The Design Of Nondestructive Testing Experimental System Based On Multi-frequency Eddy Current

Nowadays,people still cannot get rid of their dependence on natural resources and pipeline transportation is the most common way when it comes to the delivery of these resources like oil,coal,gases,etc.In this case,the safety of pipe transportation becomes extremely important to the development of an economy of a nation.It also has a strategic significance to a country suffering in wartime.However,it may cause problems when there’s a undetected leakage without repaired effectively.The leakage may lead to economy losses,environment pollution,life security cases or even worse.At present,most detection systems used in pipeline leakage are big in size,expensive.So it is of great significance to design a smaller,lower-cost leakage detection system.This thesis is carried out as bellow:Firstly.It is based on the principle of practice guided by theory and the reality that the worldwide immature eddy detection theory and incomplete research system we have.The experiment in this thesis is simulating the corresponding magnetic field distribution of excitation coil in different conditions by using ANSYS finite element simulation software.Secondly.With the difficulty on extracting detection signals while a probe coil of multi-frequency eddy current testing system was given multi-frequency signals at the same time,in this thesis we applied a new method that multiple signals with different frequency were input into multiple probe coil sensors and each sensor refers to one signal.It will be much easier to detect the expected signals in this way.Thirdly.Aiming at the commonly existed lift-off effect on eddy current testing,we presented a method to combine parallel resonance and unbalanced bridge to inhibit lift-off effect.The lift-off effect can easily cause incorrect detection results and this method to some extend can inhibit the lift-off effect.Fourthly.Due to the fact that the detecting on the same defect showed different characteristics when the probe coil is tested with different frequency signals,this thesis was using the Extreme Learning Machine to train the data and establish the model.The average correct rate reaches 87%for the tested samples which indicates that it can be used to identify and detect the depth of defects as well as classification.