Research On Quantitative Identification Of Wire Rope Damage Based On Non-destructive Testing

Steel wire rope has the characteristics of high strength and good flexibility,and is widely used in various fields to play an important role in traction or load-bearing.However,because of its relatively harsh working environment and the load which bears may change suddenly the wire rope will inevitably appear worn,broken wire,rust,and other damage,with the accumulation of time,these damages easily cause safety accidents.Therefore,the damage detection and quantitative identification of wire rope are of great significance to the orderly production of industry and the safety of the life of personnel.Based on this,this thesis aims at wire rope damage detection and quantitative identification,and mainly accomplishes the following work:First of all,the structure of the wire rope and common types of damage are introduced,the wire rope damage detection methods are studied and compared,and the wire rope damage detection sensor is designed according to the principle of the electromagnetic detection method.The damage signal acquisition circuit was also designed.This circuit converts the change in induced electric potential caused by wire rope damage into a change in output signal frequency,which is simpler and has a higher resolution of the detected damage signal compared with the traditional signs acquisition ciruit.Second,the hardware and software of the damage detection system of wire rope are designed.The hardware includes the realization of sensors,the selection of components and the welding and debugging of circuit boards.Software mainly completed the damage signal frequency measurement,data transmission and display functions,in addition,also carried out the development of the upper computer software.Localized Fault(LF)specimens and Loss of Metallic cross-sectional Area(LMA)specimens were fabricated after the system design was completed,and the raw signals of different damage types were obtained through experiments to provide samples for quantitative identification at a later stage.Then,the noise problem of the original signal of wire rope damage is studied,and the signal is processed from the time domain and frequency domain respectively.Pre-processing of signals in the time domain with singular value rejection and de-trending terms;The wavelet threshold denoising method is chosen in the frequency domain to denoise the wire rope damage signal.To obtain the best denoising effect,this thesis constructs a simulation signal according to the characteristics of the original signal of wire rope damage and carries out an experimental analysis of the effect of denoising with different parameters.An improved wavelet thresholding function is also proposed to obtain a better denoising effect compared with soft and hard thresholding functions.Combining the characteristics of the wire rope damage signal,the peak-to-peak value,wave width,waveform area,and wavelet energy were selected as the feature values for quantitative identification of wire rope damage from two perspectives of time domain and frequency domain,and the feature values were normalized.Finally,to achieve quantitative identification of wire rope damage,some methods in the field of neural networks and machine learning are investigated,including Back Propagation(BP)neural networks,Genetic Algorithm(GA)optimized BP neural networks,and Support Vector Machine(SVM),etc.A quantitative identification model of wire rope damage was established according to these methods respectively,and the extracted feature values and their corresponding damage types were used as input and output for quantitative identification experiments of LF-type and LMA-type damage using the established model.The experimental results showed that the GA-BP neural network had the highest recognition accuracy for both LF-type injury and LMA-type injury,with recognition accuracy of 93.5%and 94.6%,respectively.