Process-driven strength enhancement and progressive damage analysis of three-dimensional woven composites traction rods

With the growing emphasis on lightweight and high-strength materials, optimizing the performance of composites has become increasingly critical for engineering applications. This study presents a comprehensive methodology for optimizing the process parameters of three-dimensional (3D) woven composites traction rods using a progressive damage failure model and an embedding feature selection-based back propagation neural network (EFSBPNN) surrogate model. The forming processes were systematically analyzed to identify key design variables, followed by developing an advanced surrogate model integrating material, structural, and process parameters. A progressive damage failure analysis model was established to predict the structural response, forming the basis for the optimization strategy. Experimental validation demonstrated significant performance improvements, including an 18.56% increase in stiffness, a 40.69% enhancement in strength, and a 6.1% reduction in mass. The surrogate model achieved high accuracy across multiple metrics, and the predictive capabilities of the damage model showed a relative error of 13.5% for stiffness and 10.7% for strength. These findings highlight the potential of the proposed approach for the efficient design and optimization of composite structures.