Feature Vector Extraction And Evaluation Of Damage Information In One-dimensional Component Based On The Stress Wave Signal

One-dimensional components are common practical structures, such as piles, bearings,construction beams, pipelines and bolt, etc, all of them are one-dimensional components.Under the combined effects of various, there are often kinds of defects in the one-dimensional components after completion,those defects have influence in the quality,safety and service life of the one-dimensional components and bring projects risks and irreparable personal and property losses. Presence or absence and degree of defects in one-dimensional components should be detected after construction is completed. One-dimensional components detection techniques include stress wave reflection method, ultrasonic transmission method, high strain detection method and drilled core method. Stress wave reflection method is the most commonly used method; its equipments are simple and save resources.The current feature vector extracting and evaluation has some deficiencies: the theory of characteristics is always the Fourier analysis and wavelet analysis, this makes the high frequency stress wave signal cannot be well used; the current feature vector extraction method cannot extract the complete signal time information and multiple defects features; the evaluation methods have to be made and developed.In this paper, we use wavelet packet analysis, quantitative information entropy and the time window to extract the feature vector of stress wave signal, use the multi-parameter analysis of gray system theory, feature vector standard deviation and Euclidean distance distribution entropy to evaluate the feature vector on the reliability, stability and separability.We select the stress wave signal of pile NDT as the study object, for the characteristics of stress wave signal and the requirements of frequency and time, select sym8 wavelet packet analysis method, extract the defects information at low and high frequency waves.We propose he concept of quantitative information entropy as the new feature of stress wave signal, create the multidimensional feature vector with the quantitative information entropy, energy and variance feature vectors.We propose the time window method to extract the feature vector. Establish the norms of the time window width and the length of the step. Use the time window method to build the multidimensional feature vector of the stress wave signal.Introduce the method of multi-parameter associated analysis, the method of feature standard deviation and the method of Euclidean distance distribution entropy to evaluate the reliability, stability and separability of the multidimensional feature vector.