Inverse-Prediction of Material Property in Fused Filament Fabrication/Fused Deposition Modeling

Abstract A great body of research has focused on developing effective methods for predicting the mechanical functionality of 3D-printed products from the printing process and material data. However, the material description often relies on test results on a small number of specimens. Specimen test results are influenced by both the material and the printing process. Therefore, if the printing process differs, mechanical properties obtained from specimen tests cannot be directly applied to 3D-printed parts. However, the relation between mechanical functionality and material data becomes bijective when The geometry and printing process are fixed. In this work, we propose an intermediate process of predicting the material properties from the specimen’s test results, printing process, and geometry. The predicted material can be used, in conjunction with the product’s printing process and geometry, to predict the mechanical functionality of printed products. The authors of the work propose (1) data synthesis on various printing conditions and its specimen test results and (2) a multi-layer perceptron model that predicts the material properties. The proposed model achieved an R-squared value of 0.93 on the prediction of Young’s modulus and 0.96 on the tensile yield strength. Also, the model is evaluated to predict Young’s modulus and tensile yield strength of the material from the specimen test results achieved by finite element analysis on specimen CAD models. The average prediction accuracy for Young’s modulus is 95%, and tensile yield strength is 90%.