Heat deflection temperatures of bio-nano-composites using experiments and machine learning predictions

Abstract

Biocompatibility, biodegradability, and enhanced properties are remarkable features of bio-composites designed to reduce and replace conventional non-biodegradable polymeric materials. Therefore, it is crucial to propose reliable yet economically efficient new bio-composites. Nano-clays (NCs), short latania fibers (SLFs) and new bio fillers, e.g. pistachio shell powders (PSPs) and date seed powders (DSPs), were used to reinforce poly(propylene)/ethylene-propylene-diene-monomer (PP/EPDM) composites. Heat deflection temperature (HDT) tests were conducted. Then a machine learning (ML)-based prediction tool, the K-Nearest Neighbor Regressor (KNNR), was used to investigate HDTs of various bio-composite compositions. KNNR was employed in this study versus the Decision Tree Regressor (DTR) and Adaptive Boosting Regressor (ABR) ML approaches utilized in the previous study by Daghigh et al. (2019) for fracture toughness predictions. Furthermore, in contrast to other natural fiber composites, SLF composites have been seldom investigated. Different contents of SLFs, NCs, macro-sized PSPs and macro-sized DSPs were added to the PP/EPDM to investigate the combined effects of bio-fiber, nano-particulate, and macro bio-particulate reinforcements. This research helps develop an understanding of how such low-cost bio-reinforcements influence the HDT of PP/EPDM composites. ML predictions were used to develop lightweight, cost-effective materials where their use temperatures can be improved. KNNR ML analysis suggested the key factor influencing HDT are SLFs, NCs, DSPs, and PSPs in the order stated.