| With the rapid development of Chinese economy,the number of large bridges is increasing.After the completion of the bridges,the quality of the bridges will change with the external environment,so it is necessary to evaluate the safety of the bridge structure and provide suggestions for the maintenance of the bridge health.In this paper,through the construction of simply supported beam bridge structure health monitoring platform,real-time monitoring of simply supported beam bridge structure,and then analyze the monitoring data and predict the damage of bridge structure,provide technical support for the subsequent complete bridge monitoring system.The main research contents are as follows:(1)In this paper,a damage identification method for simply supported beam bridge is proposed based on the combination of structural vibration data and machine learning.First,an example of simply supported beam bridge is introduced,144 simply supported beam elements and a spring element are used to model the bridge.Independent random vibration sources with different amplitudes are used to excite the bridge at three points.Data sets of undamaged and different damage levels are obtained through 47 acceleration sensors.Then,data preprocessing algorithms such as Maximum Correlation Kurtosis Deconvolution(MCKD),extreme statistics,fast Fourier transform(FFT)and machine learning classification algorithms such as Support Vector Machine(SVM),Decision Tree(DT),k-Nearest Neighbor(k NN)and Discriminant Analysis(DA)are introduced to identify damage of simply supported beam bridge.Finally,the experimental results show that the features of the data set can be obtained after preprocessing,and the machine learning classification algorithm is used to classify the new data set containing features,which verifies the effectiveness of the data preprocessing algorithm to extract features,the most suitable damage identification model for the simply supported beam bridge can be obtained,and the accuracy of SVM can reach 92%.(2)In this paper,a concrete crack classification model based on the combination of acoustic emission(AE)monitoring and machine learning classification algorithm is proposed.First,an example is introduced:five AE sensors are installed asymmetrically in a group of 15×15×15 cm~3 concrete cube,the sensor is fixed on the concrete surface with silicone oil,and the acoustic emission signal data is obtained by applying three kinds of loads(shear,tension and mixing).Then,Gaussian smoothing filter is used to process the original signal data,and then two signal processing methods are used:one is to process the smoothed data by Hilbert-Huang Transform(HHT),Empirical Mode Decomposition(EMD)and Variational Mode Decomposition(VMD),and then feature extraction is performed,including Instantaneous Frequency(IF),Spectral Entropy(SE),and so on Spectral kurtosis(SK)and Spectral L2/L1 Norm(SLN);The other is to use wavelet decomposition after smoothing data by grid search optimize parameters.Finally,this paper inputs the data into the machine learning algorithm.By comparing the accuracy and stability of the two processing methods,it is found that the accuracy of traditional machine learning can reach more than 95%after Gaussian smoothing and wavelet decomposition.Among traditional machine learning algorithms,subspace k NN after super parameter optimization has the best classification effect and the most stable performance,it shows that the method is effective and feasible for the classification and prediction of concrete structure crack of simply supported beam bridge.(3)This paper uses Django web framework to design an online health monitoring system for simply supported beam bridge.By adding machine learning toolbox to the system,the functions of preprocessing,online visualization and classification analysis of monitoring data can be realized,which provides a new idea and method for the field of simply supported beam structural health monitoring. |
PDF Full Text Request