应用集成学习算法的复合材料损伤定位研究

Application of ensemble learning algorithm in damage localization of carbon fiber reinforced plastic

在碳纤维增强复合材料(carbon fiber reinforced plastic,CFRP)损伤定位研究中,传统单一模型定位误差较大,为提高CFRP的损伤定位精度,提出一种基于集成学习的损伤定位方法。重物加载引起材料应力的变化常用于模拟损伤,施加不同质量用以模拟不同的损伤程度,实验在CFRP上布置应变片,获取在不同损伤程度下的应变特征量,针对同一损伤程度,利用网格搜索法对支持向量回归机(SVR)和分类回归树(CART)模型进行参数寻优,将最优模型参数应用到集成学习模型后对比SVR、CART两种单一模型和SVR_Adaboost、随机森林(RF)、极端随机树(ERT)、梯度提升决策树(GBDT)四种集成学习模型的定位精度。结果表明:在损伤程度1的数据集上,弱学习器迭代50次的随机森林模型和极端随机树模型平均定位误差最低,为6.2 mm,并将该两种模型在损伤程度2和3的数据集上进行坐标预测。该集成学习算法可用于碳纤维增强复合材料损伤定位,且定位精度比传统单一模型定位精度高。

In the study of damage localization of Carbon fiber reinforced plastic(CFRP),the localization error of the traditional single model is large.In order to improve the damage positioning accuracy of CFRP,a damage localization method based on ensemble learning algorithm was proposed.The change of material stress caused by heavy load is often used to simulate damage,and different weights are applied to simulate different damage degrees.The experiment is to arrange strain gauges on CFRP to obtain the strain characteristic quantity under different damage degrees.For the same damage degree,the grid search method was used to optimize the parameters of Support vector regression(SVR)model and Classification and regression tree(CART)model.After applying the optimal model parameters to the ensemble learning model,the positioning accuracy of SVR,CART,SVR_Adaboost,random forest(RF),extremely randomized trees(ERT)and Gradient boosting decison tree(GBDT)are compared.The results showed that on the training set of damage degree 1,the mean positioning error of the random forest model and the extreme random tree model with 50 iterations of the weak learner is the lowest,which is 6.2 mm,and was tested on the data sets of damage degree 2 and damage degree 3.The results are indicating that the ensemble learning algorithm can be used on damage localization of CFRP,and the positioning accuracy is higher than that of traditional single models.