Multi-matrix co-optimization and artificial neural network-assisted property prediction of carbon fiber reinforced friction materials

Carbon fiber reinforced friction materials are critical in aerospace and transportation braking systems, yet conventional organic matrix often fail under high temperatures. This study introduces a hybrid matrix combining alumina-based inorganic binder (AO) with phenol-formaldehyde resin (PF) and silicone rubber (SR), leveraging AO’s high-temperature stability to compensate for organic limitations. Using an HMI-GA-BP artificial neural network, the matrix ratio was multi-objectively optimized and experimentally validated. The optimal formulation OPT-3M (AO:SR:PF= 22:12:66), compared to the PSA formulation, significantly improved compressive strength (192.0 MPa, +74 %) and shear strength (18.7 MPa, +34 %), while maintaining a volume wear rate below 0.45 × 10⁻⁷ cm³ /J and a stable friction coefficient above 0.4 across 100–350°C without thermal fade. This work offers a new approach to designing heat-resistant, low-wear composites and advances the understanding of inorganic binders in friction materials.