Research On Ultrasonic Testing Method Of Steel Strand Stress Based On RF-SVR Algorithm

The steel strand is widely used in bridge engineering structures and is often used as the main tensile member.In the actual application process,the steel strand is in a high-stress working state and harsh natural environment all year round,which is prone to different degrees of damage and stress loss.In severe cases,it will endanger the service life of the bridge and the life safety of users.Therefore,in order to ensure their stability and reliability and avoid the occurrence of sudden disasters of bridges,the long-term health detection/monitoring of in-service steel strands is particularly important.The stress state of steel strand is the key index to evaluate the safety of structure.In this thesis,the related problems of ultrasonic guided wave detection method for steel strand stress are studied.By means of time-frequency signal analysis,random forest(RF)and support vector regression(SVR)algorithm,the noise reduction method suitable for ultrasonic guided wave signal is proposed,and the SVR steel strand stress identification model based on RF feature optimization is constructed to realize the automatic identification of steel strand stress level.The main research contents are as follows :(1)According to the characteristics of ultrasonic guided wave signals under different stress levels,a quantitative method of stress characteristic parameters of steel strand is proposed.By analyzing the components and non-stationary characteristics of ultrasonic guided wave signals,the significance of signal preprocessing technology in ultrasonic guided wave detection is discussed.The ultrasonic guided wave detection test of steel strand stress is designed,and the characteristics of time domain,frequency domain and time-frequency domain of the collected guided wave detection signal are analyzed.The results show that there are obvious differences in the characteristics of ultrasonic guided wave signals under different stress levels.Based on this,a quantitative method of stress characteristic parameters of guided wave signals is proposed to extract the stress characteristics of steel strands.(2)Aiming at the problem that the detection signal is difficult to effectively characterize the stress state of steel strand due to the interference of strong background noise in the actual detection process,a preprocessing method of ultrasonic guided wave signal based on VMD-SVD combined noise reduction is proposed.A VMD signal-noise modal component identification method based on signal correlation analysis is designed to determine the noise-dominated component and the signal-dominated component.On this basis,a singular value effective order selection method based on sample entropy theory is proposed.The noise-dominated component is denoised by singular value decomposition(SVD)and reconstructed with the signal-dominated component to obtain the denoised signal,which reduces the loss of useful information in the denoising process.The simulation signal and the actual collected guided wave detection signal are tested.The results show that the proposed method can effectively restore the original signal characteristics under the condition of ensuring small residual noise interference,and provide the necessary data basis for the identification of steel strand stress level.(3)Aiming at the problem that multi-dimensional features have redundant information and the performance of recognition model is greatly affected by parameters,a SVR steel strand stress recognition method based on RF feature optimization is proposed.Based on the multi-dimensional features of guided wave detection signals under different stress states,the RF algorithm is used to rank the importance of multi-dimensional features,and the optimal feature subset is selected to eliminate redundant features in multi-dimensional features.The sparrow search algorithm(SSA)is used to solve the optimal combination of SVR penalty factor and kernel function parameters,and the optimal feature subset obtained by RF algorithm feature optimization is used as the input of the model to realize the automatic identification of steel strand stress.The experimental results show that the feature subset selected by RF algorithm can effectively improve the recognition accuracy of the model.The SVR model based on SSA algorithm has more advantages in diagnosis time and accuracy than the SVR model based on genetic algorithm and particle swarm optimization algorithm.The research paradigm of stress recognition of steel strand is developed.