多工艺参数对预浸料摩擦系数的影响及机器学习预示方法

Influence of multiple process parameters on the friction coefficient of prepregs and machine learning prediction method

  • 摘要: 在复合材料成型过程中,预浸料/预浸料和预浸料/模具之间的摩擦滑移行为会导致褶皱、孔隙等缺陷,严重影响构件力学性能。然而复杂构件成型过程中预浸料层间摩擦行为影响因素众多,现有理论模型涵盖的工艺参数有限,导致成型工艺仿真精度低,无法满足高质量成型要求。本文设计了面向多工艺参数的碳纤维预浸料摩擦试验方法,研究了速率、法向力、黏度、表面粗糙度、接触材料、纤维方向等工艺参数对摩擦系数的影响规律,以典型纤维方向 \text0^\texto\text/4\text5^\texto\text/9\text0^\texto 为例,揭示了不同纤维方向的界面摩擦机制。为实现多工艺参数下摩擦系数的快速、准确预示,建立了基于支持向量回归(SVR)的预浸料摩擦系数预示模型。开展相对纤维方向为 \text3\text0^\texto\text/\text0^\texto\text 和 \text6\text0^\texto\text/\text0^\texto\text 的预浸料/预浸料界面摩擦系数的试验和预测,两者偏差小于9%。

     

    Abstract: During the forming process of composites, the friction-sliding behavior between prepreg ply-ply and ply-tool may lead to defects such as wrinkles and pores, which seriously affect the mechanical properties of the components. However, there are many factors affecting the inter-ply friction of the prepreg plies in the forming process of complex components. The existing theoretical models contain insufficient process parameters, resulting in the accuracy of forming process simulation not meeting high-quality forming requirements. In this paper, a friction test method for carbon fiber prepregs was designed for multiple process parameters. The influence of sliding velocity, normal force, viscosity, surface roughness, contact material, and fiber orientation on the friction coefficient were studied. Taking the typical fiber orientations of \text0^\texto\text/4\text5^\texto\text/9\text0^\texto as examples, the inter-ply friction mechanism in different fiber orientations was revealed. In order to predict the friction coefficient of prepreg corresponding to multiple process parameters rapidly and accurately, a prediction model for the friction coefficient of prepreg was established using the support vector regression (SVR) method. Taking the prepreg ply-ply friction behavior with relative fiber orientation of \text3\text0^\texto\text/\text0^\texto\text and \text6\text0^\texto\text/\text0^\texto\text as examples, the experiments and predictions were conducted, and the error was less than 9%.

     

/

返回文章
返回