Research on prediction model of bond strength between reinforcement and concrete based on BP-ANN and RBF-ANN
An investigation into the predictive capabilities of neural networks for assessing the bond strength between steel bars and concrete was conducted and the output performance of neural networks was assessed.Utilizing a vast array of experimental data,this research introduced an advanced neural network-based model for predicting the bond strength in deformed steel bars and concrete,a critical consideration for both academic research and practical engineering applications in concrete structures.The study analyzed data from 290 bond anchorage tests,employing back propagation artificial neural network(BP-ANN)and radial basis function neural network(RBF-ANN)algorithms.These algorithms were used to ascertain the influence pattern of concrete strength,protective layer thickness,reinforcement diameter,anchorage length,and stirrup ratio on the bond performance between deformed steel bars and concrete.Additionally,the study established a prediction model based on an enhanced neural network algorithm,evaluating the impact of various data preprocessing techniques and neuron training counts on the model's predictive results.Comparative analyses between the proposed model and traditional models were performed,focusing on predictive accuracy and variability.The research also introduced a formula for calculating critical anchorage length.Findings indicated that the BP-ANN model,with a mean,standard deviation,and coefficient of variation of 1.009,0.188 and 0.86,respectively for the predicted-to-experimental value ratio,slightly outperformed the RBF-ANN in terms of prediction accuracy.This proposed model demonstrated enhanced precision and stability in predicting the bond strength between steel bars and concrete,offering novel insights into addressing bonding issues between steel bars and concrete.
reinforced concretebond strengthimproved neural networkinfluence parametersprediction modelbonding and anchoring testBP-ANNRBF-ANN
国家科技期刊平台
NSTL
万方数据
维普
