| Non-conductive material substrates are coated with conductive coatings to make the surface have the ability to conduct current and dissipate the charge,which improves the wear resistance,corrosion resistance,antistatic property and shield electromagnetic interference of the substrates,etc.It has been widely used in the petrochemical industry,medical devices,shipbuilding,aerospace,electronic devices and other industries.In the manufacturing of conductive coatings,the problem of uneven coating thickness is easy to occur,which makes the coating unable to meet the requirements of design and service.Therefore,the quality of the conductive coatings is guaranteed by thickness measurement.Pulsed eddy current(PEC)testing technology has the characteristics of low cost,high precision,high detection speed,abundant detection information and easy automation,which is suitable for the thickness measurement of conductive coatings.In the thickness measurement of conductive coating,the liftoff effect of the probe is one of the key factors affecting the accuracy of PEC testing.The liftoff disturbance makes the conventional characteristics of PEC signals have poor performance in characterizing the thickness of conductive coatings.The liftoff point of intersection(LOI)can suppress the liftoff effect well,but it requires more than two PEC signals to obtain a LOI point,which makes the detection efficiency low.In addition,the conductive coatings with unknown conductivity are often encountered in the detection,and it is necessary to use the inversion method based on the analytical model to determine the thicknesses of the conductive coatings through several iterations by the classical optimization algorithm.However,the classical optimization algorithm is easy to fall into local optimization in the inversion calculation,so the accuracy of inversion thickness is low.Therefore,in-depth research on the suppression of the liftoff effect in PEC testing and the thickness measurement of conductive coatings with unknown conductivity are the key issues to realize the application of PEC testing to the thickness measurement of conductive coatings.In this thesis,supported by the national Natural Science Foundation of China(NSFC)general project,combined with the advantages of the PEC analytical model,through theoretical modeling,simulation analysis,experiment and so on,the thickness measurement method of conductive coating by PEC testing is carried out.The research work mainly includes:(1)The analytical model of PEC testing is established.The expression of impedance change of coil on multilayer conductor is derived by Maxwell equations combined with the separation of variables method,truncated region eigenfunction expansion method and Cheng matrix method.On this basis,the frequency-domain model of PEC testing excited by the voltage source is established,and the time-domain signal of PEC is calculated by the Filon-spline method and fast inverse Fourier transform method.The eddy current experimental system is built and the signal acquisition software is developed.The established analytical model of PEC testing is verified and evaluated by finite element method and experiment.The results show that the simulation signal of the analytical model matches well with the finite element method results and experimental signals,and the relative errors at the peak are-0.02%and 0.1% respectively.(2)An efficient analytical model of PEC testing is established based on the variation characteristic of impedance change with frequency.According to the variation characteristic of the integral function of the impedance change model with frequency,the harmonic impedance changes are calculated by the piecewise interpolation method.Through the transformer model,the variation characteristic of impedance change with frequency is studied,the influence of thickness and conductivity on impedance change is discussed,and the adaptive interpolation method is proposed to calculate the harmonic impedance changes.The results show that the adaptive interpolation method only needs about 16 interpolation nodes,and the relative error of the simulation signal at the peak is about-0.01%,while the relative error between the simulation signal and the experimental signal at the peak is 0.1%.(3)A method for measuring the thickness of conductive coatings by slope characteristics is proposed.The attenuation characteristics of the PEC signals in the thickness change of the conductive coatings are studied.It is found that the slope of the linear attenuation of the PEC in the log-log coordinate system is related to the thickness.Thus,the slope feature is used for thickness characterization,and the power function is adopted to fit the calibration curve of the slope feature for thickness measurement.The effect of the change of excitation parameters and specimen parameters on the slope feature is discussed.The performance of the slope,peak value and zero-crossing time features for thickness measurement is compared.The results show that under the condition of liftoff change of [0,1 mm],the maximum relative error of the measured thicknesses by the slope feature is 12.08%,which is higher than that of peak value and zero-crossing time features.(4)A method for measuring the thickness of conductive coatings based on modeldriven CNN-LSTM network is proposed.CNN-LSTM network can simultaneously extract the spatial and temporal characteristics of signals.A sample set is generated by the efficient analytical model of PEC testing,and the CNN-LSTM network is driven to establish the mapping relationship between PEC signals and thicknesses of conductive coatings under different liftoffs.The influences of the sample data set,liftoff range and sample sampling points on the CNN-LSTM network are analyzed.The experimental signals are reconstructed by the spectral extrapolation method based on BP neural network to reduce the residual errors between experimental signals and simulation signals.The results show that under the condition of liftoff change of [0,1 mm],the maximum relative errors of thickness measurement by the CNN-LSTM network and slope feature are-3.34% and 11.67%,and the accuracy of the CNN-LSTM network in measuring reconstructed signal is higher than that of slope feature.(5)A method for measuring the thickness of conductive coatings with unknown conductivity by the efficient analytical model and improved grey wolf optimization algorithm is proposed.The optimization performance of the grey wolf optimization algorithm is enhanced by introducing chaotic mapping,the dimensional learning hunting search strategy and the non-linear convergence factor.The improved grey wolf optimization algorithm is used to optimize the PEC time-domain signal thickness inversion model,and the fitness value is reduced and the accuracy of inversion parameters is improved.The frequency-domain signal thickness inversion model is proposed to improve the inversion efficiency about 30%,and the decoupling thickness is obtained by introducing a reference.The influence of the residual errors of the experimental signals on parameter optimization is explored.The results show that the maximum relative error of the decoupling thicknesses of the inversion parameters is1.84%,and the accuracy of the optimization parameters can be improved by increasing the signal-to-noise ratio of the experimental PEC signals.The research results obtained in this thesis have important guiding significance for high-performance thickness measurement of conductive coatings by PEC testing,which lays a foundation for thickness measurement of PEC testing under the condition of liftoff change and specimen with unknown conductivity,and provides technical support for the application of PEC testing technology in thickness measurement of conductive coatings.There are 154 pictures,29 tables and 196 references in this thesis. |
PDF Full Text Request