SLA) 3D printing. This approach uses machine learning algorithms to analyze and predict the relationships between various printing parameters and the resulting mechanical properties, thereby allowing the engineering of better materials specifically designed for targeted applications. Artificial neural networks (ANNs) can model complex, nonlinear relationships between process parameters and material properties better than traditional methods. This research constructed four ANN models to predict critical mechanical properties, such as tensile strength, yield strength, shore D hardness, and surface roughness, based on SLA 3D printer parameters. The parameters used were orientation, lifting speed, lifting distance, and exposure time. The constructed models showed good predictive capabilities, with correlation coefficients of 0.98798 for tensile strength, 0.9879 for yield strength, 0.9823 for Shore D hardness, and 0.98689 for surface roughness. These high correlation values revealed the effectiveness of ANNs in capturing the intricate dependencies within the SLA process. Also, multi-objective optimization was conducted using these models to find the SLA printer’s optimum parameter combination to achieve optimal mechanical properties. The optimization results showed that the best combination is Edge orientation, lifting speed of 90.6962 mm/min, lifting distance of 4.8483 mm, and exposure time of 4.8152 s, resulting in a tensile strength of 40.4479 MPa, yield strength of 32.2998 MPa, Shore D hardness of 66.4146, and Ra roughness of 0.8994. This study highlights the scientific novelty of applying ANN to SLA 3D printing, offering a robust framework for enhancing mechanical strength and dimensional accuracy, thus marking a significant benefit of using ANN tools rather than traditional methods.