基于PSO-BP神经网络的玻璃钢管首层失效预测研究

Research on first-ply failure prediction of fiberglass reinforced plastic pipes based on PSO-BP neural network

预测玻璃钢管的首层失效对保障其在水利输水工程的服役安全具有重要意义。本文通过粒子群算法优化BP神经网络(PSO-BP)实现对玻璃钢管在双轴应力下管道复合层首层失效的预测,并将PSO-BP模型预测结果通过试验数据进行验证。研究表明:PSO-BP神经网络模型对玻璃钢管首层失效的平均预测准确率可达85%以上,收敛速度及预测准确率较对照BP神经网络模型均存在优势;绘制的轴向应力与环向应力的双轴失效包络线显示出PSO-BP模型预测的失效包络线与试验中测得的失效包络线十分接近,且预测失效包络线绝大部分位于试验失效包络线的内侧,故该模型是一种偏安全的预测模型,可作为一种按规范鉴定玻璃钢管合格前的有效判断手段。

Predicting the first-ply failure of fiberglass reinforced plastic(FRP)pipe is essential to ensure service safety in water conveyance projects.In this research,a particle swarm algorithm optimized backpropagation neural network(PSO-BP)is used to predict the first-ply failure of FRP pipe under biaxial stress.Experimental data verify the prediction results of the PSO-BP model as well.The results illustrated that the average prediction accuracy of the PSO-BP neural network model for the first-ply failure of the FRP pipe could reach more than 85%,which has advantages over the control backpropagation neural network model regarding convergence and accuracy.The plotted biaxial failure envelopes of axial and hoop stresses showed that the failure envelopes predicted by the PSOBP model are very close to the failure envelopes measured in the test.The predicted failure envelopes are primarily located in the test failure envelopes.Most of the predicted failure envelope is located inside the test failure one.Therefore,the model is a rational safety prediction model,which can be used as an effective approach to identify FRP pipes before they are qualified according to the specification.