| Turbine blade is the key component of aero-engine and gas turbine.Due to long-term working in high temperature and complex stress environment,it will be severely eroded by high-temperature oxidation and hot corrosion.Thermal barrier coating(TBC)is a ceramic coating deposited on superalloy.Using the excellent heat insulation performance of the coating,a large temperature drop is produced between the high temperature gas and the base alloy,so as to improve the combustion chamber temperature and the thermal efficiency of the engine.Thermal barrier coating system is a multi-layer complex structure.It will seriously threaten the safety of aero-engine if it fails.Therefore,it is necessary to obtain the parameter status of TBC system in time to ensure aircraft safety.This paper mainly studies the multi parameter inversion method of TBC system based on electromagnetic / capacitive composite detection.Based on the detection information obtained by the composite sensor in electromagnetic and capacitive modes separately,three different parameter inversion methods,namely analytical analysis,grid search and machine learning,were studied to realize the inversion of key parameters such as ceramic layer thickness,relative dielectric constant and bonding layer conductivity.The main work and innovations are as follows.1.The principle of electromagnetic / capacitive composite detection of TBC system was described.The finite element simulation model was established.The relationship between the parameters of TBC system and the detection signal was analyzed.The distribution of magnetic field and electric field in electromagnetic and capacitive working modes of composite sensor was analyzed respectively.The finite element simulation model of thermal barrier coating detection system was established.The influences of lifting off,thickness of ceramic layer,conductivity and thickness of adhesive layer on transfer impedance were analyzed.The parameters of transfer impedance and lift off,ceramic layer and bonding layer were obtained.The relationship between equivalent capacitance and lift off,thickness of ceramic layer and relative permittivity was obtained by analyzing the influences of parameters such as lift off,thickness of ceramic layer and relative permittivity under capacitance mode,which lays a foundation for the establishment of detection model.2.The detection model and parameter inversion model of TBC based on electromagnetic / capacitive composite sensor were established,and the direct inversion method based on analytical analysis and information fusion were proposed.Based on the theoretical analysis,the analytical relationship between the detection signal and the parameters of TBC system under electromagnetic and capacitive detection modes was established,the detection model and parameter inversion model of TBC were established,and the identification of model parameters was completed according to the simulation results.A multi parameter inversion method of thermal barrier coatings based on information fusion were proposed.Which is to fuse the measured signals in two modes at feature level.At the same time,the permittivity and conductivity of the adhesive layer were inversed simultaneously.Because the analytic analysis method needs to obtain part of the prior information of the sample,it has some limitations in application,which needs to be used with other methods to obtain better results.3.The database of measurement grid for parameter inversion was established,and a multi parameter inversion method combining two-dimensional grid search and analytical analysis was proposed.The three key parameters of ceramic layer thickness,relative dielectric constant and bonding layer conductivity of thermal barrier coating were inversed step by step.In order to solve the application limitation of analytical analysis method,a two-dimensional grid search algorithm was proposed,which could be combined with analytical analysis method to realize the inversion of multiple key parameters of thermal barrier coatings.Firstly,two parameters of ceramic layer thickness and bonding layer conductivity were inversed by grid search.Then,the thickness information obtained from inversion was combined with the measured equivalent capacitance,and the relative permittivity of ceramic layer was inversed by analytical method.The experimental results show that when the maximum inversion errors of the thickness of ceramic layer,relative permittivity of ceramic layer and conductivity are separately2.67%,1.76% and 1.83%.Nevertheless,this step-by-step inversion method will inevitably lead to error transmission.If the detection signal can be used to obtain the values of three parameters at the same time,the influence of substitution error could be avoided.4.A neural network model for parameter inversion of thermal barrier coatings was established,and a multi parameter inversion method based on machine learning was proposed.The simultaneous inversion of ceramic layer thickness,relative dielectric constant and bonding layer conductivity was realized,and the inversion accuracy was improved.In order to improve the accuracy of inversion,a multi parameter inversion method of TBC system based on machine learning was proposed,and the inversion methods based on RBF neural network and BP neural network were studied and compared.A three-input two-output inversion model was constructed for the inversion of ceramic layer thickness and dielectric constant,and a five-input three-output inversion model for the inversion of three key parameters of ceramic layer thickness,dielectric constant and bonding layer conductivity were established.The experimental results show that when BP neural network only inverts two parameters,the training time is about 2 minutes;the maximum inversion errors of the thickness of ceramic layer and relative permittivity are separately 1.33% and 1.97%;in the inversion of three parameters,the training time is about 5 minutes.the maximum inversion errors of the thickness of ceramic layer,the relative permittivity and conductivity are separately 1.33%,1.98% and 0.16%respectively.BP neural network algorithm training time is significantly shorter than the radial basis function neural network algorithm,and the inversion accuracy is relatively high.It has the advantages of high efficiency and high precision,and is a good choice for parameter inversion.5.The automatic detection platform of TBC system was built,and the accurate detection information was obtained for parameter inversion.The effectiveness and correctness of the inversion method were verified by the experimental results.In order to obtain accurate test data and verify the correctness of various parameter inversion methods proposed above,a thermal barrier coating system detection platform including thermal barrier coating simulation specimen,electromagnetic / capacitive composite sensor array,multi-channel switching unit and impedance analyzer was built.The virtual instrument technology was used to control the multi-channel switching unit.Each detection unit in the electromagnetic / capacitive composite sensor array was connected to the input end of the impedance analyzer to measure the transfer impedance or equivalent capacitance of the sensor.The measurement results of the impedance analyzer were automatically read through the GPIB interface.A automatic measurement system with high efficiency of parameter setting,automatic measurement and data storage was constructed.The effectiveness and accuracy of the proposed multi parameter inversion method for thermal barrier coatings were experimentally verified by using the measured data.The experimental results show that the above methods can achieve multi parameter inversion. |
PDF Full Text Request