CN112200788B - High-temperature deformation measuring device and method - Google Patents
High-temperature deformation measuring device and method Download PDFInfo
- Publication number
- CN112200788B CN112200788B CN202011108313.0A CN202011108313A CN112200788B CN 112200788 B CN112200788 B CN 112200788B CN 202011108313 A CN202011108313 A CN 202011108313A CN 112200788 B CN112200788 B CN 112200788B
- Authority
- CN
- China
- Prior art keywords
- image
- test piece
- neural network
- convolutional neural
- images
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000012360 testing method Methods 0.000 claims abstract description 149
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 120
- 230000005855 radiation Effects 0.000 claims abstract description 93
- 238000010438 heat treatment Methods 0.000 claims abstract description 87
- 238000012545 processing Methods 0.000 claims abstract description 62
- 238000001514 detection method Methods 0.000 claims abstract description 56
- 238000005259 measurement Methods 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims description 80
- 238000010586 diagram Methods 0.000 claims description 31
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000007689 inspection Methods 0.000 claims 1
- 230000008859 change Effects 0.000 abstract description 13
- 238000013528 artificial neural network Methods 0.000 abstract description 11
- 230000008878 coupling Effects 0.000 abstract 1
- 238000010168 coupling process Methods 0.000 abstract 1
- 238000005859 coupling reaction Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 32
- 230000008569 process Effects 0.000 description 22
- 239000000463 material Substances 0.000 description 15
- 230000000694 effects Effects 0.000 description 12
- 230000001808 coupling effect Effects 0.000 description 9
- 238000011156 evaluation Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 230000002349 favourable effect Effects 0.000 description 4
- 238000009529 body temperature measurement Methods 0.000 description 3
- 238000000691 measurement method Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 229910052799 carbon Inorganic materials 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 239000002131 composite material Substances 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- HBMJWWWQQXIZIP-UHFFFAOYSA-N silicon carbide Chemical compound [Si+]#[C-] HBMJWWWQQXIZIP-UHFFFAOYSA-N 0.000 description 2
- 229910010271 silicon carbide Inorganic materials 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007664 blowing Methods 0.000 description 1
- 230000001351 cycling effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000001747 pupil Anatomy 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000002834 transmittance Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N3/00—Investigating strength properties of solid materials by application of mechanical stress
- G01N3/02—Details
- G01N3/06—Special adaptations of indicating or recording means
- G01N3/068—Special adaptations of indicating or recording means with optical indicating or recording means
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N3/00—Investigating strength properties of solid materials by application of mechanical stress
- G01N3/60—Investigating resistance of materials, e.g. refractory materials, to rapid heat changes
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2203/00—Investigating strength properties of solid materials by application of mechanical stress
- G01N2203/003—Generation of the force
- G01N2203/0057—Generation of the force using stresses due to heating, e.g. conductive heating, radiative heating
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2203/00—Investigating strength properties of solid materials by application of mechanical stress
- G01N2203/0058—Kind of property studied
- G01N2203/0069—Fatigue, creep, strain-stress relations or elastic constants
- G01N2203/0075—Strain-stress relations or elastic constants
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2203/00—Investigating strength properties of solid materials by application of mechanical stress
- G01N2203/02—Details not specific for a particular testing method
- G01N2203/022—Environment of the test
- G01N2203/0222—Temperature
- G01N2203/0226—High temperature; Heating means
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2203/00—Investigating strength properties of solid materials by application of mechanical stress
- G01N2203/02—Details not specific for a particular testing method
- G01N2203/06—Indicating or recording means; Sensing means
- G01N2203/0641—Indicating or recording means; Sensing means using optical, X-ray, ultraviolet, infrared or similar detectors
- G01N2203/0647—Image analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Abstract
The present disclosure relates to a high temperature deformation measuring device and method. The device comprises: processing equipment, image acquisition equipment, heating equipment. The heating equipment is used for generating hot air flow and heating the test piece through the hot air flow; the image acquisition equipment is used for acquiring a detection image of the test piece in the heating process of the test piece; the processing equipment is used for carrying out residual error identification processing on the detection image through a convolutional neural network to obtain residual error information of the detection image; obtaining a deformation image of the test piece according to the detection image and the residual error information; and obtaining a deformation field of the test piece according to the deformation image. According to the high-temperature deformation measuring device disclosed by the embodiment of the disclosure, the coupling influence of factors such as uneven image brightness caused by high-temperature heat radiation and air refractive index change caused by hot air flow disturbance can be reduced through the convolution neural network, a clearer experimental image is obtained, and the measurement precision of a deformation field is improved.
Description
Technical Field
The disclosure relates to the field of computers, in particular to a high-temperature deformation measuring device and method.
Background
In the fields of aerospace, material processing and the like, some key structures or parts are often used in high-temperature environments, and the performance of high-temperature materials becomes an important factor for restricting the stable work of equipment. The mechanical property evaluation and measurement of the material under the high-temperature environment have important significance, and the method has important effects on the performance evaluation, the service life prediction and the development of product design of the material. The material deformation measurement under the high-temperature environment becomes an important basis for evaluating the mechanical property of the material under the high-temperature environment.
In recent years, the digital image correlation method is applied to high-temperature testing due to the advantages of simple optical path setting, high precision, good environmental adaptability and the like. However, the optical testing challenges brought by the high-temperature extreme environment still need to be solved urgently, such as the coupling effect of factors such as uneven image brightness caused by high-temperature heat radiation, air refractive index change caused by hot air flow disturbance and the like, so that the identification precision of the image acquired by the high-temperature digital image correlation technology is difficult to improve, and the precision of deformation field calculation is seriously influenced.
Disclosure of Invention
In view of the above, the present disclosure provides a high temperature deformation measurement apparatus and method.
According to an aspect of the present disclosure, there is provided a high temperature deformation measuring apparatus, the apparatus including: processing equipment, image acquisition equipment, heating equipment. The heating apparatus is for: generating hot air flow, and heating the test piece through the hot air flow; the image acquisition device is configured to: acquiring a detection image of the test piece in the heating process of the test piece; the processing device is configured to: carrying out residual error identification processing on the detection image through a convolutional neural network to obtain residual error information of the detection image, wherein the residual error information is used for representing the influence of thermal radiation and thermal current disturbance on the test piece; obtaining a deformation image of the test piece according to the detection image and the residual error information; and obtaining a deformation field of the test piece according to the deformation image.
In a possible implementation manner, performing residual identification processing on the detected image through a convolutional neural network to obtain residual information of the detected image, including: carrying out segmentation processing on the detection image to obtain a plurality of sub-images of the detection image; and respectively carrying out residual error identification processing on the plurality of sub-images through a convolutional neural network to obtain residual error images of the plurality of sub-images, wherein the residual error information of the detection image comprises the residual error images of the plurality of sub-images.
In a possible implementation manner, obtaining a deformation image of the test piece according to the detection image and the residual information includes: performing difference processing on the plurality of sub-images and residual images of the plurality of sub-images respectively to obtain deformed sub-images; and splicing the deformation sub-images to obtain the deformation image.
In a possible implementation manner, obtaining a deformation field of the test piece according to the deformation image includes: obtaining image information of a first preset channel of the deformed image; and obtaining the deformation field according to the image information of the first preset channel.
In a possible implementation manner, the heating device is disposed between the test piece and the image acquisition device, and an included angle between a wind direction of a hot air flow generated by the heating device and a normal plane of the test piece is a preset angle, wherein the preset angle is greater than 0 ° and smaller than 90 °.
In a possible implementation manner, the apparatus further includes a temperature measuring device, configured to acquire temperature information of the test piece, and the image acquiring device is further configured to: acquiring a plurality of first sample images during heating of the test piece, the processing device being further configured to: the convolutional neural network is trained through a plurality of first sample images and temperature information acquired by the first sample images at the same time, when the first sample images are acquired, the test piece is arranged between the heating equipment and the image acquisition equipment, and an included angle between the wind direction of hot air flow generated by the heating equipment and a normal plane of the test piece is 90 degrees.
In one possible implementation, the training the convolutional neural network by using a plurality of first sample images and temperature information acquired simultaneously with the first sample images includes: acquiring image information of a second preset channel of the first sample image; obtaining a temperature field of the surface of the test piece according to the image information of the second preset channel and the temperature information obtained simultaneously with the first sample image; obtaining a thermal radiation map corresponding to the first sample image according to the temperature field; inputting the first sample image into the convolutional neural network to obtain a thermal radiation training image; determining a first network loss function value of the convolutional neural network according to the thermal radiation training diagram and the thermal radiation diagram; training the convolutional neural network according to the first network loss function value.
In one possible implementation, the image acquisition device is further configured to: acquiring a room temperature image while the test piece is not heated and acquiring a plurality of second sample images during heating of the test piece, the processing device being further configured to: through the normal temperature image and a plurality of second sample images, it is right the convolutional neural network trains, when obtaining the second sample image, heating equipment set up in the test piece with between the image acquisition equipment, the wind direction of the hot gas flow that heating equipment produced with the normal plane of test piece between the contained angle is 0.
In one possible implementation manner, training the convolutional neural network through the normal-temperature image and the plurality of second sample images includes: performing difference processing on the second sample image and the normal temperature image to obtain a hot air flow disturbance noise map of the test piece; inputting the second sample image into the convolutional neural network to obtain a hot air flow training image; determining a second network loss function value of the convolutional neural network according to the hot airflow training graph and the hot airflow disturbance noise graph; training the convolutional neural network according to the second network loss function value.
According to another aspect of the present disclosure, there is provided a high temperature deformation measuring method, the method including: carrying out residual error identification processing on the detection image through a convolutional neural network to obtain residual error information of the detection image, wherein the residual error information is used for representing the influence of thermal radiation and thermal current disturbance on the test piece; obtaining a deformation image of the test piece according to the detection image and the residual error information; and obtaining a deformation field of the test piece according to the deformation image.
According to the high-temperature deformation measuring device disclosed by the embodiment of the disclosure, the convolutional neural network can be trained respectively aiming at the influence generated by thermal radiation and thermal current disturbance, the influence generated by the coupling effect of uneven brightness caused by thermal radiation and air refractive index change caused by thermal current disturbance can be inhibited, the accuracy of residual error information identification is improved, and a clearer experimental image is obtained. The influence of thermal radiation and thermal current disturbance on the test piece is obtained through the convolutional neural network, the coupling effect of factors such as uneven image brightness caused by high-temperature thermal radiation, air refractive index change caused by thermal current disturbance and the like is reduced, the precision of a deformation field is improved, and the evaluation accuracy of the mechanical property of the material under the high-temperature environment is improved.
Other features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments, which proceeds with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments, features, and aspects of the disclosure and, together with the description, serve to explain the principles of the disclosure.
FIG. 1 shows a schematic view of a high temperature deformation measurement device according to an embodiment of the present disclosure;
FIG. 2 shows a schematic diagram of a convolutional neural network training process, in accordance with an embodiment of the present disclosure;
FIG. 3 shows a schematic diagram of a convolutional neural network training process, in accordance with an embodiment of the present disclosure;
FIG. 4 shows a schematic view of a high temperature deformation measurement device according to an embodiment of the present disclosure;
fig. 5 shows a flow chart of a high temperature deformation measurement method according to an embodiment of the present disclosure.
Detailed Description
Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. In the drawings, like reference numbers can indicate functionally identical or similar elements. While the various aspects of the embodiments are presented in drawings, the drawings are not necessarily drawn to scale unless specifically indicated.
The word "exemplary" is used exclusively herein to mean "serving as an example, embodiment, or illustration. Any embodiment described herein as "exemplary" is not necessarily to be construed as preferred or advantageous over other embodiments.
Furthermore, in the following detailed description, numerous specific details are set forth in order to provide a better understanding of the present disclosure. It will be understood by those skilled in the art that the present disclosure may be practiced without some of these specific details. In some instances, methods, means, elements and circuits that are well known to those skilled in the art have not been described in detail so as not to obscure the present disclosure.
Fig. 1 shows a schematic view of a high temperature deformation measurement apparatus according to an embodiment of the present disclosure, as shown in fig. 1, the apparatus including: a processing device 11, an image acquisition device 12, a heating device 13.
The heating device 13 is configured to: generating hot air flow, and heating the test piece through the hot air flow;
the image acquisition device 12 is configured to: acquiring a detection image of the test piece in the heating process of the test piece;
the processing device 11 is configured to:
carrying out residual error identification processing on the detection image through a convolutional neural network to obtain residual error information of the detection image, wherein the residual error information is used for representing the influence of thermal radiation and thermal current disturbance on the test piece;
obtaining a deformation image of the test piece according to the detection image and the residual error information;
and obtaining a deformation field of the test piece according to the deformation image.
According to the high-temperature deformation measuring device disclosed by the embodiment of the disclosure, the influence of thermal radiation and thermal current disturbance on the test piece can be obtained through the convolutional neural network, the influence of coupling effect of factors such as uneven image brightness caused by high-temperature thermal radiation and air refractive index change caused by thermal current disturbance is reduced, a clearer experimental image is obtained, the measurement precision of a deformation field is improved, and the evaluation accuracy of mechanical property of the material under the high-temperature environment is improved.
In one possible implementation, the test piece may be a carbon/silicon carbide composite material, and may be used as a thermal protection material in the field of aerospace, and the disclosure does not limit the material of the test piece.
In one possible implementation, the test piece may be fixed in front of a lens of an image capturing device, which may be a CCD (charge coupled device) camera, through a test piece holder, and may be used to capture an image of the test piece in a high temperature environment. A filter (for example, a blue filter) may be added in front of the lens of the image capturing device to filter most of the strong light radiation.
In a possible implementation mode, a compensation light source, for example, a blue light source, can be arranged near the camera to compensate the ambient light of the experimental environment, and can be used in cooperation with the blue light filter to obtain a clear image on the basis of avoiding overexposure caused by strong light radiation.
In one possible implementation, the image acquisition device may be connected to the processing device to transmit the acquired image to the processing device, and the processing device may process the acquired image of the test piece to obtain the deformation field of the test piece.
In a possible implementation manner, the heating device may be a device that generates and injects a high-temperature air flow, and when the test piece is heated, the test piece may be blown by a hot air flow, so as to generate an influence of hot air flow disturbance and a thermal radiation influence on the test piece, and a working condition of an aircraft in the aerospace field in flight may be simulated, that is, a working condition of the aircraft that simultaneously bears the high temperature and the hot air flow. The heating device is arranged between the test piece and the image acquisition device, and an included angle between the wind direction of hot air flow generated by the heating device and a normal plane of the test piece is a preset angle, wherein the preset angle is larger than 0 degree and smaller than 90 degrees.
In an example, as shown in FIG. 1, a high temperature air flow generated by a heating device may side blow the test piece, the side blown hot air flow may heat the test piece, generating thermal radiation, and the hot air flow perturbation may cause an air refractive index change, generating a hot air flow perturbation. That is, the heating device blowing the test piece from the side may simultaneously generate the influence of thermal radiation and thermal current disturbance to the test piece.
In a possible implementation manner, during the heating process, the image acquisition device can continuously acquire the detection image of the test piece during the heating process, and the processing device can acquire the residual error information of the detection image through the convolutional neural network, namely, the influence of thermal radiation and thermal current disturbance on the test piece is acquired.
In one possible implementation, the convolutional neural network can identify the noise generated by the external environment on the detection image by using the influence generated by thermal radiation and thermal airflow disturbance as the noise. Because the noise of thermal radiation and thermal airflow disturbance is complex, the convolutional neural network can be a deep learning neural network and can comprise a plurality of network levels, however, the deep neural network can cause the calculation amount to rise sharply and can not obtain the features of a plurality of scales, so that the receptive field can be enlarged and the scales of the features can be increased through the hole convolution without increasing the calculation amount.
In one possible implementation, the convolutional neural network may be a 32-layer neural network, in which the convolutional layer and the ReLU (activation layer) are passed: as a first layer of the convolutional neural network, the input image is processed using a standard convolution and ReLU (modified Linear Unit) function. The convolutional layer consists of 64 filters of size 3 × 3 × 1, which can generate 64 feature images, and the ReLU function acts to non-linearize the output. The feature map can be obtained by using a combination of standard convolution and hole convolution, a batch normalization process and a ReLU function as 2-31 layers of a convolutional neural network by using a scaled Conv (hole convolution), a BN (batch normalization) and a ReLU (activation layer) as layers of the convolutional neural network, wherein the convolutional layer comprises 64 filters with the size of 3 × 3 × 64, and the 30 intermediate layers of the convolutional neural network are formed by cycling ten times in the order of "hole convolution with the standard convolution-hole rate of 1-hole convolution with the hole rate of 2". The final layer of the residual network can be defined by Conv (convolutional layer), which uses standard convolution, reconstructs the image with 13 × 3 × 64 filter, and obtains the residual information.
In a possible implementation manner, before the convolutional neural network is used for acquiring the residual error information, the convolutional neural network can be trained, a coupling effect may be generated between the influences generated by thermal radiation and thermal airflow disturbance, the convolutional neural network is not favorable for identifying the residual error information, and the deformed residual error caused by the expansion of a test piece in the heating process is not favorable for identifying the two influences. Based on the reasons, the two influences are exerted on the test piece and the convolutional neural network is trained to be not beneficial to convergence of convolutional neural network training, so that the convolutional neural network can be trained respectively aiming at the influences caused by thermal radiation influence and thermal current disturbance, the convolutional neural network can obtain the capacity of identifying the two influences respectively, and then residual error information of the detected image can be identified.
In one possible implementation, the convolutional neural network may be trained on the effects of thermal radiation, such that the convolutional neural network obtains the ability to identify the effects of thermal radiation.
Fig. 2 is a schematic diagram illustrating a convolutional neural network training process according to an embodiment of the present disclosure, and as shown in fig. 2, the high temperature deformation measuring apparatus further includes a temperature measuring device for acquiring temperature information of the test piece, and the image acquiring device 12 is further configured to: acquiring a plurality of first sample images during heating of the test piece, the processing device 11 being further configured to: training the convolutional neural network through a plurality of first sample images and temperature information acquired simultaneously with the first sample images.
In a possible implementation manner, when the first sample image is acquired, the test piece is arranged between the heating device and the image acquisition device, and an included angle between a wind direction of hot air flow generated by the heating device and a normal plane of the test piece is 90 °.
In a possible implementation manner, since the heating device and the image acquiring device are respectively located at two sides of the test piece, the influence of the hot air flow on the image acquired by the image acquiring device is small, in this case, the image acquiring device can acquire the first sample image in the heating process of the test piece, and train the capability of the neural network to identify the influence generated by the thermal radiation through the first sample image.
In a possible implementation manner, the temperature measuring device may be an infrared temperature measuring device, and may be configured to detect the temperature of a specific position on the test piece during the heating process of the test piece, and transmit the unread information obtained by the detection to the processing device. The processing equipment can acquire the first sample image in real time and the temperature information of the test piece when shooting the first sample image, and train by using the first sample image and the temperature information.
In one possible implementation, the training the convolutional neural network by using a plurality of first sample images and temperature information acquired simultaneously with the first sample images includes: acquiring image information of a second preset channel of the first sample image; obtaining a temperature field of the surface of the test piece according to the image information of the second preset channel and the temperature information obtained simultaneously with the first sample image; obtaining a thermal radiation map corresponding to the first sample image according to the temperature field; inputting the first sample image into the convolutional neural network to obtain a thermal radiation training image; determining a first network loss function value of the convolutional neural network according to the thermal radiation training diagram and the thermal radiation diagram; training the convolutional neural network according to the first network loss function value.
In one possible implementation, as shown in fig. 2, the test piece may be heated by a heating device, so that the temperature of the test piece rises, generating thermal radiation. In an example, when the flame center temperature of the controllable heating device reaches 600 ℃, the image acquisition device is started to acquire a first sample image, and the temperature information is acquired through the temperature measurement device. The temperature of hot air flow of the heating device can be gradually raised, and a plurality of first sample images and temperature information can be obtained in the heating process. And stopping acquiring the first sample image and the temperature information until the central temperature of the flame of the heating device reaches 1500 ℃. An image set consisting of the first sample image and a temperature data set consisting of the temperature information acquired during the heating process may be used to train the convolutional neural network.
In one possible implementation manner, the thermal radiation diagram of the first sample image can be obtained through the image information and the temperature information of the second preset channel of the first sample image, and the thermal radiation diagram can be used as a training target of the convolutional neural network.
In one possible implementation manner, the second preset channel may be an R (red) channel and a G (green) channel of the first sample image, and image information of the R channel and the G channel of the first sample image (for example, an R value and a G value of each pixel point of the first sample image) may be extracted.
In a possible implementation manner, the temperature field of the surface of the test piece can be obtained through the image information of the second preset channel and the temperature information obtained simultaneously with the first sample image.
In an example, the colorimetric wavelength coefficients may be obtained first through image information of a second preset channel. In an example, the colorimetric wavelength coefficients may be obtained by the following formula (1):
wherein I is the colorimetric wavelength coefficient, C2Is the second Planck radiation constant, λGAt green wavelength, λRAt a red wavelength.
Subsequently, the calculated point temperature (i.e., the temperature of any pixel point) can be obtained by the colorimetric wavelength coefficient and the temperature information, and in an example, the calculated point temperature can be obtained by the following formula (2):
wherein T is the calculated point temperature, T0Is the temperature of the reference point (i.e. the reference position on the test piece measured by the thermometric apparatus), BRGTo calculate the colorimetric values of the red and green channels, BRG0The colorimetric values of the red channel and the green channel are taken as reference points. The calculated temperature of each pixel point, namely, the temperature field of the test piece, can be determined by the temperature of the single point measured by the temperature measuring equipment and the image information of each pixel point in the first sample image according to the formula (2).
Further, the calculated temperature field data formula can be integrated within the filtering wavelength range (for example, 465 ± 5nm) of the optical filter to obtain the radiation intensity of each pixel point, as shown in the following formula (3):
wherein T is absolute temperature, k and h are Planck constants, c is light speed in vacuum, and k is Boltzmann constant.
In one possible implementation, the thermal radiation noise at each point can be determined by the following equation (4), i.e., determining the thermal radiation map:
wherein G israAnd (T) is the thermal radiation noise of the calculation point, m is the conversion coefficient between the image gray value and the response current, u is the photoelectric conversion coefficient, T is the exposure time in the image acquisition process, r ' is 2a/f ' is the relative aperture, a is the radius of the entrance pupil, f ' is the focal length of the image acquisition equipment, and K is the transmittance coefficient of the camera lens group. The thermal radiation noise of each pixel point can be calculated through a formula (4), and then the thermal radiation diagram is obtained. The thermal radiation map is a result obtained by calculating image information of the first sample image and actually measured temperature information, is an accurate thermal radiation map, and can be obtained by subtracting a thermal radiation map from a noise-free map (for example, an image of a test piece obtained at normal temperature), and the accurate noise map is used as a training target of the convolutional neural network.
In a possible implementation manner, the first sample image may be input into a convolutional neural network, a thermal radiation training diagram, that is, a thermal radiation diagram obtained by prediction of the convolutional neural network, may be obtained, and the thermal radiation training diagram may have an error.
In one possible implementation, the first network loss function value of the convolutional neural network may be determined from the thermal radiation training graph and the thermal radiation graph, for example, the first network loss function value may be determined by the following equation (5):
wherein L (Θ) is a first network loss function value, R (y)i(ii) a Theta) is the ith pixel point of the thermal radiation prediction chart, yiIs the ith (i is less than or equal to N, N is the pixel point number of the thermal radiation prediction graphAmount, i and N are positive integers) pixel points, xiThe ith pixel point of a noiseless image (for example, an image photographed at normal temperature) is Θ, which is a learning parameter. That is, the pixel value in the thermal radiation diagram of the ith pixel point is subtracted from the pixel value in the noiseless diagram, so that the accurate noise (thermal radiation noise) of the ith pixel point can be obtained, then, the two-norm between the noise of the ith pixel point of the thermal radiation training diagram and the accurate noise can be obtained, the two-norm of the N pixel points can be averaged, and the first network loss function value can be obtained.
In one possible implementation, the convolutional neural network may be trained with a first network loss function value. For example, the first network loss function value can be used for back propagation, and the network parameters of the convolutional neural network are gradually adjusted by a gradient descent method, so that the error of the convolutional neural network is reduced, and the precision of the convolutional neural network is improved. In an example, the training process may be iteratively performed for a plurality of times until a training condition is satisfied, for example, the training condition includes that the training time reaches a preset time, the first network loss function value is smaller than a preset threshold or converges to a preset interval. The present disclosure does not limit the training conditions.
In one possible implementation, the convolutional neural network may identify the effects of thermal radiation after the training conditions are met. The convolutional neural network trained in the above way can inhibit errors caused by uneven brightness due to high-temperature radiation, improve the identification capability of thermal radiation, and improve the detection precision.
In one possible implementation, after the convolutional neural network is able to identify the effects of thermal radiation, the convolutional neural network may continue to be trained so that it can identify the effects of thermal current disturbances.
Fig. 3 shows a schematic diagram of a convolutional neural network training process according to an embodiment of the present disclosure, and as shown in fig. 3, the image acquisition device 12 is further configured to: acquiring a room temperature image when the test piece is not heated and acquiring a plurality of second sample images during heating of the test piece, the processing device 11 being further configured to: and training the convolution neural network through the normal-temperature image and the plurality of second sample images.
In a possible implementation manner, when the second sample image is acquired, the heating device is disposed between the test piece and the image acquisition device, and an included angle between a wind direction of hot air flow generated by the heating device and a normal plane of the test piece is 0 °.
In a possible implementation, since the hot air flow generated by the heating device does not directly blow the test piece, the temperature change of the test piece is hardly generated, and the generated influence is the influence caused by the air refractive index change caused by the hot air flow disturbance, namely, the residual error generated by the hot air flow disturbance. Because the influence of thermal radiation hardly exists, the decoupling of the influence generated by thermal current disturbance and the influence generated by thermal radiation can be realized in the training process, the capability of the convolutional neural network for identifying the two influences can be respectively improved in the training process, and the training can be converged.
In one possible implementation manner, training the convolutional neural network through the normal-temperature image and the plurality of second sample images includes: performing difference processing on the second sample image and the normal temperature image to obtain a hot air flow disturbance noise map of the test piece; inputting the second sample image into the convolutional neural network to obtain a hot air flow training image; determining a second network loss function value of the convolutional neural network according to the hot airflow training graph and the hot airflow disturbance noise graph; training the convolutional neural network according to the second network loss function value.
In one possible implementation, the ambient temperature image, i.e., the image without noise (thermal radiation and thermal current disturbance) interference, may be acquired by the image acquisition device before the heating device heats up. Subsequently, the heating device can be turned on to generate a hot gas flow. In an example, the acquisition of the second sample image using the image acquisition device is initiated when the flame core temperature of the controllable heating device reaches 600 ℃. The temperature of the hot air flow of the heating device can be gradually raised, and a plurality of second sample images are obtained in the heating process. And stopping acquiring the second sample image until the central temperature of the flame of the heating device reaches 1500 ℃. A set of images made up of the second sample images acquired during the heating process can be used to train the convolutional neural network.
In a possible implementation manner, the second sample image may be subtracted from the normal temperature image, and since only noise interference caused by thermal airflow disturbance exists in the second sample image, a thermal airflow disturbance noise map, that is, a noise map representing thermal airflow disturbance, may be obtained by subtracting the second sample image from the normal temperature image.
In a possible implementation manner, the second sample image may be input into a convolutional neural network to obtain a hot airflow training map, that is, a hot airflow interference noise map obtained by prediction of the convolutional neural network, where the hot airflow training map may have errors, and the accuracy of the neural network may be improved through training to reduce the errors between the hot airflow training map and the hot airflow interference noise map.
In one possible implementation, the second network loss function value may be determined based on an error between the hot gas flow training map and the hot gas flow disturbance noise map, and the convolutional neural network may be trained. In an example, the second network loss function value may be determined in a similar manner to the first network loss function value. That is, the second norm of the noise of the ith pixel point of the hot airflow training diagram and the accurate noise (the difference between the second sample image and the normal temperature image is used for obtaining the noise) can be obtained, and the second norm of the N pixel points can be averaged to obtain the second network loss function value.
In one possible implementation, the convolutional neural network may be trained with the second network loss function values. For example, the second network loss function value can be used for back propagation, and the network parameters of the convolutional neural network are gradually adjusted by a gradient descent method, so that the error of the convolutional neural network is reduced, and the precision of the convolutional neural network is improved. In an example, the training process may be iteratively performed for a plurality of times until a training condition is satisfied, for example, the training condition includes that the training time reaches a preset time, the second network loss function value is smaller than a preset threshold or converges to a preset interval. The present disclosure does not limit the training conditions.
In one possible implementation, the convolutional neural network may identify the effects of thermal airflow disturbances after the training conditions are met. The convolution neural network trained in the above way can inhibit errors caused by refractive index changes caused by hot air flow, improve the recognition capability of interference generated by the hot air flow, and improve the detection precision.
In one possible implementation, after the above two types of training are completed, the convolutional neural network can obtain the capability of identifying the influence caused by thermal radiation and thermal airflow disturbance, that is, the capability of obtaining residual information of the detection image. The sequence of the training is not limited by the present disclosure, that is, the ability of the convolutional neural network to identify thermal current disturbance can be trained first, and then the ability of the convolutional neural network to identify thermal radiation interference can be trained.
In one possible implementation, after training, the convolutional neural network may be tested, for example, a heating device may be disposed at the position in fig. 1, the test piece may be heated to obtain a test image, and the training effect of the neural network may be determined according to the recognition capability of the convolutional neural network on the residual error of the test image. If the training effect is good, namely the recognition precision of the residual error is high, the convolutional neural network can be used in actual measurement, and if the recognition precision of the residual error is low, the neural network can be continuously trained until the training effect meets the measurement requirement.
In one possible implementation, as shown in fig. 1, when residual information of a detection image is acquired through a trained convolutional neural network, an image acquisition device may be a high-resolution camera, and a processing device may have insufficient processing capability on a high-definition image acquired by the image acquisition device. Therefore, the high-definition image can be segmented to obtain a plurality of sub-images, and the processing device can process the sub-images to meet the processing requirement.
In a possible implementation manner, performing residual identification processing on the detected image through a convolutional neural network to obtain residual information of the detected image, including: carrying out segmentation processing on the detection image to obtain a plurality of sub-images of the detection image; and respectively carrying out residual error identification processing on the plurality of sub-images through a convolutional neural network to obtain residual error images of the plurality of sub-images, wherein the residual error information of the detection image comprises the residual error images of the plurality of sub-images.
In one possible implementation, after the image acquisition device acquires the detection image, the processing device may divide the image, for example, into a plurality of (e.g., 4, 8, etc.) sub-images, or may determine the size of the sub-image according to the processing capability (e.g., the image resolution capable of being processed) of the processing device, so as to divide the detection image.
In one possible implementation, the trained convolutional neural network may process a plurality of sub-images separately, and a residual image of each sub-image may be obtained. The residual information of the detection image may include residual images of the plurality of sub-images, i.e., the residual images of the plurality of sub-images may constitute the residual information of the detection image. In residual information, influence generated by thermal radiation and thermal current disturbance can be coupled, the trained convolutional neural network can inhibit influence generated by uneven brightness caused by thermal radiation and air refractive index change caused by thermal current disturbance, coupling effect is relieved, and accuracy of the residual information is improved.
In a possible implementation manner, a deformation image of the test piece can be obtained according to the detection image and the residual error information, the detection image and the residual error information can be subjected to difference processing, and a difference is made between the detection image and the residual error information, namely, a difference is made between the image of the test piece subjected to noise interference after high-temperature deformation and noise interference of the test piece, so that the influence of the noise interference can be removed, and the deformation image can be obtained.
In one possible implementation, if the detection image is subjected to the segmentation processing, the step includes: performing difference processing on the plurality of sub-images and residual images of the plurality of sub-images respectively to obtain deformed sub-images; and splicing the deformation sub-images to obtain the deformation image.
In a possible implementation manner, the difference processing may be performed on the residual images of the plurality of sub-images and the plurality of sub-images, so as to obtain the deformation of the test piece in each sub-image, that is, the deformation sub-image. And splicing the deformation subimages to obtain a deformation image corresponding to the detection image, namely, an image representing the deformation of the test piece.
In one possible implementation, the deformation field of the test piece, i.e., the deformation of each pixel point, may be determined according to the deformed image of the test piece. This step may include: obtaining image information of a first preset channel of the deformed image; and obtaining the deformation field according to the image information of the first preset channel.
In a possible implementation manner, the first preset channel may be a B (blue) channel of a deformed image (e.g., a B value of each pixel), and a deformation field of each pixel may be determined according to the image information. The deformation field can represent the deformation degree of the test piece at high temperature, and can be used for evaluating the mechanical property of the material at high temperature.
According to the high-temperature deformation measuring device disclosed by the embodiment of the disclosure, the convolutional neural network can be trained respectively aiming at the influence generated by thermal radiation and thermal current disturbance, the influence generated by uneven brightness caused by thermal radiation and air refractive index change caused by thermal current disturbance can be inhibited, the coupling effect is reduced, and the accuracy of residual error information is improved. The influence of thermal radiation and thermal current disturbance on the test piece is obtained through the convolutional neural network, the coupling effect of factors such as uneven image brightness caused by high-temperature thermal radiation, air refractive index change caused by thermal current disturbance and the like is reduced, the measurement precision of a deformation field is improved, a clearer experimental image is obtained, and the evaluation accuracy of the mechanical property of the material under the high-temperature environment is improved.
Fig. 4 shows a schematic diagram of a high temperature deformation measuring apparatus according to an embodiment of the present disclosure, and as shown in fig. 4, the apparatus includes a processing device 11, an image acquisition device 12, and a heating device 13. The heating device 13 can be used to generate a hot air flow, and the test piece is heated by the hot air flow. The image acquisition device may be a CCD camera and may be used to acquire images of the test piece in a high temperature environment. The processing device 11 may be an industrial computer operable to process the acquired images to determine a deformation field of the test piece.
In one possible implementation, the test piece may be a carbon/silicon carbide composite material, the test piece may be fixed in front of the lens of the image capturing device by a test piece clamp, and a filter (for example, a blue filter) may be additionally installed in front of the lens of the image capturing device to filter most of the strong light radiation. A compensation light source, such as a blue light source, can also be arranged near the camera to compensate the ambient light of the experimental environment, so as to obtain a clearer image. In addition, temperature measuring devices (e.g., infrared temperature measuring devices) may also be utilized
In one possible implementation, the processing device 11 may determine that the test piece is affected by thermal radiation and thermal current disturbances through a convolutional neural network, which may be trained prior to use. However, since there may be a coupling effect between the influences caused by thermal radiation and thermal airflow disturbance, which is not favorable for the convolutional neural network to identify residual error information, and there is a deformation residual error caused by the expansion of the test piece during the heating process, which is not favorable for identifying the above two influences, the convolutional neural network is trained for two influences at the same time, which may cause difficulty in training convergence, and therefore, the convolutional neural network can be trained for two influences respectively.
In one possible implementation, the convolutional neural network may be trained on the influence of thermal radiation, and the heating device 13 may be set in position 1, i.e. the test piece is set between the heating device and the image acquisition device, and the angle between the wind direction of the hot gas flow generated by the heating device and the normal plane of the test piece is 90 °. Because the heating equipment and the image acquisition equipment are respectively positioned at two sides of the test piece, the influence of hot air flow disturbance on the image acquired by the image acquisition equipment is small, and the noise in the image is mainly the influence generated by heat radiation. In this case, the image acquisition device may acquire the first sample image during the heating process of the test piece, and measure the temperature information of the test piece through the temperature measurement device, i.e., may perform training using the first sample image and the temperature information.
In a possible implementation manner, the heating device can heat the test piece, so that the temperature of the test piece rises, and thermal radiation is generated, for example, when the central temperature of the flame of the controllable heating device reaches 600 ℃, the image acquisition device starts to acquire the first sample image, the temperature measurement device acquires temperature information, and when the central temperature of the flame of the heating device reaches 1500 ℃, the acquisition of the first sample image and the temperature information is stopped. The convolutional neural network may be trained using the plurality of first sample images and temperature information acquired in this process.
In a possible implementation manner, image information of an R channel and an G channel of the first sample image may be extracted, a colorimetric wavelength coefficient may be obtained by using formula (1), and further, a temperature of each pixel point, that is, a temperature field may be calculated by using formula (2). Subsequently, the radiation intensity of each pixel point can be determined by using the formula (3), and then a thermal radiation pattern can be obtained by using the formula (4). In the training process, the thermal radiation diagram and the noiseless diagram (for example, the image of the test piece obtained at normal temperature) can be subjected to subtraction to obtain an accurate noise diagram, and the accurate noise diagram is used as a training target of the convolutional neural network.
In a possible implementation manner, the first sample image may be input to the convolutional neural network, a thermal radiation training graph may be obtained, a second norm between the noise of the ith pixel point of the thermal radiation training graph and the accurate noise (i.e., the noise of the ith pixel point in the accurate noise graph) may be obtained, and the second norms of the N pixel points may be averaged to obtain the first network loss function value. The first network loss function value can be used for back propagation to train the convolutional neural network, and after the training process is iterated for multiple times, the convolutional neural network can obtain the capability of identifying the influence generated by thermal radiation.
In one possible implementation, the convolutional neural network may be trained for the effects of thermal flow disturbances. The heating device 13 may be arranged in position 2, i.e. the heating device is arranged between the test piece and the image acquisition device, the wind direction of the hot gas flow generated by the heating device making an angle of 0 ° with the normal plane of the test piece. In this case, the hot air flow generated by the heating device does not directly blow the test piece, and therefore, the temperature change of the test piece is hardly caused, which is an influence caused by the refractive index change of the air due to the disturbance of the hot air flow.
In a possible implementation manner, when the central temperature of the flame of the heating device reaches 600 ℃, the image acquisition device starts to acquire the second sample image, and when the central temperature of the flame of the heating device reaches 1500 ℃, the acquisition of the second sample image is stopped. The plurality of second sample images obtained in this process may be used to train a convolutional neural network.
In a possible implementation manner, the second sample image and the normal temperature image may be subtracted, and since only noise interference generated by thermal airflow disturbance exists in the second sample image, a thermal airflow disturbance noise map (accurate noise map) may be obtained by subtracting the second sample image and the normal temperature image, and the noise map may be used as a training target of the convolutional neural network.
In a possible implementation manner, the second sample image may be input to the convolutional neural network to obtain a hot airflow training diagram, and two norms of the noise of the ith pixel point of the hot airflow training diagram and the accurate noise (the difference between the second sample image and the normal temperature image is used to obtain the noise) may be obtained, and the two norms of the N pixel points may be averaged to obtain the second network loss function value. The second network loss function value can be used for back propagation to train the convolutional neural network, and after the training process is iterated for multiple times, the convolutional neural network can obtain the capability of identifying the influence generated by hot air flow disturbance.
In one possible implementation, the convolutional neural network is enabled to obtain the ability to identify the effects of thermal radiation and thermal current disturbances through the training process described above. The heating device 13 may be arranged in position 3, i.e. the heating device is arranged between the test piece and the image acquisition device, the angle between the wind direction of the hot gas flow generated by the heating device and the normal plane of the test piece being a preset angle, wherein the preset angle is greater than 0 ° and less than 90 °. The heating equipment can generate hot air flow to blow the test piece, the influence of hot air flow disturbance and the influence of heat radiation are generated on the test piece at the same time, in the heating process, the image acquisition equipment can continuously acquire the detection image of the test piece in the heating process, and the processing equipment can acquire residual error information of the detection image through the convolutional neural network, namely, the influence of the heat radiation and the hot air flow disturbance on the test piece is acquired.
In a possible implementation manner, after the image obtaining device obtains the detection image, the processing device may segment the image, for example, the image may be segmented into a plurality of sub-images, and the convolutional neural network processes the plurality of sub-images respectively, so as to obtain a residual image of each sub-image (in the example, in the training process, the processing device may also segment the first sample image or the second sample image, and perform training by using the segmented sub-images respectively). And performing difference processing on the sub-images and the residual images of the sub-images to obtain deformed sub-images, and performing splicing processing on the deformed sub-images to obtain deformed images corresponding to the detection images. Further, the deformation field of the test piece, that is, the deformation of each pixel point, can be determined according to the deformed image of the test piece. The image information of the B channel of the deformed image can be extracted, and the deformation field of each pixel point is determined by utilizing the image information. The deformation field can be used for representing the deformation degree of the test piece at high temperature and evaluating the mechanical property of the material at high temperature.
In a possible implementation manner, the high-temperature deformation measuring device can obtain a convolution neural network which can be used for identifying influences caused by thermal radiation and thermal airflow disturbance, so as to suppress errors caused by the influences and improve the measurement accuracy of a deformation field. The method can provide an effective measurement means for thermal assessment of high-temperature materials, and can be used for material performance assessment in the fields of aerospace and the like. The present disclosure does not limit the application field of the high-temperature deformation measuring device.
In one possible implementation, the present disclosure also provides a high temperature deformation measurement method that may be performed by a processing device to measure a deformation field of a test piece.
Fig. 5 shows a flow chart of a high temperature deformation measurement method according to an embodiment of the present disclosure, which may include, as shown in fig. 5:
step S11, residual error identification processing is carried out on the detection image through a convolutional neural network, and residual error information of the detection image is obtained, wherein the residual error information is used for representing the influence of thermal radiation and hot air flow disturbance on the test piece;
step S12, obtaining a deformation image of the test piece according to the detection image and the residual error information;
and step S13, obtaining a deformation field of the test piece according to the deformation image.
The flowchart and block diagrams in the figures illustrate the systems, methods and architectures, functions and operations according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, a segment, or a portion of functionality. In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
Having described embodiments of the present disclosure, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (10)
1. A high temperature deformation measuring device, characterized in that the device comprises: a processing device, an image acquisition device, a heating device,
the heating apparatus is for: generating hot air flow, and heating the test piece through the hot air flow;
the image acquisition device is configured to: acquiring a detection image of the test piece in the heating process of the test piece;
the processing device is configured to:
carrying out residual error identification processing on the detection image through a convolutional neural network to obtain residual error information of the detection image, wherein the residual error information is used for representing the influence of thermal radiation and thermal current disturbance on the test piece;
obtaining a deformation image of the test piece according to the detection image and the residual error information;
and obtaining a deformation field of the test piece according to the deformation image.
2. The apparatus of claim 1, wherein performing residual identification processing on the detected image through a convolutional neural network to obtain residual information of the detected image comprises:
carrying out segmentation processing on the detection image to obtain a plurality of sub-images of the detection image;
and respectively carrying out residual error identification processing on the plurality of sub-images through a convolutional neural network to obtain residual error images of the plurality of sub-images, wherein the residual error information of the detection image comprises the residual error images of the plurality of sub-images.
3. The apparatus of claim 2, wherein obtaining a deformation image of the test piece from the inspection image and the residual information comprises:
performing difference processing on the plurality of sub-images and residual images of the plurality of sub-images respectively to obtain deformed sub-images;
and splicing the deformation sub-images to obtain the deformation image.
4. The apparatus of claim 1, wherein obtaining a deformation field of the test piece from the deformation image comprises:
obtaining image information of a first preset channel of the deformed image;
and obtaining the deformation field according to the image information of the first preset channel.
5. The apparatus of claim 1, wherein the heating device is disposed between the test piece and the image acquisition device, and an angle between a wind direction of the hot air flow generated by the heating device and a normal plane of the test piece is a preset angle, wherein the preset angle is greater than 0 ° and less than 90 °.
6. The apparatus of claim 1, further comprising a temperature measuring device for acquiring temperature information of the test piece,
the image acquisition device is further configured to: a plurality of first sample images are acquired during heating of the test piece,
the processing device is further configured to: training the convolutional neural network by a plurality of first sample images and temperature information acquired simultaneously with the first sample images,
when the first sample image is obtained, the test piece is arranged between the heating device and the image obtaining device, and an included angle between the wind direction of hot air flow generated by the heating device and a normal plane of the test piece is 90 degrees.
7. The apparatus of claim 6, wherein the training of the convolutional neural network with a plurality of first sample images and temperature information acquired simultaneously with the first sample images comprises:
acquiring image information of a second preset channel of the first sample image;
obtaining a temperature field of the surface of the test piece according to the image information of the second preset channel and the temperature information obtained simultaneously with the first sample image;
obtaining a thermal radiation map corresponding to the first sample image according to the temperature field;
inputting the first sample image into the convolutional neural network to obtain a thermal radiation training image;
determining a first network loss function value of the convolutional neural network according to the thermal radiation training diagram and the thermal radiation diagram;
training the convolutional neural network according to the first network loss function value.
8. The apparatus of claim 1, wherein the image acquisition device is further configured to: acquiring a room temperature image while the test piece is not heated and acquiring a plurality of second sample images during heating of the test piece,
the processing device is further configured to: training the convolutional neural network through the normal-temperature image and the plurality of second sample images,
when the second sample image is obtained, the heating device is arranged between the test piece and the image obtaining device, and an included angle between the wind direction of hot air flow generated by the heating device and a normal plane of the test piece is 0 degree.
9. The apparatus of claim 8, wherein training the convolutional neural network with the normothermic image and the plurality of second sample images comprises:
performing difference processing on the second sample image and the normal temperature image to obtain a hot air flow disturbance noise map of the test piece;
inputting the second sample image into the convolutional neural network to obtain a hot air flow training image;
determining a second network loss function value of the convolutional neural network according to the hot airflow training graph and the hot airflow disturbance noise graph;
training the convolutional neural network according to the second network loss function value.
10. A method of high temperature deformation measurement, the method comprising:
carrying out residual error identification processing on the detection image through a convolutional neural network to obtain residual error information of the detection image, wherein the residual error information is used for representing the influence of thermal radiation and thermal current disturbance on the test piece;
obtaining a deformation image of the test piece according to the detection image and the residual error information;
and obtaining a deformation field of the test piece according to the deformation image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011108313.0A CN112200788B (en) | 2020-10-16 | 2020-10-16 | High-temperature deformation measuring device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011108313.0A CN112200788B (en) | 2020-10-16 | 2020-10-16 | High-temperature deformation measuring device and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112200788A CN112200788A (en) | 2021-01-08 |
| CN112200788B true CN112200788B (en) | 2021-03-26 |
Family
ID=74009845
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011108313.0A Active CN112200788B (en) | 2020-10-16 | 2020-10-16 | High-temperature deformation measuring device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112200788B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112884751B (en) * | 2021-03-09 | 2021-10-01 | 清华大学 | Device and method for synchronously measuring deformation and temperature by using ultraviolet fluorescent mark |
| CN113030174B (en) * | 2021-04-09 | 2022-03-25 | 清华大学 | Temperature deformation measurement method and system based on partition filtering |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919905A (en) * | 2019-01-08 | 2019-06-21 | 浙江大学 | A kind of Infrared Non-destructive Testing method based on deep learning |
| CN110838139A (en) * | 2019-11-04 | 2020-02-25 | 上海联影智能医疗科技有限公司 | Training method of image registration model, image registration method and computer equipment |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7103460B1 (en) * | 1994-05-09 | 2006-09-05 | Automotive Technologies International, Inc. | System and method for vehicle diagnostics |
| CN104457603B (en) * | 2014-08-18 | 2017-02-22 | 清华大学 | Object deformation measurement method under high-temperature environment |
| RU2694021C1 (en) * | 2015-12-14 | 2019-07-08 | Моушен Метрикс Интернешэнл Корп. | Method and apparatus for identifying portions of fragmented material within an image |
| CN106500612B (en) * | 2016-10-12 | 2019-08-06 | 北京科技大学 | The non-contact measurement system and method for material members high temperature deformation |
| CN108053061B (en) * | 2017-12-08 | 2021-08-31 | 天津大学 | Solar irradiance prediction method based on improved convolutional neural network |
| CN108538011B (en) * | 2018-03-07 | 2020-12-15 | 北京宇航时代科技发展有限公司 | Laser radar fire detection method |
| CN109739181B (en) * | 2019-01-22 | 2021-03-16 | 宁波大学 | Machine tool spindle thermal error modeling detection method based on detection neural network |
| CN111414977B (en) * | 2020-03-09 | 2023-06-20 | 西南交通大学 | Weighted integration temperature sensitive point combination selection method for machine tool spindle thermal error modeling |
| CN111429540B (en) * | 2020-04-22 | 2020-12-08 | 清华大学 | Temperature field and deformation field synchronous measurement device and method |
| CN111611907B (en) * | 2020-05-18 | 2023-10-31 | 沈阳理工大学 | Image-enhanced infrared target detection method |
| CN111721225B (en) * | 2020-06-23 | 2021-04-16 | 清华大学 | Dynamic measurement method and device for temperature deformation in high-temperature environment |
-
2020
- 2020-10-16 CN CN202011108313.0A patent/CN112200788B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109919905A (en) * | 2019-01-08 | 2019-06-21 | 浙江大学 | A kind of Infrared Non-destructive Testing method based on deep learning |
| CN110838139A (en) * | 2019-11-04 | 2020-02-25 | 上海联影智能医疗科技有限公司 | Training method of image registration model, image registration method and computer equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112200788A (en) | 2021-01-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112200788B (en) | High-temperature deformation measuring device and method | |
| CN112067147B (en) | Method and device for synchronously measuring temperature and deformation | |
| CN111721225B (en) | Dynamic measurement method and device for temperature deformation in high-temperature environment | |
| CN111429540B (en) | Temperature field and deformation field synchronous measurement device and method | |
| CN110514392B (en) | Temperature deformation measurement system and method based on wide-angle lens and image distortion correction | |
| EP1678485B2 (en) | Method and ir-camera for determining the risk of condensation | |
| CN111707382B (en) | Dynamic optical compensation method and device for synchronous measurement of temperature deformation | |
| Wang et al. | Improvement method of high-temperature digital image correlation measurement accuracy based on image processing | |
| CN112634373A (en) | Zero-expansion ceramic calibration plate-based dynamic correction method for vision measurement system | |
| CN113865717A (en) | Transient high-temperature colorimetric temperature measuring device based on high-speed camera | |
| CN111289111A (en) | Self-calibration infrared body temperature rapid detection method and detection device | |
| CN113532548A (en) | System and method for synchronously measuring temperature deformation in high-temperature environment | |
| CN114323591A (en) | Optical distortion measuring method and testing system based on digital image correlation method | |
| US11846498B2 (en) | Displacement amount measuring device, displacement amount measuring method, and recording medium | |
| Shaw-Lecerf et al. | Methodology for Validation of Unsteady Pressure-Sensitive Paint Measurements using Pressure Transducers | |
| CN109387488A (en) | A kind of method for fast measuring and instrument of optical glass refractive index | |
| KR102492065B1 (en) | Apparatus and method for optical measuring | |
| CN111750992B (en) | Spectrum estimation method based on self-adaptive weighted linear regression | |
| CN114034405A (en) | Non-contact temperature measurement method and system | |
| Dziarski et al. | Effect of unsharpness on the result of thermovision diagnostics of electronic components | |
| Youngquist et al. | Optical distortion evaluation in large area windows using interferometry | |
| Vaughan et al. | Correlation of a thermal model using infra-red camera data for thermo-elastic predictions | |
| Quix et al. | Model deformation measurement capabilities at ETW | |
| CN112461372B (en) | NETD test evaluation device and method for high-low temperature medium-wave thermal imager | |
| Nekrylov et al. | Choosing parameters of active reference mark optical-electronic systems spatial position control |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |