Ultrasonic Identification Of The Stress Analysis Of Loaded Concrete Based On Support Vector Regression

Ultrasonic has strong penetrating ability and is suitable for non-destructive testing of various materials.The testing of concrete defects used used ultrasonic is mature.However,as a multiphase composite material,it is difficult to accurately measure ultrasonic parameters of concrete,and determining the correlation between ultrasonic parameters and concrete stress is a difficult task.The paper studies the law of ultrasonic spread in concrete,and uses Support Vector Regression to anslysis the numerical model and experimental data.The main research work of the thesis includes:1.The numerical simulation method of meso-concrete is studied.The concrete material is considered to be composed of three materials: aggregate,mortar and bonding interface.The aggregate is randomly placed,the aggregate unit is made of homogeneous elastic material,the mortar unit is made of damage plastic material,and the interface is simulated by the cohesion unit.A combination of damage plastic model and cohesion unit is proposed to establish a random aggregate numerical model of meso-concrete.The random aggregate model and the pure cement mortar model established by this method are compared with the ultrasonic propagation law under the ultrasonic single excitation source,and the two excitation modes of the ultrasonic phased array are simulated.The results show that the numerical simulation effect obtained by combining the two methods of damage plasticity model and cohesive force unit is better.Compared with the pure cement mortar model,the presence of aggregate in the random aggregate model will affect the propagation of ultrasonic waves in the material;ultrasonic phase control In the array simulation,the ultrasonic phased array can effectively enhance the energy of the ultrasonic signal and improve the resolution of the concrete material state.2.Use Fourier transform to process the numerical simulation ultrasonic parameter data,combine the time domain data and frequency domain data to obtain five ultrasonic wave speed,first wave amplitude,main frequency amplitude,nonlinear coefficient and ultrasonic spectrum area.Parameters,the correlation between five ultrasonic parameters and stress is studied separately,and the regression equations of ultrasonic parameters and stress of ten aggregate numerical models are juxtaposed and fitted according to the principle of data regression analysis.It was found that the fitting effect of simple juxtaposition curve is poor,and the coefficient of determination is basically below 0.9.Especially in order to increase the coefficient of determination,the use of high-order polynomial fitting will cause obvious over-fitting.Therefore,the support vector regression method was used to study the correlation between the ultrasonic parameters and stress of ten aggregate numerical models,and the influence of selecting different normalization methods and different kernel functions for support vector regression on the prediction results was summarized.Aiming at the special problem of the thesis,the first normalization method is proposed.The results show that: support vector regression can analyze multi-factor problems,and summarize the support vector regression model of the correlation between five ultrasonic parameters and stress.The results are better than the collocation curve fitting effect,and the first normalization method and diameter are used.The prediction result obtained from the basic kernel function is the best.3.Designed and completed the ultrasonic test to identify the stress state of concrete.Comparing the test results with the numerical simulation results,it is found that the numerical simulation method does not consider the original micropores in the concrete,and the ultrasonic parameter results of the numerical simulation do not have a slight increase in the energy of the receiving point in the low stress stage.The numerical simulation method is generally feasible in other aspects.Three models of simple support vector regression,support vector regression optimized by traversal algorithm and support vector regression optimized by particle swarm algorithm were used to analyze the test data,and the support vector regression prediction results of the three methods were compared.The results show that the particle swarm optimization optimization support vector regression model has a shorter calculation time and better accuracy,and is most suitable for solving the problem of the correlation between ultrasonic parameters and stress.The article studies the ultrasonic waves spread process of in the loaded concrete materials,using numerical simulation and experiment to verify each other;using SVR to analyze the ultrasonic parameter results,using PSO to find the appropriate hyperparameters,which provides a certain reference for researchers who use ultrasonic to identify the stress state of loading concrete.