Utilization of red mud as copper substitute in eco-friendly resin-based brake composites: Performance evaluation and machine learning prediction

Abstract

Red mud, a highly alkaline industrial solid waste generated during the production of alumina, poses significant environmental challenges due to its massive stockpiling. This study proposes a novel strategy to utilize red mud as a filler in copper-free resin-based brake composites, aiming to replace environmentally problematic and costly copper. This approach not only offers a pathway towards copper-free brake composites but also presents a potential solution for the eco-friendly disposal of red mud. The effects of red mud content on the physical properties, mechanical properties, and friction-wear performance of the composites were systematically investigated, coupled with a comprehensive analysis of the wear mechanisms. Results indicate that incorporating an appropriate amount of red mud enhances the density, hardness, and impact strength of the composites. Crucially, it effectively improves the friction coefficient at medium to high temperatures, mitigating thermal fade. While density and hardness increased proportionally with red mud content, impact strength exhibited a complex non-linear trend, initially decreasing, then increasing, before decreasing again. The composite containing 35 wt% red mud demonstrated the most favorable overall friction-wear performance. An optimized formulation was derived using Response Surface Methodology (RSM): 18 wt% phenolic resin, 5 wt% bamboo fiber, 15 wt% alumina, 39 wt% red mud, 3 wt% graphite, 5 wt% nitrile rubber powder, and 15 wt% barium sulfate. This formulation maintained a stable friction coefficient between 0.48 and 0.50 across a test temperature range of 100 ∼ 350 °C. Furthermore, five machine learning methods were employed to establish predictive models correlating the composite formulation with the friction coefficient. Cross-validation revealed that the Particle Swarm Optimization-Back Propagation (PSO-BP) neural network exhibited the best generalization capability on the test set (R² = 0.87284, prediction accuracy = 97.5987 %), identifying it as the optimal prediction model. This study successfully demonstrates the feasibility of using red mud to replace copper in brake composites, yielding high-performance, eco-friendly, copper-free resin-based brake composites.