Prediction and Optimization of Torsional Performance for Multiaxial Three‐Dimensional Woven Composites

ABSTRACT Current research on the torsional performance of multiaxial three‐dimensional woven composites (M3DWC), particularly for plate‐shaped components such as engine vanes, remains insufficient. Moreover, there is a lack of efficient methods for performance prediction and process parameter optimization. In this study, M3DWC with varying preform structures, weaving needle angles, and yarn tension were designed and torsional tests were conducted to characterize their failure behavior. Based on the experimental data, this study established response surface methodology (RSM) and support vector regression (SVR) prediction models to accurately reveal the complex non‐linear relationship between process parameters and torsional performance, thereby effectively improving the accuracy and efficiency of performance prediction. Furthermore, a hybrid strategy combining genetic algorithms (GA) and multi‐start local search (MLS) was employed to optimize process parameters, achieving the concurrent optimization of multi‐objective performance. The results demonstrate that the specimen with the preform structure of 0°/45°/90°/135°, weaving needle angle of 48.45°, and yarn tension of 1.19 N exhibits the optimal comprehensive torsional performance. This finding provides a theoretical foundation for the design and fabrication of variable vanes made from M3DWC.