Friction and wear characteristics of anti-skid masterbatch filled acrylonitrile butadiene styrene (ABS) based polymer composite using Taguchi and machine learning techniques

Abstract

The effect of anti-skid masterbatch (ASM) filled acrylonitrile butadiene styrene (ABS) composite on friction coefficient (COF) and specific wear rate (SWR) characteristics are explored. The composites were developed by varying percentage (by wt) of ASM filler loadings with the ABS matrix through a co-rotating, intermeshing, twin-screw extruder using melt-mixing and injection molding process. The developed composite samples underwent various experimental runs following the L₁₆ orthogonal array (OA) design of experiments (DOE) at three factors (filler content, load and frequency) and four levels to assess the friction and wear behavior of the composite. The investigation identifies the optimal parameter settings for COF and SWR at A₄B₁C₂ and A₄B₂C₁ parameter levels. Results from the confirmation test demonstrated a ~ 24% increase and a ~ 8% reduction in COF and SWR characteristics, respectively. ANOVA tests revealed that load and frequency had a significant effect on COF and SWR. In addition, supervised machine learning (ML) models were employed to predict the COF and SWR behavior of the ABS/ASM composite. The study found that random forest regressor (RFR) and gradient boosting regressor (GBR) models out-performed decision tree regressor (DTR) models, with R² values of 0.9109 and 0.8909 for COF, and 0.9079 and 0.8979 for SWR, respectively. These models closely matched the optimized experimental results and exhibited lower performance matrix values, such as mean absolute error (MAE) and root mean square error (RMSE), further validating their accuracy. The Taguchi statistical model also showed strong predictive power with R² values of 0.9587 for COF and 0.7074 for SWR. The significance of this study lies in its contributions to the field of composite material development, especially in the optimization of tribological properties for practical applications. The findings not only highlight the potential of ABS/ASM composites in improving friction and wear characteristics but also showcase the effectiveness of ML models in predicting these behaviors with high accuracy. This research opens up new avenues for the application of ASM-filled composites, particularly in floor liner applications, and sets a foundation for further advancements in the use of data-driven approaches in material design and performance prediction.