Wear behaviour prediction framework for agro-waste reinforced metal matrix composites using Taguchi–GRA multi response optimization and machine learning techniques

Abstract

This study explores the dry sliding wear behaviour of cost-effective, eco-friendly Al6061 aluminium matrix composites (AMCs) reinforced with corn cob ash (CCA), an agro-waste byproduct, via bottom-pouring stir-casting, to overcome the challenges posed by conventional ceramic reinforcements, including strength-to-weight ratio, wear resistance, and cost. The composite was optimised using a multi-response identifier and Taguchi-grey relational analysis (GRA). The test was performed using a Design of Experiments (DoE) L36 orthogonal array with various loads and sliding velocities. The optimal dry sliding wear test parameters were 4% CCA reinforcement, 10 N load, and 1 m/s velocity. Seven different types of machine learning and statistical regression methods were utilised to improve prediction performance. The results showed that the ANFIS and ANN models outperform other models, with R² values of 0.99 and 0.956, respectively. The results revealed that combining modern ML models with Taguchi-GRA is an effective approach for producing lightweight, wear-resistant agro-based AMCs, which will help expand their applications in the aerospace and automotive industries.