NEURAL NETWORK-BASED PREDICTION OF MODIFIED CFRP COMPOSITES USED IN AEROSPACE APPLICATION

Abstract

CFRPs, or carbon fiber reinforced polymers, are renowned for having superior wear characteristics. In this study, we compared the wear properties of polycar bonate (PC)/acrylonitrile butadiene styrene (ABS) modified di-glycidyl ether of bisphenol A (DGEBA) based CFRPs with electrophoretically deposited (EPD) car bon fibers. We studied the wear rate (WR) and coefficient of friction (COF) of EPD-modified CF with flexible silanized graphene oxide (SGO) bonds using pin on-disk experiments. The findings showed significant decreases in COF and WR of 25% and 34%, respectively. Additionally, the behaviour of the CFRP system with PC/ABS varied depending on the blend ratios. The COF and WR of blends such 70/30, 50/50, and 30/70 showed varied tendencies, however, 10/90 and 90/10 showed consistent decremental patterns, with 90/10 achieving maximum reductions of 19% and 37.9%, respectively. We also looked at how applied load and sliding speed affected wear severity, which changed the wear mechanism from abrasion to delamination. Two testing scenarios were used in the study: room temperature (RT) and cryo-treated (CT). The sample composition was found to be the most important parameter in terms of volume loss using analysis of variance (ANOVA). A Taguchi analysis was also car ried out to optimize the parameters and provide the desired result. We used three predictive models to forecast wear behaviour: the Levenberg-Marquardt algorithm, Artificial Neural Network, and linear regression. The Levenberg-Marquardt algo rithm showed the best performance based on mean squared error (MSE) loss, with 2.7% MSE for RT and 2.15% MSE for CT, demonstrating its potential for precise wear behaviour prediction in CFRP composites. The Levenberg-Marquardt algorithm has a lot of potential for forecasting how CFRP composites will wear. This discovery aids in the comprehension and improvement of CFRP materials for a range of applications needing dependable wear performance.