Machine Learning and Regression Analysis Approaches for Investigation of Mechanical Properties of FDM Manufactured Re-Entrant Auxetic Structures Under Flexural Loading

Abstract

This paper describes an experimental study on re-entrant auxetic structures manufactured by fused deposition modeling (FDM). The feedstock materials of NPR structures are acrylonitrile butadiene styrene (ABS) and poly-lactic acid (PLA). Experimental study is performed to examine the effect of design factors (angle, width, and length of arm) of unit cell of auxetic structures on three responses namely strength, stiffness, and specific energy absorption (SEA) under flexural loading. From the experimental results, it is found that flexural strength improves with increase in all three design factors of ABS structures; while it improves with increase in angle and reduction in width and length of arm for PLA structures. Furthermore, based on experimental study, regression models of responses are developed using analysis of variance (ANOVA). Also, machine learning (ML) models using neural networks are developed to predict all three responses. Results of regression models are compared with NN models to assess accuracy of prediction. Finally, optimal configuration of auxetic structure is determined using gray relational analysis (GRA) to improve the responses; and reduce weight and fabrication time.