基于仿生优化算法构建混凝土拱结构应力预测模型

Stress Prediction Model of Concrete Arch Structure Based on Bio-inspired Optimization Algorithm

为获得混凝土拱结构应力预测的标准模型,以实测应力数据为基础,分析得出了混凝土拱结构拱顶、拱脚处正应力和剪应力的变化规律,以BP神经网络模型为基础,基于鲸鱼算法(WOA)、灰狼算法(GWO)、布谷鸟算法(CSA)、蝙蝠算法(BA)和天牛须算法(BAS)5种仿生优化算法,构建了WOA-BP、GWO-BP、CSA-BP、BA-BP、BAS-BP共5种优化模型,并将模型结果与M5树模型(M5T)、随机森林模型(RF)、小波神经网络模型(WNN)计算结果进行了对比,引进积日数(DOY)为模型输入因子,分析得出了最优模型。结果表明,混凝土拱结构自生体积变形及应力变化呈现周期性变化趋势,不同模型对应力的预测精度有所差异,仿生优化算法可提高模型精度,其中WOA-BP模型与实测值的拟合方程斜率分别为0.996、0.998、0.958、0.997,拟合效果最优,同时在4种应力预测中,WOA-BP模型的R_(RMSE)分别为0.021、0.074、0.021、0.062 MPa,R^(2)分别为0.998、0.935、0.992、0.974,E_(NS)分别为0.998、0.930、0.971、0.973,M_(MAE)分别为0.015、0.061、0.018、0.049 MPa,WOA-BP模型在所有模型中精度最高,可作为混凝土拱结构应力预测的标准模型使用。

In order to obtain the standard model of stress prediction of concrete arch structure, this paper analyzed the change law of normal stress and sheared stress at the arch top and arch foot of the concrete arch structure based on the measured stress data. Based on the BP neural network model, five bio-inspired optimization algorithms, whale algorithm(WOA), gray wolf algorithm(GWO), cuckoo algorithm(CSA), bat algorithm(BA) and beetle algorithm(BAS), were used to optimize the model. Five optimization models including WOA-BP, GWO-BP, CSA-BP, BA-BP, and BAS-BP were constructed. The model results were compared with the results of M5 tree model(M5 T), random forest model(RF), and wavelet neural network model(WNN). The day of year(DOY) was introduced as the input factor of the model, and the optimal model was obtained. The results showed that the autogenous volumetric deformation and stress changes of concrete arch structure present a cyclical trend. Different models have different stress prediction accuracy. The bio-inspired optimization algorithms can improve the accuracy of the model. Among them, the slopes of the fitting equations between the WOA-BP model and the measured value were 0.996, 0.998, 0.958 and 0.997, respectively, and the fitting effect was the best. In the stress prediction, the Rof the WOA-BP model were 0.021, 0.074, 0.021, 0.062 MPa, R~2 were 0.998, 0.935, 0.992, and 0.974, Ewere 0.998, 0.930, 0.971, and 0.973, and Mwere 0.015, 0.061, 0.018 and 0.049 MPa, respectively. The WOA-BP model has the highest accuracy among all models and can be used as a standard model for stress prediction of concrete arch structures.