Strength Prediction Research On Pullout Method Testing The Strength Of Ultra-high Strength Concrete By Neural Network Model

Since the ultra-high strength concrete has been successfully developed,because of its characteristics of high strength,light weight,and corrosion resistance,it can meet the needs of various structural forms and the requirements of use in harsh environments,has been widely used in bridges,high-rise buildings and utility tunnel.After twenty years of development,ultra-high strength concrete has already achieved factory prefabrication,the world’s first fully prefabricated concrete bridge was constructed in Changsha in 2016,this technology will push for further development of ultra-high strength concrete in practical applications.Although the time of the application of ultra-high strength concrete is more than two decades,but the strength testing method of ultra-high strength concrete remains the traditional drilling core method,there is no second practical way.And the drilling core method is very damaging to the relatively thin ultra-high strength concrete component,if the sample number rise,the safety implications for the monitored structures will also rise.If the pullout method which has less influence can be applied to test the strength of ultra-high strength concrete,the strength defection problem will be solved.This paper designed the experiment to verify applicability of the cast-in-place pullout method and the post-install pullout method testing the strength of ultra-high strength concrete,using the least square method which is commonly used to simulate the experimental data,getting the strength formula,and the experimental data were processed by the neural network model,to get the model that can predict the strength of ultra-high strength concrete.The main research content and conclusions are as follows:(1)systematically summarized the development history of ultra-high strength concrete,the domestic and overseas research situation of pullout method and neural network model,besides,briefly summarized the main content of this article;(2)carried out the strength test experiment of ultra-high strength concrete by cast-in-place pullout method and post-install pullout method,which designed 3 levels of strength,each strength level set 6 sets of specimens and test blocks.Using the least square method to fit for the experimental data and calculating the related parameters,get the conclusion that cast-in-place pullout method and post-install pullout method are appropriate for testing the strength of ultra-high strength concrete,and get the corresponding strength formula;(3)described the principles of the neural network model from the three aspects:biological neuron model,neuron model and neural network model.Analyzed the advantages and disadvantages of several major neural network models,the BP network model and radial basic function model were adopted to deal with the experimental data and the prediction model was obtained;(4)get the conclusion that cast-in-place pullout method has higher detection accuracy while the post-install pullout method has wider strength application range,by comparing the detection accuracy and detection range of two detection methods(cast-in-place pullout method and post-install pullout method);get the conclusion that ultra-high strength concrete and steel fiber cement mortar have similar trend and the effect of steel fiber on enhancing the pulling force is more obivious,by comparing the strength curves of ultra-high strength concrete,common concrete and 3 fiber cement mortar;get the conclusion that the neural network model has higher prediction accuracy by comparing the prediction accuracy of the least square method curve and the neural network model.