CN116412767A - Novel terahertz thickness measurement method for thermal barrier coating based on model front three-peak driving - Google Patents
Novel terahertz thickness measurement method for thermal barrier coating based on model front three-peak driving Download PDFInfo
- Publication number
- CN116412767A CN116412767A CN202310256929.XA CN202310256929A CN116412767A CN 116412767 A CN116412767 A CN 116412767A CN 202310256929 A CN202310256929 A CN 202310256929A CN 116412767 A CN116412767 A CN 116412767A
- Authority
- CN
- China
- Prior art keywords
- peak
- terahertz
- layer
- thermal barrier
- barrier coating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000012720 thermal barrier coating Substances 0.000 title claims abstract description 53
- 238000000691 measurement method Methods 0.000 title claims abstract description 8
- 238000005259 measurement Methods 0.000 claims abstract description 32
- 238000000034 method Methods 0.000 claims abstract description 32
- 238000004088 simulation Methods 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000009826 distribution Methods 0.000 claims abstract description 18
- 238000012360 testing method Methods 0.000 claims abstract description 16
- 238000004458 analytical method Methods 0.000 claims abstract description 11
- 238000002474 experimental method Methods 0.000 claims abstract description 11
- 239000000919 ceramic Substances 0.000 claims description 46
- 238000013527 convolutional neural network Methods 0.000 claims description 35
- 230000004913 activation Effects 0.000 claims description 14
- 238000010606 normalization Methods 0.000 claims description 13
- 230000003746 surface roughness Effects 0.000 claims description 13
- 238000013507 mapping Methods 0.000 claims description 11
- 230000005540 biological transmission Effects 0.000 claims description 10
- 238000001514 detection method Methods 0.000 claims description 8
- 238000002310 reflectometry Methods 0.000 claims description 8
- 238000001328 terahertz time-domain spectroscopy Methods 0.000 claims description 6
- 239000006185 dispersion Substances 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000008033 biological extinction Effects 0.000 claims description 3
- 238000005520 cutting process Methods 0.000 claims description 3
- 230000001934 delay Effects 0.000 claims description 3
- 229910052751 metal Inorganic materials 0.000 claims description 3
- 239000002184 metal Substances 0.000 claims description 3
- 229910052727 yttrium Inorganic materials 0.000 claims description 3
- BQCADISMDOOEFD-UHFFFAOYSA-N Silver Chemical compound [Ag] BQCADISMDOOEFD-UHFFFAOYSA-N 0.000 claims description 2
- 229910052709 silver Inorganic materials 0.000 claims description 2
- 239000004332 silver Substances 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 abstract description 4
- 238000009659 non-destructive testing Methods 0.000 abstract description 3
- 238000013528 artificial neural network Methods 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 abstract description 2
- 230000003993 interaction Effects 0.000 abstract 1
- 238000011056 performance test Methods 0.000 abstract 1
- 238000011160 research Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 6
- 238000010801 machine learning Methods 0.000 description 6
- 239000011248 coating agent Substances 0.000 description 4
- 238000000576 coating method Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000001066 destructive effect Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- PXHVJJICTQNCMI-UHFFFAOYSA-N Nickel Chemical compound [Ni] PXHVJJICTQNCMI-UHFFFAOYSA-N 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000002366 time-of-flight method Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 241000284466 Antarctothoa delta Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 239000000956 alloy Substances 0.000 description 1
- 229910045601 alloy Inorganic materials 0.000 description 1
- 230000003872 anastomosis Effects 0.000 description 1
- 239000011324 bead Substances 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 229910010293 ceramic material Inorganic materials 0.000 description 1
- 239000007822 coupling agent Substances 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 239000007789 gas Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 230000005865 ionizing radiation Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 229910052759 nickel Inorganic materials 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000007750 plasma spraying Methods 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000002604 ultrasonography Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/02—Measuring arrangements characterised by the use of optical techniques for measuring length, width or thickness
- G01B11/06—Measuring arrangements characterised by the use of optical techniques for measuring length, width or thickness for measuring thickness ; e.g. of sheet material
- G01B11/0616—Measuring arrangements characterised by the use of optical techniques for measuring length, width or thickness for measuring thickness ; e.g. of sheet material of coating
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T50/00—Aeronautics or air transport
- Y02T50/60—Efficient propulsion technologies, e.g. for aircraft
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Geometry (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Length-Measuring Devices Using Wave Or Particle Radiation (AREA)
Abstract
The invention provides a novel terahertz thickness measurement method of a thermal barrier coating based on model front three-peak driving, which comprises the following steps: preparing a thermal barrier coating sample, carrying out a terahertz nondestructive testing experiment, and collecting a time domain signal; exploring the interaction mechanism of terahertz waves and a thermal barrier coating, and analyzing the refractive index and the flight time required by thickness measurement of the front three peaks; constructing a terahertz signal analysis model, and extracting the first three peaks as a training set; designing a weight distribution layer, reducing the difference between experiments and simulations caused by terahertz wave scattering, and updating network parameters in a counter-propagation mode; and extracting the first three peaks of the experimental signal to serve as a test set, and carrying out performance test of the model-driven neural network. The invention researches out the front three peak values through the terahertz propagation mechanism and can be used for measuring the thickness of the thermal barrier coating, so that the coincidence degree of an experimental test set and a simulation training set is ensured, and the accurate measurement of the terahertz thickness of the thermal barrier coating is completed by a deep learning method under the condition that a large amount of samples are not damaged.
Description
Technical Field
The invention relates to the field of terahertz thickness measurement, in particular to a novel terahertz thickness measurement method of a thermal barrier coating based on model front three-peak driving.
Background
The aeroengine is known as a bright bead on an industrial crown, and is an important component of an aeroplane. To increase the thrust-to-weight ratio of the engine, the combustor temperature is continuously increased, which is far above the turbine blade melting point. To solve the above problems, it is necessary to coat the blade surface with a thermal barrier coating to resist the erosion of high temperature gases and particles. Wherein the thickness of the top ceramic layer is related to the thermal resistance, and the heat insulation performance is determined. However, the ceramic layer obtained by the current mainstream atmospheric plasma spraying mode presents a wave-shaped stacking shape, and the phenomenon of uneven thickness exists. If the thickness is too large, the ceramic layer is easy to fall off under the action of the ultra-large centrifugal force of the high-speed rotation of the blade, and the thermal resistance is reduced, the heat conducting performance is enhanced, the surface temperature of the blade matrix is increased, and the blade is easy to damage. Therefore, developing periodic ceramic layer thickness detection is critical in improving turbine blade service life.
Current thermal barrier coating thickness measuring methods are classified into destructive and non-destructive. When the thickness measurement is carried out by adopting a metallographic method and other lossy detection methods, the precision is higher, but the ceramic layer is required to be destroyed, so that the sample is wasted. Non-destructive methods mainly include ultrasound, eddy current, infrared and radiation. When the thickness of the ceramic layer is measured based on an ultrasonic method, a coupling agent needs to be coated on the surface, and the coating is easy to pollute. The thickness of the ceramic layer can be detected by the eddy current method, but the bonding layer is made of nickel-based alloy, so that the conductivity is weaker, and the thickness measuring precision is affected. When the detection is performed based on an infrared method, the phenomenon of uneven thermal emission on the surface of the ceramic layer causes the increase of thickness measurement errors. The detection result obtained by the ray method is visual, but the ionizing radiation is larger, and the harm to the health of personnel is easy to cause.
In recent years, the terahertz nondestructive testing method is widely applied to thickness measurement of thermal barrier coatings, terahertz is an electromagnetic wave with the frequency ranging from 0.1THz to 10THz, has strong penetrability to ceramic materials, and can realize non-contact and non-ionization thickness measurement. The current terahertz ceramic layer thickness measuring method mainly comprises a time-of-flight method, an inversion method and a machine learning method. And acquiring the flight time by positioning the first two reflection peaks, preparing a standard sample piece, extracting the refractive index, and realizing thickness measurement. However, the ceramic layer contains a large number of randomly distributed pores, so that the refractive indexes of all areas are inconsistent, and the thickness measurement error of the time-of-flight method is increased. The inversion method reduces residual errors between simulation and experimental signals by constructing an optimization algorithm iteration solution model, realizes simultaneous measurement of refractive index and thickness, and reduces measurement errors caused by non-uniformity of a ceramic layer microstructure. However, the method needs to execute a large amount of iterative operation every time of thickness measurement, and has low efficiency. In contrast, the machine learning method establishes mathematical mapping between the input features and the thickness through back propagation optimization weights and bias parameters, and the trained machine learning method can realize ceramic layer thickness prediction without iterative operation. However, the back propagation is a full supervision mode, a large number of thermal barrier coating samples need to be destroyed so as to obtain accurate thickness values as labels, sample waste is easily caused, and cost is increased.
Disclosure of Invention
Aiming at the problems, the invention provides a novel method for measuring the thickness of the thermal barrier coating terahertz based on model front three-peak driving, which solves the problem that a large amount of samples are required to be destroyed in the thermal barrier coating thickness measuring method based on data driving machine learning, provides a training set generated by analyzing a model, and explores the mathematical mapping relation between front three reflection peaks and thickness, thereby being used as convolutional neural network input and reducing redundant information with larger experimental difference. Meanwhile, the terahertz wave scattering caused by the ceramic layer material is considered, a weight distribution layer is provided, and the third peak weight with poor anastomosis degree is reduced, so that the terahertz thickness measuring method for the thermal barrier coating based on model front three-peak driving is constructed, high-efficiency and high-precision measurement is realized, and the method has important theoretical significance and engineering application value for online evaluation of the manufacturing quality of the blade coating.
The technical scheme of the invention is as follows:
the method comprises the following steps:
1) Constructing a terahertz signal analysis model considering the surface roughness of the thermal barrier coating, measuring the reflection signal of the silver total reflection mirror by using a terahertz time-domain spectroscopy system, inputting the reflection signal as a reference signal into the analysis model, setting parameters such as different thickness, refractive index, roughness and the like, generating simulation signals, extracting the first three peaks of each simulation signal, and constructing a training set;
2) Preparing a thermal barrier coating test piece, measuring time domain signals at different positions of a sample by using a terahertz time domain spectroscopy system, and extracting the first three peaks of the time domain signals; then, a linear cutting method is adopted to destroy the sample, a metallographic microscope is utilized to obtain accurate thickness values of all detection positions, and a test set is constructed;
3) And (3) establishing a convolutional neural network structure with a weight distribution layer, updating parameters by using the training set in the step (1), and verifying the performance of the convolutional neural network by using the test set in the step (2), so as to realize the online measurement of the thickness of the thermal barrier coating.
The step 1) specifically comprises the following steps:
1.1 Terahertz signal scattering phenomenon is caused according to the surface roughness of the thermal barrier coating, the amplitude of a reflection peak is reduced, the phenomenon is described according to kirchhoff theory,wherein R is s For the specular reflectivity of rough surfaces, R 0 The total reflectivity, lambda is the wavelength of incident terahertz, sigma is roughness, and e is;
1.2 Generating a terahertz simulation signal considering the surface roughness of the thermal barrier coating by utilizing the square relation between the reflectivity and the reflection coefficient:
wherein Y is M The IFFT is inverse Fourier transform, r is the terahertz simulation signal 01 R is the reflection coefficient of terahertz waves on the surface of the ceramic layer 12 For the reflection coefficient of terahertz waves at the interface between the ceramic layer and the metal bonding layer, t 01 For the transmission coefficient of terahertz waves on the surface of the thermal barrier coating, t is t 10 The transmission coefficient of terahertz waves in the thermal barrier coating is i is an imaginary unit, E 0 (omega) is the frequency domain reference signal, ">Is a phase factor, c is the speed of light, ω is the angular velocity of terahertz waves, d 1 For the thickness of the ceramic layer> And +.>The complex refractive indexes of the air, the ceramic layer and the bonding layer are respectively shown, wherein kappa is an extinction coefficient, n is a real part of the refractive index, t is a transmission coefficient, and r is a reflection coefficient. Sigma (sigma) 1 For the surface roughness of the ceramic layer, sigma 2 Is the interface roughness of the ceramic layer and the matrix.
The step 2) specifically comprises the following steps:
2.1 Positioning the first reflection peak by using the maximum value of the terahertz time-domain signal, and increasing the direction positioning peak-valley value between the time delays;
2.2 Using the peak-to-valley time in the step 2.1) as a starting point, and positioning a second reflection peak by using the signal maximum value;
2.3 Solving a first reflection peak and a second reflection peak time delay;
2.4 Adding the second peak-to-peak time and the time delay in the step 2.3), and judging whether the signal amplitude corresponding to the added time is the minimum value of the nearby amplitude; if yes, extracting a third reflection peak, otherwise, positioning the third peak based on a nearby minimum value coordinate;
2.5 The mapping relation between the front three peaks and the thickness is established, and the front three peak signals are input as a convolutional neural network and are used for measuring the thickness of the thermal barrier coating.
The method for establishing the mapping relation between the front three peaks of the terahertz signal and the thickness of the thermal barrier coating specifically comprises the following steps:
according to the terahertz thickness measurement formulaWherein Δt is the time of flight, n 1 The refractive index of the ceramic layer, c is the light speed; delta t calculation requires only the first two peak time differences delta t=t peak1 -t peak2 The method comprises the steps of carrying out a first treatment on the surface of the Wherein Deltat is the time of flight, t peak1 Peak time, t, of the first reflection peak peak2 Peak time for the second peak; n is n 1 Calculating the three peaks before the extractionE n (ω) is the nth reflection peak frequency domain representation, r 01 R is the reflection coefficient of terahertz waves on the surface of the ceramic layer 12 For the reflection coefficient of terahertz waves at the interface between the ceramic layer and the metal bonding layer, t 01 For the transmission coefficient of terahertz waves on the surface of the thermal barrier coating, t 10 The transmission coefficient of the terahertz wave in the thermal barrier coating is shown as the transmission coefficient of the terahertz wave in the thermal barrier coating; since the peak is part of the reflection peak, it is known that the first three peaks carry refractive index information.
The step 3) specifically comprises the following steps:
3.1 A convolutional neural network structure is constructed, and the convolutional neural network structure comprises a weight distribution layer, a convolutional layer, a batch normalization layer, an activation layer and a full connection layer, so that a complete thickness measurement frame is formed;
3.2 Performing parameter training on the convolutional neural network by using the front three-peak simulation data set, establishing mathematical mapping between the front three-peak and the thickness, and ending weight and bias updating after reaching convergence conditions;
3.3 Inputting the first three-peak experimental signals into the convolutional neural network after training, and predicting the thickness value of the ceramic layer.
As a further scheme of the invention: the weight distribution layer constructed in the step 3.1) reduces the third peak weight due to residual error between simulation and experiment caused by terahertz wave dispersion, thereby improving the coincidence degree of a simulation training set and an experiment test set, in addition, the third peak amplitude is negative, the weight upper bound is limited by the ratio of the sum of the first three peak amplitudes and the sum of the absolute amplitudes of the first three peaks, the larger the absolute amplitude of the third peak is realized, the smaller the weight upper bound is, the thickness measuring precision of a convolution neural network to an actual sample is ensured,wherein Y represents the output of the adaptive weight layer, w 3 For the third peak weight, peak i M=1, 2,3 for the mth reflection peak.
As a further scheme of the invention: step 3.1) constructing a convolution layer, a batch normalization layer, an activation layer and a full connection layer, wherein the convolution layer is expressed as: z is Z i,j l =conv (Y, w) +b, wherein Z i,j l Indicating the output of the first convolution layer, i and j are output data dimensions, b is bias, w is weight, and conv indicates convolution operation; the active layer is expressed as: s is S l =σ(Z i,j l ) Wherein S is l For the first active layer output, σ is the active function; the batch normalization layer is expressed as:in the formula, E () is expected, epsilon is a smaller positive number, and the numerical stability is ensured. The full connection is denoted as F l =W l F l-1 +B l Wherein F is l And outputting for the first full connection layer.
The invention aims to solve the problem that a large number of samples are required to be destroyed in the traditional machine learning method, and a thickness measuring method is driven by a model. In the invention, the difference between simulation and experimental signals is reduced mainly by constructing a weight distribution layer, the simulation signals are used as training sets, the convolutional neural network is trained to establish the mathematical mapping between the front three peak values and the thickness, the actual thermal barrier coating sample is prepared, terahertz nondestructive testing and destructive experiments are carried out, and the experimental signals with marks are constructed as test sets, so that the performance verification is realized.
Compared with the existing detection method, the method has the following advantages:
1. aiming at the problem that a large amount of samples are required to be destroyed in the traditional machine learning training, a terahertz signal analysis model is constructed, so that a simulation signal is generated for weight and bias parameter training, and the cost can be reduced;
2. the refractive index and the flight time of the ceramic layer required by thickness measurement are carried by the three front peaks of the terahertz signals are explored, so that the terahertz signals can replace complete signals, redundant information input is reduced, similarity between simulation and experimental data sets is ensured, and algorithm solving efficiency can be improved.
Therefore, the invention provides a novel method for measuring the thickness of the ceramic layer driven by the model, and a large amount of coating damage is reduced. And generating a training set through an analysis model, exploring the propagation mechanism of terahertz waves in the thermal barrier coating, analyzing the former three peaks to replace complete signals to be used as convolutional neural network input, and reducing redundant information with larger difference with experimental signals. Meanwhile, the weight distribution layer is innovatively constructed, residual errors between experiments and simulation are reduced, measurement accuracy is improved, a large amount of samples are not required to be wasted, and thickness measurement errors caused by uneven microstructures are reduced.
Drawings
Fig. 1 is a waveform diagram of a terahertz reference signal.
Fig. 2 is a waveform diagram of terahertz signals carrying information with a thickness of 200 μm and a refractive index of 5-0.05 i.
FIG. 3 is a waveform diagram of a terahertz signal carrying information with a thickness of 300 μm and a refractive index of 4-0.05 i.
FIG. 4 is a model driven convolutional neural network framework diagram.
Fig. 5 is a model accuracy evaluation chart.
Fig. 6 is a measurement accuracy evaluation chart.
FIG. 7 is a graph of a sample thickness profile of a thermal barrier coating ceramic layer.
Detailed Description
In order to clearly illustrate the technical features of the present patent, the following detailed description will make reference to the accompanying drawings.
Specific implementations of the invention are described in detail below in connection with specific embodiments.
As shown in fig. 1 to 5, the embodiment of the invention provides a terahertz thickness measurement method of a thermal barrier coating based on model front three-peak driving, which comprises the following steps:
1) The method comprises the steps of constructing a terahertz signal analysis model considering the surface roughness of a ceramic layer table, and based on a terahertz time-domain spectroscopy system (for example: teraMetrix T-Ray 5000) to obtain a model input reference signal. And analyzing the thickness, refractive index and roughness range of the ceramic layer, and inputting the ceramic layer into a terahertz signal analysis model by combining a reference signal to form a large number of simulation signals. Specifically, terahertz simulation signals (for example, 50000 different refractive index and thickness signals) carrying different thicknesses, refractive indexes and roughness are obtained by changing material parameters, wherein the two terahertz simulation signals respectively carry thickness information of 300 μm and 200 μm, the refractive indexes are 4-0.05i and 5-0.05i, and the root mean square roughness is 5 μm and 10 μm.
The propagation rule of the terahertz wave in the thermal barrier coating is explored, and the flight time and refractive index mechanism of key information required by thickness measurement carried by the front three peaks of the terahertz signal are revealed, so that the front three peaks replace the complete signal to be input as a convolutional neural network.
2) Preparing a thermal barrier coating sample, measuring a terahertz signal at the center of the sample, destroying the sample by adopting a linear cutting mode, measuring a specific thickness value by adopting a metallographic method, and evaluating the coincidence degree of a three peak before simulation and a three peak before experiment to obtain relative errors of 0.1%, 0.4% and 23.9%, wherein the coincidence degree of a third peak is reduced due to terahertz wave scattering.
3) And (3) establishing a weight distribution layer and combining with the convolutional neural network to form a deep learning thickness measurement frame so as to realize ceramic layer thickness measurement.
In the step 1), the construction of the analysis model considering the surface roughness of the ceramic layer specifically comprises the following steps:
and according to the terahertz signal scattering phenomenon caused by the surface roughness of the thermal barrier coating, the amplitude of a reflection peak is reduced, and the phenomenon is described according to kirchhoff theory.Wherein R is s For the specular reflectivity of rough surfaces, R 0 Let λ be the incident THz wavelength and σ be the roughness, for the total reflectivity.
Generating a terahertz simulation signal considering the surface roughness of the thermal barrier coating by utilizing the square relation between the reflectivity and the reflection coefficient:
wherein E is 0 (omega) is the frequency domain reference signal, E R (omega) is a terahertz frequency domain simulation signal, E n (ω) is the n-th reflection peak frequency domain representation, ">Is a phase factor, c is the speed of light, ω is the angular velocity, d 1 For the thickness of the ceramic layer>j=0,1,2,/>And +.>The complex refractive indexes of the air, the ceramic layer and the bonding layer are respectively shown, wherein kappa is an extinction coefficient, n is a real part of the refractive index, t is a transmission coefficient, and r is a reflection coefficient. Sigma (sigma) 1 For the surface roughness of the ceramic layer, sigma 2 Is the interface roughness of the ceramic layer and the matrix.
In step 2), the constructed reflection peak positioning specifically includes:
2.1 The maximum value of the terahertz time-domain signal is used for positioning the first reflection peak, and the time delay is increased to position the peak-valley value in the direction.
2.2 With the peak-to-valley time in step 2.1) as the starting point, locating the second reflection peak with the signal maximum.
2.3 Solving for the first and second reflection peak time delays.
2.4 Adding the second peak-to-peak time to the time delay of step 2.3), and determining whether the signal amplitude corresponding to the added time is the minimum value of the nearby amplitude. If yes, extracting a third reflection peak, otherwise, positioning the third peak based on the nearby minimum value coordinates.
2.5 The mapping relation between the front three peaks and the thickness is established, and the front three peak signals are input as a convolutional neural network and are used for measuring the thickness of the thermal barrier coating.
In the step 2), the mapping relation between the front three peaks of the terahertz signal and the thickness of the thermal barrier coating is established, and the method specifically comprises the following steps:
according to the terahertz thickness measurement formulaWherein Δt is the time of flight, n 1 The refractive index of the ceramic layer and the light velocity of c. Delta t calculation requires only the first two peak time differences delta t=t peak1 -t peak2 . Wherein Deltat is the time of flight, t peak1 Peak time, t, of the first reflection peak peak2 Is the peak time of the second peak. n is n 1 Calculating the three peaks before the extractionE n And (ω) is the nth reflection peak frequency domain representation. Since the peak is part of the reflection peak, it is known that the first three peaks carry refractive index information.
In step 3), the construction of the convolutional neural network with the weight distribution layer specifically comprises the following steps:
3.1 A convolutional neural network structure is constructed, and the convolutional neural network structure comprises a weight distribution layer, a convolutional layer, a batch normalization layer, an activation layer and a full connection layer, so that a complete thickness measurement frame is formed;
3.2 Performing parameter training on the convolutional neural network by using the front three-peak simulation data set, establishing mathematical mapping between the front three-peak and the thickness, and ending weight and bias updating after reaching convergence conditions;
3.3 Inputting the first three-peak experimental signals into the convolutional neural network after training, and predicting the thickness value of the ceramic layer.
In step 3.1), the weight distribution layer specifically includes:
and as the terahertz wave dispersion causes residual error between simulation and experiment, the third peak weight is reduced, so that the coincidence degree of the simulation training set and the experiment testing set is improved. In addition, the third peak amplitude is negative, the weight upper bound is limited by the ratio of the sum of the first three peak amplitudes and the sum of the absolute amplitudes of the first three peaks, the larger the absolute amplitude of the third peak is, the smaller the weight upper bound is, the thickness measuring precision of the convolutional neural network to the actual sample is ensured,wherein Y represents the output of the adaptive weight layer, w 3 For the third peak weight, peak i M=1, 2,3 for the mth reflection peak.
In step 3.1), the construction of the convolution layer, the batch normalization layer, the activation layer and the full connection layer specifically comprises the following steps:
the convolution layer is expressed as: z is Z i,j l =conv (Y, w) +b, wherein Z i,j l Indicating the output of the first convolution layer, i and j are output data dimensions, b is bias, w is weight, and conv indicates convolution operation; the active layer is expressed as: s is S l =σ(Z i,j l ) Wherein S is l For the first active layer output, σ is the active function; the batch normalization layer is expressed as:in the formula, E () is expected, epsilon is a smaller positive number, and the numerical stability is ensured. The full connection is denoted as F l =W l F l-1 +B l Wherein F is l And outputting for the first full connection layer. Specifically, the construction of the convolutional neural network structure comprises 1 weight distribution layer, 4 convolutional layers and 4 batchesThe input signal size is 400 multiplied by 1, the selected learning rate is 0.05, the batch size is 512, the maximum number of rounds is 500, and the specific parameters of each layer are shown in the following table:
convolutional neural network parameter setting table with weight distribution layer
Name of each layer | Parameters (parameters) |
Weight distribution layer | 1 weight, the value ranges are [0,0.1 ]] |
Convolutional layer 1 | The convolution kernel has a size of 7×1, a number of channels of 1, and a number of 16 |
Batch normalization layer 1 | Epsilon is set to 10 -5 |
Activation layer 1 | ReLU activation function |
Convolutional layer 2 | The convolution kernel has a size of 5×1, a number of channels of 1, and a number of 16 |
Batch normalization layer 2 | Epsilon is set to 10 -5 |
Activation layer 2 | Learning parameter ranges [0,1 ] using ReLU activation functions] |
Convolutional layer 3 | The convolution kernel has a size of 3×1, a number of channels of 1, and a number of 16 |
Batch normalization layer 3 | Epsilon is set to 10 -5 |
Activation layer 3 | Learning parameter ranges [0,1 ] using ReLU activation functions] |
Convolutional layer 4 | The convolution kernel has a size of 1×1, a number of channels of 1, and a number of 16 |
Batch normalization layer 1 | Epsilon is set to 10 -5 |
Activation layer 4 | ReLU activation function |
Pooling layer 1 | The size of the pooling layer is 4 multiplied by 1 |
Full tie layer 2 | Number of |
Full tie layer 3 | The number of neurons is 1 |
As shown in fig. 5 and 6, in the embodiment of the invention, actual 17 samples of thermal barrier coating with inconsistent thickness are prepared, thickness measurement is performed at a selection center of each sample, scanning step is 2mm, the collection point number of each sample is 3, total 51 experimental signals are taken as a test set, the test set is input into a convolutional neural network after training, and the average relative error of thickness measurement is 1.05%. Finally, the thermal barrier coating is scanned point by point through a terahertz time-domain spectroscopy system (for example, teraMetrix T-Ray 5000), terahertz experimental signals of all positions of a test piece are obtained and input into a network, and the thickness values of all detection positions are achieved without iterative solution.
According to the invention, a model-driven convolutional neural network thickness measuring frame is constructed, a large number of simulation training sets are generated based on a terahertz signal analysis model, and refractive index and flight time can be measured by exploring the front three peaks of a signal, so that the complete signal is replaced as input, and redundant information with larger difference with an experimental signal is reduced. Meanwhile, the terahertz wave dispersion phenomenon causes poor third peak coincidence, a weight distribution layer is innovated, the third peak weight is intelligently reduced, the similarity between a training set and a testing set is improved, and the thickness measuring precision of the convolutional neural network on an actual ceramic layer is ensured. The method provided by the invention can finish the updating of the weight and the bias parameters of the convolutional neural network without damaging a large amount of samples, thereby reducing the training cost. The invention can efficiently and non-contact detect the thickness of each position of the sample, accurately measure the thickness of the thermal barrier coating with uneven microstructure on the premise of avoiding a large amount of waste of the sample, and has important theoretical significance and engineering application value for the quality evaluation of the coating in the preparation stage.
The foregoing description of the preferred embodiments of the present invention should not be taken as limiting the invention, but rather should be understood to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Other embodiments of the present disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure. This application is intended to cover any adaptations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
Claims (6)
1. The terahertz thickness measurement method for the thermal barrier coating based on model front three-peak driving is characterized by comprising the following steps of:
1) Constructing a terahertz signal analysis model considering the surface roughness of the thermal barrier coating, measuring the reflection signal of the silver total reflection mirror by using a terahertz time-domain spectroscopy system, inputting the reflection signal as a reference signal into the analysis model, setting parameters such as different thickness, refractive index, roughness and the like, generating simulation signals, extracting the first three peaks of each simulation signal, and constructing a training set;
2) Preparing a thermal barrier coating test piece, measuring time domain signals at different positions of a sample by using a terahertz time domain spectroscopy system, and extracting the first three peaks of the time domain signals; then, a linear cutting method is adopted to destroy the sample, a metallographic microscope is utilized to obtain accurate thickness values of all detection positions, and a test set is constructed;
3) And (3) establishing a convolutional neural network structure with a weight distribution layer, updating parameters by using the training set in the step (1), and verifying the performance of the convolutional neural network by using the test set in the step (2), so as to realize the online measurement of the thickness of the thermal barrier coating.
2. The method for terahertz thickness measurement of the thermal barrier coating based on model front three-peak driving according to claim 1, wherein the step 1) specifically comprises the following steps:
1.1 Terahertz signal scattering phenomenon is caused according to the surface roughness of the thermal barrier coating, reflection peak amplitude is reduced, and the phenomenon is described according to kirchhoff theory;wherein R is s For the specular reflectivity of rough surfaces, R 0 Lambda is the incident THz wavelength and sigma is the roughness, which is the total reflectance;
1.2 Generating a terahertz simulation signal considering the surface roughness of the thermal barrier coating by utilizing the square relation between the reflectivity and the reflection coefficient:
wherein Y is M The IFFT is inverse Fourier transform, r is the terahertz simulation signal 01 R is the reflection coefficient of terahertz waves on the surface of the ceramic layer 12 For the reflection coefficient of terahertz waves at the interface between the ceramic layer and the metal bonding layer, t 01 For the transmission coefficient of terahertz waves on the surface of the thermal barrier coating, t 10 The transmission coefficient of terahertz waves in the thermal barrier coating is i is an imaginary unit, E 0 (omega) is the frequency domain reference signal, ">Is a phase factor, c is the speed of light, ω is the angular velocity of terahertz waves, d 1 For the thickness of the ceramic layer> And +.>The complex refractive indexes of the air, the ceramic layer and the bonding layer are respectively shown, wherein kappa is an extinction coefficient, n is a real part of the refractive index, t is a transmission coefficient, and r is a reflection coefficient. Sigma (sigma) 1 For the surface roughness of the ceramic layer, sigma 2 Is the interface roughness of the ceramic layer and the matrix.
3. The method for terahertz thickness measurement of the thermal barrier coating based on model front three-peak driving according to claim 1, wherein the step 2) specifically comprises the following steps:
2.1 Positioning the first reflection peak by using the maximum value of the terahertz time-domain signal, and increasing the direction positioning peak-valley value between the time delays;
2.2 Using the peak-to-valley time in the step 2.1) as a starting point, and positioning a second reflection peak by using the signal maximum value;
2.3 Solving a first reflection peak and a second reflection peak time delay;
2.4 Adding the second peak-to-peak time and the time delay in the step 2.3), and judging whether the signal amplitude corresponding to the added time is the minimum value of the nearby amplitude; if yes, extracting a third reflection peak, otherwise, positioning the third peak based on a nearby minimum value coordinate;
2.5 The mapping relation between the front three peaks and the thickness is established, and the front three peak signals are input as a convolutional neural network and are used for measuring the thickness of the thermal barrier coating. .
4. The method for terahertz thickness measurement of the thermal barrier coating based on model front three-peak driving according to claim 1, wherein the step 3) specifically comprises the following steps:
3.1 A convolutional neural network structure is constructed, and the convolutional neural network structure comprises a weight distribution layer, a convolutional layer, a batch normalization layer, an activation layer and a full connection layer, so that a complete thickness measurement frame is formed;
3.2 Performing parameter training on the convolutional neural network by using the front three-peak simulation data set, establishing mathematical mapping between the front three-peak and the thickness, and ending weight and bias updating after reaching convergence conditions;
3.3 Inputting the first three-peak experimental signals into the convolutional neural network after training, and predicting the thickness value of the ceramic layer.
5. The terahertz thickness measurement method of the thermal barrier coating based on model front three-peak driving of claim 1 is characterized in that the weight distribution layer constructed in the step 3.1) reduces the third peak weight due to residual error between simulation and experiment caused by terahertz wave dispersion, thereby improving the coincidence degree of a simulation training set and an experiment test set; in addition, the third peak amplitude is negative, the weight upper bound is limited by the ratio of the sum of the first three peak amplitudes and the sum of the absolute amplitudes of the first three peaks, so that the larger the absolute amplitude of the third peak is, the smaller the weight upper bound is, and the thickness measuring precision of the convolutional neural network on an actual sample is ensured, wherein Y=w 3 ·peak 3 Wherein Y represents the output of the adaptive weight layer, w 3 For the third peak weight, peak i M=1, 2,3 for the mth reflection peak.
6. The ceramic layer thickness terahertz measurement method based on a convolutional neural network according to claim 1, wherein step 3.1) constructs a convolutional layer, a batch normalization layer, an activation layer and a full connection layer, and the convolutional layer is expressed as: z is Z i,j l =conv (Y, w) +b, wherein Z i,j l Indicating the output of the first convolution layer, i and j are output data dimensions, b is bias, w is weight, and conv indicates convolution operation; the active layer is expressed as: s is S l =σ(Z i,j l ) Wherein S is l For the first active layer output, σ is the active function; the batch normalization layer is expressed as:in the formula, E () is expected, epsilon is a smaller positive number, and the numerical stability is ensured; the full connection is denoted as F l =W l F l-1 +B l Wherein F is l And outputting for the first full connection layer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310256929.XA CN116412767A (en) | 2023-03-16 | 2023-03-16 | Novel terahertz thickness measurement method for thermal barrier coating based on model front three-peak driving |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310256929.XA CN116412767A (en) | 2023-03-16 | 2023-03-16 | Novel terahertz thickness measurement method for thermal barrier coating based on model front three-peak driving |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116412767A true CN116412767A (en) | 2023-07-11 |
Family
ID=87049013
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310256929.XA Pending CN116412767A (en) | 2023-03-16 | 2023-03-16 | Novel terahertz thickness measurement method for thermal barrier coating based on model front three-peak driving |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116412767A (en) |
-
2023
- 2023-03-16 CN CN202310256929.XA patent/CN116412767A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Fromme et al. | Measurement of the scattering of a Lamb wave by a through hole in a plate | |
Jingjing et al. | A Lamb wave quantification model for inclined cracks with experimental validation | |
Zhang et al. | Longitudinal wave scattering from rough crack-like defects | |
Fahim et al. | Model-based damage reconstruction in composites from ultrasound transmission | |
Benstock et al. | The influence of surface roughness on ultrasonic thickness measurements | |
CN113720907B (en) | Composite material layered damage identification method for contour and depth sequence identification | |
CN106949861B (en) | A kind of method of non-linear ultrasonic on-line monitoring metal material strain variation | |
Hassan et al. | State-of-the-art review on the acoustic emission source localization techniques | |
CN107764195A (en) | Coating high accuracy thickness detecting method based on Fresnel matrix THz ripple propagation models | |
CN111553877A (en) | Damage identification and service life evaluation method based on terahertz ceramic matrix composite blade | |
Nokhbatolfoghahai et al. | Evaluation of the sparse reconstruction and the delay-and-sum damage imaging methods for structural health monitoring under different environmental and operational conditions | |
JP3413603B2 (en) | Measuring method of residual stress | |
Burghardt et al. | Robot-operated quality control station based on the UTT method | |
Morii et al. | A new inverse algorithm for tomographic reconstruction of damage images using Lamb waves | |
Dou et al. | Simultaneous measurement of thickness and sound velocity of porous coatings based on the ultrasonic complex reflection coefficient | |
Lee et al. | Demonstration of model-assisted probability of detection framework for ultrasonic inspection of cracks in compressor blades | |
Hameed et al. | Transverse damage localization and quantitative size estimation for composite laminates based on Lamb waves | |
Boffa et al. | About the combination of high and low frequency methods for impact detection on aerospace components | |
Wei et al. | Damage quantification of aluminum plates using SC-DTW method based on Lamb waves | |
CN114427838A (en) | Method and system for predicting and evaluating thickness of medium based on reflection terahertz spectrum | |
CN116412767A (en) | Novel terahertz thickness measurement method for thermal barrier coating based on model front three-peak driving | |
CN116562118A (en) | Terahertz measurement method for thickness of ceramic layer of thermal barrier coating based on Elman neural network | |
Cao et al. | In situ elastic constant determination of unidirectional CFRP composites via backwall reflected multi-mode ultrasonic bulk waves using a linear array probe | |
Kurc et al. | Non-contact robotic measurement of jet engine components with 3D optical scanner and UTT method | |
Zhu et al. | Estimation study of structure crack propagation under random load based on multiple factors correction |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |