Tribological performance and machine learning-based wear prediction of Acacia arabic-Pencil Cactus fiber composites

Natural fiber reinforced polymer composites are attracting increasing interest due to their sustainability and promising tribological performance. In this work, Acacia Arabica and Pencil Cactus fibers were employed as reinforcements to study the dry sliding wear behavior of polyester composites. Wear tests were conducted using a pin-on-disc apparatus at sliding speeds of 600, 800 and 1000 rpm under normal loads of 40, 60 and 80 N with a constant fiber content of 25%. Treated composites exhibited markedly lower wear loss and coefficient of friction compared to untreated counterparts, with wear loss ranging from 0.00501 to 0.01429 g and coefficient of friction between 0.301 and 0.615. The treated hybrid composite (TACFPC) demonstrated superior performance, achieving a maximum wear reduction of 41.6%. FESEM and EDX analyses confirmed enhanced fiber – matrix interfacial bonding, reduced fiber pull-out and improved surface stability in treated composites. Machine learning-based wear prediction models showed good predictive capability, with the deep neural network achieving an R2 = 0.8037, RSME = 0.000366 mm3/m, and MAE = 0.000300 mm3/m. The optimum condition for minimum wear was identified at 600 rpm, 40 N load and TACFPC, corresponding to a predicted wear rate of 0.002515 mm3 /m and an S/N ratio of 51.99 dB.