Research On Stress Recognition Of Compressed Concrete By Ultrasonic Wave Based On Random Forest Algorithm

Concrete,as the main building material for bridge structure,often produces internal defects due to the variation in stress levels.Ultrasonic detection techniques have been widely used for non-destructive testing of concrete,such as old bridge inspection.However,concrete is not an isotropic material.Therefore,it is valuable to study how to effectively obtain ultrasonic response characteristics to evaluate stress levels and establish numerical models that can guide practical engineering.In this paper,the propagation process of ultrasonic waves in concrete was simulated,and indoor experiments were designed using the random forest algorithm to analyze and process numerical simulation and experimental results.The main research content and achievements are as follows:1.A random convex polygon aggregate microscopic concrete model was established.PYTHON scripts were used in ABAQUS to randomly place convex polygon aggregates and polygonal initial voids,endowing the aggregates with homogeneous elastic materials,mortar with damage plastic materials,and the interface layer with cohesive unit properties.The uniaxial compression curve of concrete was obtained after loading,which could well simulate the compression process of concrete compared with the specification.The propagation behavior of ultrasonic waves in compressed concrete was simulated using displacement loading.The results showed that the random polygon aggregates in concrete had a significant impact on the propagation process of ultrasonic waves,and the wave velocity could characterize the variation of concrete stress.2.The ultrasound parameters in the numerical model were processed by Fourier transforms,and five ultrasound parameters,including wave velocity,first wave amplitude,main frequency amplitude,weighted spectral area,and nonlinear coefficient,were selected to fit the relationship between ultrasound parameters and stress using single-factor data regression.The fitting result was good,but there were overfitting problems for some curves due to the small sample size,and this belonged to a multi-factor regression problem.Therefore,multiple linear regression models,support vector regression models,and random forest regression models were established,and the results showed that these three models had better results than single-factor fitting.The random forest regression model produced the best prediction results.3.An ultrasonic wave identification uniaxial compression concrete stress test was designed.By analyzing the waveform data of 30 specimens collected in the experiment and comparing them with the numerical simulation results,it was found that the numerical model establishment method was basically feasible.Simple random forest regression prediction analysis of the test data showed that the effect was generally mediocre.Cross-validation methods and sparrow search algorithms were used to optimize hyperparameters,determine appropriate hyperparameters,and obtain good prediction results.When analyzing the prediction results,it was found that the prediction result was higher at the low-stress stage.However,to improve the prediction accuracy throughout the stage,a double transverse wave test was designed for 12 specimens,and the results showed that the double transverse wave method could improve the prediction accuracy at the low-stress stage,thus compensating for the defect of the random forest regression model’s high prediction results on concrete stress at the low-stress stage.This paper analyzed and studied the propagation of ultrasonic waves in uniaxial compression concrete,and compared and verified it using numerical models and indoor experiments.The random forest regression analysis was used to analyze the correlation between ultrasound parameters and stress,and hyperparameters were optimized using cross-validation and sparrow search algorithms.The double transverse wave method was used to fit the relationship between ultrasound parameters and stress at the low-stress stage,providing a reference for practical engineering applications such as old bridge inspection,repair,and reinforcement.