Lattice Structure Optimization of 3D Printed TPMS under Different Loading Conditions Using Regression Machine Learning

Abstract

Modern manufacturing techniques have been significantly transformed by additive manufacturing (AM). Because of its capabilities like customized part manufacturing and, the ability to manufacture intricate and complex parts with reduced waste of material, additive manufacturing is becoming more popular. However, the properties of the parts manufactured by this method significantly vary with the variation in process parameters. Optimizing these parameters helps to extract enhanced mechanical properties. In addition, lattice structures have created new possibilities for increasing strength while lowering part weight through optimized lattice structures. The effect of lattice structure and process parameters on the specimen made using the fused deposition method (FDM) is the major focus of this study. In this work, three distinct TPMS-base (Triply Periodic Minimal Surfaces) lattice architectures are examined for a range of layer height levels. Investigations are conducted using the L9 orthogonal array. The FDM technique uses PLA plastic filament. The Taguchi method was used for optimization, and samples were evaluated on the UTM and Izod impact testing machines. Moreover, an ML model is created by applying machine learning to the collected data. In tensile and impact test data, neural network and Gaussian process regression models showed low error rates and predicted good accuracy. The neural network model for the flexural test data showed a moderate level of accuracy, suggesting potential for improvement. The models' performance was highlighted by their low RMSE, MSE, and MAE values, which show that they can predict material properties. The overall findings indicated that layer height has less impact on tensile and flexural strength than lattice structure. In contrast to the lattice structure, layer height influences the toughness.