HVD-Based Damage Detection for Prefabricated Composite Wallboard Connectors
Description
According to the experimental results, the relative curvature value of HVD marginal spectral entropy can be calculated and the conclusion can be drawn : ( 1 ) Under the single damage condition, the relative curvature of the HVD marginal spectrum entropy at the damaged connector shows a significant change, and its value reaches the peak level. The research shows that when the position of the damaged connector remains unchanged, with the gradual aggravation of the damage degree, the relative curvature change range of the marginal spectral entropy of the connector is further expanded, thus significantly improving the identifiability of the damage position. ( 2 ) For multiple damages, the relative curvature value of marginal spectral entropy can determine the location of the damaged connector, but the judgment of the degree of damage is insufficient. Based on the test results, the entropy difference of HVD power spectral density can be calculated and the following conclusions can be drawn : ( 1 ) For single damage, the difference of HVD power spectral density entropy at the damaged connector changes obviously, and the value is the maximum. When the location of the damaged connector is the same, with the gradual increase of the degree of damage, the change of the power spectral density entropy difference of the damaged connector is more significant, and the recognition of the damage location is improved. ( 2 ) For multiple damages, the power spectral density entropy difference can not only determine the location of the damaged connectors, but also distinguish the damage degree of different damaged connectors. The analysis based on the relative curvature of the marginal spectral entropy shows that the method can accurately identify the damage location in the single damage condition, and the curvature difference increases significantly with the increase of the damage degree, but it is difficult to distinguish the damage degree difference under the multi-damage condition. In contrast, the power spectral density entropy difference method is not only sensitive to single damage location, but also can effectively distinguish the damage degree of different connectors in multiple damage scenarios, showing higher comprehensive performance.
Files
Steps to reproduce
The data in this database are derived from scale experiments. In this experiment, 10 working conditions with different damage conditions were designed. The rubber hammer was used to hit the center of the wall plate to apply the impact load, and the impact was perpendicular to the plate surface. Every two seconds, after the vibration of the specimen was attenuated to a stable state, the above excitation and acquisition process was repeated, and the number of repetitions was not less than ten times. In this experiment, four IEPE acceleration sensors are used. The measurement range of the sensor is set to 0-100m / s2, and the actual collected data is about 70m / s2. The DH5930 mobile data acquisition equipment is used for real-time acquisition of dynamic signals. The sampling frequency of the acceleration acquisition system is determined to be 5000Hz. In order to ensure the effectiveness of data acquisition, the acquisition range should be controlled at 70 % to 80 % of the acquisition equipment range, and the best signal-to-noise ratio can be obtained at this time. In this study, before the formal data acquisition, it is necessary to complete the connection between the acceleration acquisition system and the four sensors, and preheat the equipment for about 30 minutes to improve the accuracy of the data. In the signal processing part, the data collected by the four acceleration sensors are imported into the MATLAB environment for processing. In order to separate the dual-channel signals generated by each excitation, the peak detection function is used to locate the extreme points in the acceleration sequence, so as to determine the time domain position of the transient signal. In view of the fact that the transient signal has the characteristics of energy concentration and stable process, the error caused by the truncation operation can be ignored. The movable rectangular window function is used to intercept the signal to ensure that the transient signal obtained by the sensor T-1 to T-4 forms a one-to-one correspondence, and then the Fourier transform is performed on the signal time-frequency diagram to obtain the corresponding signal frequency domain diagram. In this study, when using the HVD method to extract the feature vector of the collected signal, the AMD filtering cut-off frequency is set to [ 170,180 ] Hz, and the first-order vibration information is extracted.