Microstructure-Based Prediction of Mechanical Properties of Austempered Ductile Iron Using Multiple Linear Regression Analysis

Abstract

Multiple linear regression analysis (MLRA) was used to predict the mechanical properties of austempered ductile iron (ADI) including yield and tensile strength, uniform elongation, hardening exponent, as well as fracture energy by building a model that uses characteristic features of microstructural constituents as input parameters. The complex multi-scale microstructure of ADI, which is composed of spherical graphite particles over 10 mu m diameter; and an ausferritic matrix with sub-micron sized features, makes it ideal for prediction of mechanical properties. For that purpose, low alloyed ductile iron samples austempered between 300 and 400 degrees C for 45-180 min were tensile tested, and also multi-scale microstructural characterization were carried out using optical microscope, SEM, and EBSD technique. Moreover, a sensitivity analysis was performed to determine which microstructural parameter(s) each mechanical property is most sensitive to. The results show that tensile and yield strength are most sensitive to size and morphology of matrix phases. Moreover, the size and aspect ratio of acicular ferrite correlate well with those of high-carbon austenite; since both form during transformation of parent austenite into ausferrite during austempering treatment. Equiaxed parent austenite grains transform into ausferrite with acicular morphology during the austempering treatment; and presence of equiaxed austenite grains in the austempered samples indicates untransformed regions during austempering treatment. Ductility was found to be more sensitive to nodularity of graphite particles, and this sensitivity was attributed to the size difference between graphite particles and grain size of matrix phases.