Research On Grout Transmission Method Detection Of Prestressed Holes Based On Bp Neural Network

In recent years,China’s transportation infrastructure has been developed rapidly,and with the rapid development of transportation construction,post-tensioned prestressed concrete beams have been favored by the engineering community due to their high strength materials,excellent spanning ability,lightness of their own structure,and their strong anti-cracking ability.During the production of post-tensioned prestressed concrete beams,the grouting process is extremely important,and the safety and service life of the concrete beams are directly affected by it,so it is necessary to test the grouting quality.The more mature grouting quality inspection methods for prestressed pores are mostly stress wave inspection methods,which are non-destructive,fast,real-time and can reflect the grouting quality comprehensively.However,most of these detection methods are transverse detection,which can only identify local defects and have low detection efficiency.In this paper,we propose a longitudinal transmission detection method based on BP neural network algorithm to intelligently discriminate the collected signals by BP neural network,so as to quickly evaluate the grout compactness of pre-stressed holes.The main research content of this paper includes the following three parts:(1)Using ABAQUS software,a numerical simulation study of stress wave propagation in prestressed pores with different grouting degrees was carried out to obtain the relationship curve between pore grout density and wave velocity,which decreases with increasing grout density.The numerical simulation results were input into Matlab simulation software,and the relationship curves of signal frequency as well as energy with grouting degree and the corresponding frequency and energy values were obtained by Fourier transform.As the grout density increases,the signal energy decays significantly and the frequency drifts to lower frequencies;(2)Through finite element simulation and simulation analysis,wave speed,frequency and energy are determined as BP network characteristic parameters,and the simulated data and the actual detection data are input into the BP neural network as sample data to construct the BP network model.By comparing the predicted value with the expected value error,the network structure is adjusted and the BP network structure is determined as a 3-layer topology with 4 nodes in the input layer,6 nodes in the hidden layer,1 output node,500 iterations and a learning rate of 0.0001.The BP neural network predicts the hole grout compactness error within ±0.03,which meets the actual demand;(3)The beam yard of the first section of Huaguan Expressway in Guangzhou made 1:1 footage test box girders,set 0,70%,80%,90%,95% and 100% aperture grouting degrees,and tested the apertures with this testing method.During the implementation of this method,the detection accuracy was improved by using the stable Chirp signal as the source signal through the signal research,and the error between the detection results and the actual set bore grouting degree was within 3%,which proved the feasibility of this method.Meanwhile,the 50 m T-beam of Yang-Xin Expressway Yellow River Special Bridge was tested,and the grouting test results met the actual engineering requirements.