Predicting the Tensile Strength of 4D Printed PLA/EPO/Lignin Biocomposites Using Machine Learning

Abstract

The allure of 4D printing and machine learning (ML) for various applications is unquestionable, and researchers are striving hard to improve their performance. In this work, machine learning has been applied to predict the tensile strength of the 4D printed materials. The study investigated the reinforcement of polylactic acid (PLA) filament with lignin from oil palm empty fruit bunches (OPEFB) in the presence of epoxidized palm oil (EPO) as 4D printable filament. The alkaline extraction method was carried out used sodium hydroxide (NaOH), followed by precipitation with mineral acids utilizing one-factor-at-a-time (OFAT). Thereafter, the tensile strength of the 4D printed material was evaluated by tensile testing machine followed by machine learning prediction in which convolutional neural network (CNN) was adopted. The morphology of the 4D printed materials was determined by scanning electron microscope (SEM). The SEM micrograph of the tensile test of biocomposites revealed layer-by-layer formation of the filaments on the printed unfilled PLA biocomposite indicating lower inter-filament bonding. In the first trial, the actual result of the experiment was evaluated to be 24.44 MPa while the CNN prediction was 25.53 MPa. In the second attempt, the actual result of the experiment was 31.61 MPa whereas the prediction from CNN was 27.55 MPa. The coefficient of determination value obtained from CNN prediction is 0.12662. The current study indicates that machine learning is an important tool to optimize and/or predict the properties of 4D printing materials.