| With the development of social economy,reinforced concrete structure has been widely used in all kinds of construction projects,but the safety problems and economic losses caused by steel corrosion have aroused widespread social concern.As a substitute for steel bars,FRP bars have the advantages of light weight,high strength and corrosion resistance,so their application prospects are very wide.If FRP bars and concrete want to give full play to their respective performance in the structure,they need to have a good bond property between them to ensure better cooperation.Among them,the research on the bond strength between FRP bars and concrete at home and abroad is mostly based on the condition of room temperature,and the research on the bond strength between FRP bars and concrete at high temperature is relatively few.Therefore,this paper will focus on the two aspects of bond slip curve and bond strength between FRP bars and concrete,combined with the artificial neural network method which has been widely used in various fields in recent years to systematically study the bond performance between FRP bars and concrete.The main research contents and conclusions of this paper are as follows:（1）The research results of bond performance between FRP bars and concrete structures at home and abroad are summarized and analyzed,and the bond mechanism between FRP bars and concrete is systematically elaborated from two aspects: the composition of bond stress and the form of bond failure,in addition,the factors affecting the bond performance between FRP bars and concrete are summarized.Ten kinds of bond-slip constitutive relations,such as Mavlar model,BPE model and continuous model,were summarized and analyzed,which laid a theoretical foundation for the proposed new bond-slip constitutive relations in this paper..（2）Based on 45 groups of GFRP reinforcement concrete pull-out test data,which considered the diameter of bars,the test loading rate and the surface form of bars,a new constitutive model of bond slip between FRP bars and concrete was proposed,the model curve consists of micro-slip segment,ascending segment,descending segment and residual segment.The micro-slip segment is represented by a linear equation,the ascending segment is represented by the Weibull Model curve equation,and the descending segment is represented by a linear segment in the BPE Model.The bond stress value of the residual segment remains unchanged.The new bond-slip constitutive model was verified by using the FRP reinforcement concrete pull-out test data from the research group and relevant literatures.At the same time,it was compared with the commonly used bond-slip constitutive model.It was found that the model could fit the test data well and the accuracy was completely satisfied.（3）Based on the experimental data in the literature,292 sets of FRP bar pull-out tests at room temperature were established,among which 242 sets of data were used for training and50 sets of data were used for simulation prediction.The artificial neural network was used to predict the bond strength between FRP bars and concrete.A three-layer artificial neural network model was trained by using the back propagation algorithm.The input layer of the model included seven parameters: FRP bar type,surface form,FRP bar diameter,anchorage length,failure mode,concrete compressive strength and normalized concrete protective layer thickness.The output layer is the bond strength of FRP reinforced concrete interface.The results show that the BP neural network model has good prediction and generalization ability,and the prediction error is small.Moreover,the BP neural network model can comprehensively consider many factors affecting the bond strength of FRP reinforced concrete interface,and give accurate prediction results.（4）Based on the experimental data in the existing literature,a database of 151 FRP reinforced concrete pull-out tests under high temperature was established,of which 140 were used for training and 11 were used for prediction.Compared with the prediction model at normal temperature,the input layer of the prediction model at high temperature has one more temperature variable.The artificial neural network model is trained by using the test data to establish the prediction model of interface bond strength,and the predicted values of the model were compared with the experimental data and the corresponding mathematical model calculation results to verify the validity of the model in predicting the interface bond strength of FRP reinforced concrete. |
