Multi-objective optimization of mechanical properties of additively manufactured tri-hexagon pattern specimens using machine learning algorithms

This paper provides a thorough investigation of research carried out on predictive modeling and multi-objective optimization of mechanical properties in additive manufacturing, with a specific emphasis on specimens manufactured using a tri-hexagon pattern in additive manufacturing. This study examines the complex correlation between geometric factors, printing circumstances, and material qualities to improve the performance and dependability of parts in additive manufacturing procedures. The research emphasizes the importance of geometric elements, specifically the tri-hexagon pattern, in determining mechanical behavior and structural integrity. This conclusion is based on a thorough examination of existing literature. Utilizing machine learning methods like logistic regression and artificial neural networks, prediction models are created to precisely predict mechanical qualities based on geometric features and printing conditions. In addition, this paper examines the utilization of multi-objective optimization techniques, such as the Non-dominated Sorting Genetic Algorithm II (NSGA-II), to optimize conflicting objectives such as strength, stiffness, and weight simultaneously. NSGA-II demonstrates a convergence pattern by decreasing the average distance between individuals in the population from 0.5 to 0.2 during a span of 50 generations. The solutions in the last iteration exhibit an average tensile strength (TS) of 1.1 MPa and an average modulus of elasticity (ME) of 31.3 MPa, highlighting the algorithm's effectiveness in achieving desirable values for both targets simultaneously. In addition, this research conducted using Response Surface Methodology (RSM) demonstrates, through an overlay plot that the optimal printing orientation is 0° and the most favorable infill density is 20%. This finding is supported by a coefficient of determination exceeding 90%, which confirms the effectiveness and efficiency of the proposed algorithms.