Predicting and analyzing the tensile properties of oxygen plasma-treated banana fiber-reinforced recycled high-density polyethylene composites via Taguchi design and machine learning models
Year: 2026
Type: article
Abstract: The growing demand for sustainable, eco-friendly materials has intensified interest in natural fiber-reinforced polymer composites, yet the synergistic optimization of interfacial properties and predictive modeling of mechanical performance in recycled thermoplastic systems remains inadequately explored. This study investigated the physio-mechanical performance of oxygen plasma-treated banana fiber-reinforced recycled high-density polyethylene (rHDPE) composites and compared five supervised machine learning models—Linear Regression, Decision Tree, Random Forest, Support Vector Regression, and an artificial neural network (ANN)—for predicting tensile properties. Composites were fabricated by compression molding using a Taguchi L9 design, varying fiber orientation, fiber content (10–20 wt.%), and plasma treatment time (5–15 min). The optimal formulation (transverse orientation, 20 wt.% fiber, 5 min plasma treatment) achieved a tensile strength of 38.25 MPa—a value approximately 104% higher than typical literature values reported for neat rHDPE and 26% higher than reported untreated fiber benchmarks, with tensile modulus showing a 6.2–11.1 times increase relative to literature rHDPE baselines. Leave-one-out cross-validation (LOOCV) revealed fold-wise relative accuracy ranging from 12%–96% across the nine held-out formulations depending on the algorithm. Given the small dataset (n = 9), model rankings remain strictly exploratory; however, LOOCV metrics tentatively favored the SVR for tensile strength, strain at break and Random Forest for modulus. This performance gap highlights why cross-validation, rather than in-sample fit alone, should underpin predictive modeling claims for small experimental datasets. The study establishes a data-driven framework combining plasma-assisted interfacial engineering with machine learning for the efficient design of green composites, showing strong potential for semi-structural and lightweight applications-such as automotive interior panels, packaging, and construction components-supporting circular-economy valorization of agricultural waste and post-consumer recycled plastic.
Source: Discover Composites
Language: English
Subfield: Polymers and Plastics
Field: Materials Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7225250491