Research On Pile Defect Identification Based On Multi-Higher Order Moment And Defect Location

Nowadays,the stress wave reflection method has became the main detection way of pile foundation census due to its many advantages.As a further improvement of the pile defect recognition accuracy,and for classifying integrated pile and defective piles,a novel feature based on higher order statistical moments is proposed in this paper.Further more,in order to locate defects,a method of finding local peaks of signals is proposed in this study.Through processing stress wave reflected signals,using numerical simulation,feature extraction,pattern recognition,finite element simulation verification and a series of methods in this study at the same time.Eventually,the purpose of identifying and locating internal defects of piles automatically is achieved.Based on the one-dimensional stress wave theory of pile-soil system,we establish the pile-soil system model,and the wave equation has been derived.According to the initial condition and the boundary condition,the finite difference method is used to solve the equation.Besides,curves of velocity of integrated piles and various types of defective piles are obtained by simulation experiment.And the comparative analysis with the stress wave reflected signals of different types of piles is performed.In order to improve the accuracy of stress wave reflection method,wavelet packet analysis is used to process the stress wave reflected signals.DB7 wavelet is applied to decompose the signals into the third level,and a sliding window which moves with the time is constructed.The sliding window is applied to extract higher order moment features of the eight wavelet packet bands on the third level.The extracted higher order moment features are fused by principal component analysis and defined as multi-higher order moment features.The multi-higher order moment feature is input into support vector machine for classification and recognition.By comparing the classification results with the traditional pile identification features(power spectral density,variance and entropy),the experimental results show that the multi-higher order moment feature has a 23.33% improvement in classification accuracy than the traditional pile detection features.And compared with the 42-dimensional feature,the multi-higher order moment feature has a shorter classification time by 0.76 seconds.The method of locating the internal defects of piles by finding local peaks of signals is proposed in this study.This part of the research is based on the three-dimensional structure.The three-dimensional finite element model of the pile-soil system is built by ABAQUS software,and the defects are set at different positions inside the piles.The defective piles include shrinking pile,expanding pile,segregated pile and fractured pile.According to the simulation results of finite element method,the defect location is calculated and it is compared with the real location of the defect.The experimental results show that the localization error can be controlled within 8.50% by finding the local peak of the signal.At the same time,the finite element simulation curves of the pile with different positions of defects are drawn.The influence on the velocity response surves and positioning accuracy which are caused by different defective position has been analyzed.