Multi-spectral temperature measurement method based on facet optimization processing technology, device, storage medium and electronic equipment

Abstract

The present invention discloses a multi-spectral temperature measurement method based on facet optimization processing technology, which includes the following steps: step 1, converting spectral information into spectral energy values; step 2, dividing each curved surface of a turbine blade into a plurality of triangular facets; step 3, adjusting the triangular facets once based on energy gradient to obtain the triangular facets after the primary adjustment; step 4, adjusting the triangular facets twice based on neural network model technology; step 5, calculating the weight of the triangular facets after the secondary adjustment; step 6, using the parameter information of the triangular facets after the secondary adjustment, the spectral energy values ​​and the weight values ​​of the triangular facets, to calculate the radiation temperature of the turbine blade to be measured. The present invention provides a multi-spectral temperature measurement method that enhances the overall measurement accuracy of a complex blade curved surface by refining and adjusting the facets.

Description

Technical Field

[0001] The present invention relates to the technical field of turbine blade temperature measurement, and more specifically, to a multi-spectral temperature measurement method based on facet optimization processing technology, a device thereof, a storage medium and an electronic device. Background Art

[0002] During the operation of aircraft engines, temperature measurement of turbine blades is one of the key technologies to ensure engine performance and safety. Traditional temperature measurement methods have many limitations in environments with high temperature, high pressure and high-speed rotation of turbines. In order to solve these problems, non-contact multi-spectral radiation temperature measurement technology came into being. It calculates the surface temperature of an object by analyzing the spectral radiation information of different wavelengths emitted from the surface of the object. Multi-spectral radiation temperature measurement technology can provide more accurate temperature information than other non-contact temperature measurement methods, such as colorimetric temperature measurement. This technology can not only adapt to complex working environments, but also reduce the interference of environmental factors by analyzing data at multiple wavelengths, thereby improving the accuracy and reliability of measurements.

[0003] However, in practical applications, the complex geometry and material properties of turbine blades require temperature measurements to be not only high-precision but also high-resolution. At this point, the surface element partitioning technology is particularly important. By dividing the surface of the turbine blade into multiple small measurement units or "surface elements", the temperature changes in each area can be captured more accurately, which is important for analyzing the temperature of the entire blade and predicting its life.

[0004] In addition, multi-spectral temperature measurement is often affected by reflected radiation errors during implementation. This error mainly comes from the reflection of thermal radiation from other engine components or adjacent blades on the surface of the turbine blade, especially in high temperature environments. Effectively identifying and compensating for this error is the key to improving the accuracy of multi-spectral temperature measurement.

[0005] In the existing multi-spectral radiation temperature measurement technology for aircraft engine turbine blades, although the use of spectral information in multiple spectral bands improves the overall accuracy of temperature measurement, it still faces several challenges. When dealing with complex curved surfaces and their reflected radiation, traditional temperature measurement technology is often unable to accurately subdivide and process the data of each measurement area, especially in areas with large temperature changes or large influence of reflected radiation. In addition, the existing technology fails to fully utilize the spectral data correlation between various areas on the turbine blade surface, thus limiting the further improvement of temperature measurement accuracy. Summary of the invention

[0006] An object of the present invention is to solve at least the above problems and to provide at least the advantages which will be described hereinafter.

[0007] Another object of the present invention is to provide a multi-spectral temperature measurement method based on facet optimization processing technology, which solves the technical problem that the prior art cannot accurately subdivide and process the data of each measurement area for complex surfaces, and cannot fully utilize the correlation between the spectral data of each area, resulting in limited improvement in measurement accuracy.

[0008] In order to achieve these purposes and other advantages according to the present invention, a multi-spectral temperature measurement method based on facet optimization processing technology is provided, which comprises the following steps:

[0009] Step 1: Obtain spectral information of the turbine blade to be tested, and convert the spectral information into spectral energy value;

[0010] Step 2: Obtain a 3D model of the turbine blade to be tested, and divide each curved surface of the turbine blade into a plurality of triangular face elements;

[0011] Step 3: based on the angle threshold between the face element normal vector and the reference direction vector, extract the parameter information and spectral energy value of the triangular face element of the target surface, calculate the energy gradient of each triangular face element on the target surface, adjust the triangular face element once based on the energy gradient, obtain the triangular face element after the adjustment, and extract the parameter information and spectral energy value of the triangular face element after the adjustment; wherein the parameter information of the triangular face element includes the area, side length and angle of the triangular face element;

[0012] Step 4: Based on the neural network model technology, the triangular surface element after the primary adjustment is adjusted for a second time to obtain parameter information of the triangular surface element after the secondary adjustment, and the spectral energy value of the triangular surface element after the secondary adjustment is extracted;

[0013] Step 5: Calculate the weight W of the triangular face after secondary adjustment based on formula (1) i ;

[0014] Formula (1);

[0015] Where, k in formula (1) is the adjustment coefficient, x i is the data quality score of the ith triangle face, x mean is the average of the data quality scores of all triangular facets;

[0016] Step 6: Using the parameter information, spectral energy value and weight value of the triangular facet obtained after the secondary adjustment, based on the emissivity model and the temperature inversion algorithm, the radiation temperature of the turbine blade to be measured is calculated.

[0017] Preferably, in the multi-spectral temperature measurement method based on the facet optimization processing technology, in step 1, the spectral information is converted into spectral energy value based on formula (2);

[0018] P = 10 × log 10 L formula (2);

[0019] In formula (2), P is the spectral information, the unit is dBm, and L is the spectral energy value, the unit is mW.

[0020] Preferably, in the multi-spectral temperature measurement method based on the facet optimization processing technology, step 2 is specifically as follows:

[0021] S2.1 Take the midpoint of each side length of the curved surface of the 3D model of the turbine blade to be tested, connect adjacent midpoints in sequence, and divide the curved surface into a plurality of primary triangular face elements;

[0022] S2.2, according to step S2.1, divide each primary triangular face into multiple secondary triangular face elements by taking the midpoint of the side length;

[0023] S2.3, repeat step S2.2 to perform multiple face element divisions until the number of divisions is equal to the preset division level value, thereby obtaining multiple triangular face elements.

[0024] Preferably, in the multi-spectral temperature measurement method based on the facet optimization processing technology, in step 3, the energy gradient of each triangular facet is calculated based on formula (3);

[0025] △E=[ max(E)- min(E)] / S Formula (3);

[0026] In formula (3), max(E) and min(E) represent the highest spectral energy value and the lowest spectral energy value inside the triangular surface element, respectively, and S is the area of ​​the triangular surface element;

[0027] Using formula (4) based on the energy gradient of the triangular face element, the area size of the triangular face element is adjusted;

[0028] S i =αexp(-β△E i ) Formula (4);

[0029] In formula (4), S i represents the area of ​​the i-th triangle face after one adjustment, △E i is the energy gradient at the center of the triangular face, α and β are adjustment parameters, the initial value of α is 0.02, and the initial value of β is 0.05.

[0030] Preferably, in the multi-spectral temperature measurement method based on the facet optimization processing technology, step 4 specifically includes:

[0031] Step S4.1, using the spectral energy value and parameter information of the triangular face element before adjustment as input values ​​and the parameter information of the triangular face element after adjustment as output values, constructing a neural network model 1; using the historical sample data of the turbine blades, training, testing and verifying the neural network model 1;

[0032] Step S4.2, inputting the parameter information and spectral energy value of the triangular facet after the primary adjustment in the historical sample data of the turbine blade into the neural network model 1 as input values, and obtaining the parameter information of the triangular facet after the secondary adjustment corresponding to the historical sample data;

[0033] Step S4.3, using the parameter information and spectral energy value of the triangular face element after the primary adjustment as input values, and the parameter information of the triangular face element after the secondary adjustment as output values, constructing a second neural network model, and using the historical sample data of the turbine blades to train, test and verify the second neural network model;

[0034] Step S4.4: The parameter information and spectral energy value of the triangular face element after the primary adjustment of the target surface are input into the trained neural network model 2 as input values ​​to obtain the parameter information of the triangular face element after the secondary adjustment.

[0035] Preferably, in the multi-spectral temperature measurement method based on the facet optimization processing technology, in step 4, the neural network model is a hybrid architecture formed by a convolutional neural network (CNN) combined with a fully connected layer (Dense Layers); wherein the convolution layer includes three layers of convolution, each convolution layer is followed by batch normalization and a ReLU activation function, and each convolution layer is followed by a pooling layer; the fully connected layer includes two fully connected layers with 128 and 64 nodes respectively;

[0036] The batch size of the neural network model is 32, the number of iterations is 50, and the learning rate is initially set to 0.001. K-fold cross validation (K=5) is used to evaluate the generalization ability of the neural network model, and the quality of the output results is evaluated by the average mean square error (MSE) on the test set.

[0037] Preferably, in the multi-spectral temperature measurement method based on the facet optimization processing technology, x in formula (1) i is obtained by calculating formula (5), x mean It is obtained by calculating formula (6);

[0038] x i =SNR i ×(1-R i ) Formula (5);

[0039] Formula (6);

[0040] Among them, SNR in formula (5) i is the signal-to-noise ratio of the i-th bin, which is calculated by formula (7); is the adjustment factor, which is calculated by formula (8); in formula (6), N is the total number of bins;

[0041] Formula (7);

[0042] Formula (8);

[0043] Among them, P in formula (7) signal,i and P noise,i Represent the signal power and noise power of the i-th bin respectively; in formula (8), T i is the temperature of the i-th panel; θ is the evaluation factor, and its initial value is set to 0.1.

[0044] The present invention also provides a multi-spectral temperature measurement device based on facet optimization processing technology, which is used to implement the above method.

[0045] The present invention also provides a storage medium, in which a computer program is stored. When the computer program is loaded and executed by a processor, the above method is implemented.

[0046] The present invention also provides an electronic device, comprising: a processor and a memory; wherein the memory is used to store a computer program; and the processor is used to load and execute the computer program so that the electronic device executes the above method.

[0047] The present invention has at least the following beneficial effects:

[0048] The present invention proposes a multi-spectral temperature measurement method based on facet optimization processing technology. The core technology of the method lies in a new method combining adaptive facet subdivision and spectral data processing. The method develops an adaptive subdivision algorithm that can adjust the facet size and shape according to the spectral information distribution of the blade surface. In areas where the temperature gradient is large or the reflected radiation influence is concentrated, the algorithm divides the facet into finer areas; while in edge areas where the temperature changes gently and the reflected radiation influence is small, the facet is larger to reduce the amount of calculation and reduce the time complexity of the algorithm. In addition, the algorithm also includes using machine learning to optimize the shape of each facet, and using the principle of local similarity or spatial continuity to accurately match the curved surface structure of the turbine blade. At the same time, the method enables each facet to not only process its spectral data independently, but also share spectral information with adjacent facets, and weightedly adjust the priority and precision of each facet according to the spectral data of each band collected from multiple channels, thereby enhancing the overall measurement accuracy of the complex blade surface.

[0049] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Schematic diagram of the process of the multi-spectral temperature measurement method based on the facet optimization processing technology described in Example 1 of the present invention. DETAILED DESCRIPTION

[0051] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments so that those skilled in the art can implement the invention with reference to the description.

[0052] It should be understood that the terms such as “having”, “including” and “comprising” used herein do not exclude the existence or addition of one or more other elements or combinations thereof.

[0053] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified.

[0054] In the description of the present invention, the terms "lateral", "longitudinal", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside" and "outside" etc. to indicate directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings and are only for the convenience of describing the present invention and simplifying the description. They do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as a limitation on the present invention.

[0055] like Figure 1 As shown, the present invention provides a multi-spectral temperature measurement method based on facet optimization processing technology, which includes the following steps:

[0056] Step 1: Obtain spectral information of the turbine blade to be tested, and convert the spectral information into spectral energy value;

[0057] Step 2: Obtain a 3D model of the turbine blade to be tested, and divide each curved surface of the turbine blade into a plurality of triangular face elements;

[0058] Step 3: based on the angle threshold between the face element normal vector and the reference direction vector, extract the parameter information and spectral energy value of the triangular face element of the target surface, calculate the energy gradient of each triangular face element on the target surface, adjust the triangular face element once based on the energy gradient, obtain the triangular face element after the adjustment, and extract the parameter information and spectral energy value of the triangular face element after the adjustment; wherein the parameter information of the triangular face element includes the area, side length and angle of the triangular face element;

[0059] Step 4: Based on the neural network model technology, the triangular surface element after the primary adjustment is adjusted for a second time to obtain parameter information of the triangular surface element after the secondary adjustment, and the spectral energy value of the triangular surface element after the secondary adjustment is extracted;

[0060] Step 5: Calculate the weight W of the triangular face after secondary adjustment based on formula (1) i ;

[0061] Formula (1);

[0062] Where, k in formula (1) is the adjustment coefficient, x i is the data quality score of the ith triangle face, x mean is the average of the data quality scores of all triangular facets;

[0063] Step 6: Using the obtained parameter information of the triangular facet after secondary adjustment, the spectral energy value (obtained in step 4) and the weight value of the triangular facet (obtained in step 5), based on the emissivity model and the temperature inversion algorithm, the radiation temperature of the turbine blade to be measured is calculated.

[0064] In another technical solution, a multi-spectral temperature measurement method based on facet optimization processing technology, in step 1, spectral information is converted into spectral energy value based on formula (2);

[0065] P = 10 × log 10 L formula (2);

[0066] In formula (2), P is the spectral information, the unit is dBm, and L is the spectral energy value, the unit is mW.

[0067] In another technical solution, the multi-spectral temperature measurement method based on the facet optimization processing technology, step 2 is specifically as follows:

[0068] S2.1 Take the midpoint of each side length of the curved surface of the 3D model of the turbine blade to be tested, connect adjacent midpoints in sequence, and divide the curved surface into a plurality of primary triangular face elements;

[0069] S2.2, according to step S2.1, divide each primary triangular face into multiple secondary triangular face elements by taking the midpoint of the side length;

[0070] S2.3, repeat step S2.2 to perform multiple face element divisions until the number of divisions is equal to the preset division level value, thereby obtaining multiple triangular face elements.

[0071] In another technical solution, a multi-spectral temperature measurement method based on facet optimization processing technology, in step 3, the energy gradient of each triangular facet is calculated based on formula (3);

[0072] △E=[ max(E)- min(E)] / S Formula (3);

[0073] In formula (3), max(E) and min(E) represent the highest spectral energy value and the lowest spectral energy value inside the triangular surface element, respectively, and S is the area of ​​the triangular surface element;

[0074] Using formula (4) based on the energy gradient of the triangular face element, the area size of the triangular face element is adjusted;

[0075] S i =αexp(-β△E i ) Formula (4);

[0076] In formula (4), S i represents the area of ​​the i-th triangle face after one adjustment, △E i is the energy gradient at the center of the triangular face, α and β are adjustment parameters, the initial value of α is 0.02, and the initial value of β is 0.05.

[0077] In another technical solution, the multi-spectral temperature measurement method based on the facet optimization processing technology, step 4 specifically includes:

[0078] Step S4.1, using the spectral energy value and parameter information of the triangular face element before adjustment as input values ​​and the parameter information of the triangular face element after adjustment as output values, constructing a neural network model 1; using the historical sample data of the turbine blades, training, testing and verifying the neural network model 1;

[0079] Step S4.2, inputting the parameter information and spectral energy value of the triangular facet after the primary adjustment in the historical sample data of the turbine blade into the neural network model 1 as input values, and obtaining the parameter information of the triangular facet after the secondary adjustment corresponding to the historical sample data;

[0080] Step S4.3, using the parameter information and spectral energy value of the triangular face element after the primary adjustment as input values, and the parameter information of the triangular face element after the secondary adjustment as output values, constructing a second neural network model, and using the historical sample data of the turbine blades to train, test and verify the second neural network model;

[0081] Step S4.4: The parameter information and spectral energy value of the triangular face element after the primary adjustment of the target surface are input into the trained neural network model 2 as input values ​​to obtain the parameter information of the triangular face element after the secondary adjustment.

[0082] In another technical solution, a multi-spectral temperature measurement method based on facet optimization processing technology, in step 4, the neural network model is a hybrid architecture formed by a convolutional neural network (CNN) combined with a fully connected layer (Dense Layers); wherein the convolution layer includes three layers of convolution, each convolution layer is followed by batch normalization and ReLU activation function, and each convolution layer is followed by a pooling layer; the fully connected layer includes two fully connected layers with 128 and 64 nodes respectively;

[0083] The batch size of the neural network model is 32, the number of iterations is 50, and the learning rate is initially set to 0.001. K-fold cross validation (K=5) is used to evaluate the generalization ability of the neural network model, and the quality of the output results is evaluated by the average mean square error (MSE) on the test set.

[0084] In another technical solution, a multi-spectral temperature measurement method based on facet optimization processing technology is used. In formula (1), x i is obtained by calculating formula (5), x mean It is obtained by calculating formula (6);

[0085] x i =SNR i ×(1-R i ) Formula (5);

[0086] Formula (6);

[0087] Among them, SNR in formula (5) i is the signal-to-noise ratio of the i-th bin, which is calculated by formula (7); is the adjustment factor, which is calculated by formula (8); in formula (6), N is the total number of bins;

[0088] Formula (7);

[0089] Formula (8);

[0090] Among them, P in formula (7) signal,i and P noise,i Represent the signal power and noise power of the i-th bin respectively; in formula (8), T i is the temperature of the i-th panel; θ is the evaluation factor, and its initial value is set to 0.1.

[0091] The present invention provides a novel pre-processing method for multi-spectral temperature measurement of aircraft engine turbine blades. The method uses algorithms and data processing technology to perform more refined facet division on the surface of turbine blades, thereby improving the temperature measurement accuracy and efficiency. This scheme mainly includes three aspects: facet subdivision and temperature distribution adjustment, machine learning optimization of facet shape, and multi-channel spectral data weight adjustment. First, a spectrometer is used to collect radiation data with a wavelength of 1100nm-1600nm at 1000K-1200K from the surface of the blade to be measured, and the data is pre-processed, filtered, and denoised before input into this algorithm. The algorithm dynamically adjusts the size and shape of the facet based on the basic theoretical knowledge of multi-spectral radiation temperature measurement and data characteristics, while taking into account the influence of reflected radiation. The adjusted model can output the temperature distribution on the surface of the turbine blade through the existing emissivity model and temperature inversion algorithm.

[0092] The multi-spectral temperature measurement method based on facet optimization processing technology proposed in the present invention divides the target surface of the collected turbine blade to be measured into triangular facets in sequence, adjusts the divided triangular facets based on energy gradient values, and adjusts the triangular facets again using neural network model technology, thereby achieving refined processing of complex surfaces and improving temperature measurement accuracy and efficiency. In addition, the present invention also performs weighted adjustment on the priority of the facets to be measured, so that each facet obtains a different priority, further enhancing the overall measurement accuracy of the complex blade surface.

[0093] The method used in the present invention is: using the angle relationship between the normal vector of the face element and the reference vector to extract the face of the turbine blade to be measured, and according to the temperature distribution, performing face element segmentation on the area with higher temperature and greater reflected radiation influence. Then, using the machine learning method, learning the relationship between the face element shape and the temperature measurement effect in the historical data, adjusting the shape configuration of the face element of the current face to be measured; traversing the face elements, weighting and adjusting the priority of the face elements of the face to be measured according to the data collected from different spectral channels, so that each face element obtains a different priority. The processed face element spectral data is input into the known emissivity model and the temperature measurement algorithm to complete the temperature inversion.

[0094] The present invention also provides a multi-spectral temperature measurement device based on the facet optimization processing technology, which is used to implement all the steps of the multi-spectral temperature measurement method based on the facet optimization processing technology.

[0095] The present invention also provides a storage medium, which includes: ROM, RAM, magnetic disk or optical disk and other media that can store program codes, wherein a computer program is stored. When the computer program is loaded and executed by a processor, all or part of the steps of the multi-spectral temperature measurement method based on the above-mentioned facet optimization processing technology are implemented.

[0096] The present invention also provides an electronic device, which is a device including a processor (CPU / MCU / SOC) and a memory (ROM / RAM), such as a desktop computer, a portable computer, a smart phone, etc. In particular, a computer program is stored in the memory, and when the processor loads and executes the computer program, all or part of the steps of the multi-spectral temperature measurement method based on the facet optimization processing technology are implemented.

[0097] Example 1

[0098] The present invention provides a multi-spectral temperature measurement method based on facet optimization processing technology, which comprises the following steps:

[0099] Step S100, obtaining spectral information of the turbine blade to be tested, and converting the spectral information into spectral energy values; specifically, inputting a 3D model of the turbine blade to be tested into a computer system, using a spectrometer to collect spectral information P (decibel milliwatt value / dBm) of the turbine blade at different temperatures and wavelengths, and converting it into a spectral energy value L (mW) using formula (1);

[0100] P = 10 × log 10 L formula (2);

[0101] In formula (2), P is the spectral information, the unit is dBm, and L is the spectral energy value, the unit is mW;

[0102] Step S200, obtaining a 3D model of the turbine blade to be tested, and dividing each curved surface of the turbine blade into a plurality of triangular facets; specifically:

[0103] S201 takes the midpoint of each side length of the curved surface of the 3D model of the turbine blade to be measured, connects adjacent midpoints in sequence, and divides the curved surface into a plurality of primary triangular face elements;

[0104] S202, dividing each primary triangular face element by taking the midpoint of the side length to form a plurality of secondary triangular face elements according to S201;

[0105] S203, repeatedly executing S202, performing multiple face element divisions, until the number of divisions is equal to the preset division level value, and obtaining multiple triangular face elements.

[0106] During the face division process, data is initialized first, an empty array is set for data storage in face division, and an array 'subdividedVertices' for storing newly generated vertices and an array 'subdividedFaces' for storing new faces are set. At the same time, a map 'midpointIndexMap' is set to track the midpoints of the calculated edges to avoid repeated calculations.

[0107] Then, using the 'subdivideMesh' function, we process each triangle face one by one and calculate the midpoint of each face's three edges. This step involves checking if the midpoint already exists in the 'midpointIndexMap'. If not, we calculate a new midpoint and update the vertex array and map.

[0108] Again, four new triangles are generated based on the three vertices of each original triangle and the three newly calculated midpoints. Each time a subdivision is performed, a triangle becomes four small triangles.

[0109] Finally, the face subdivision process is recursively performed according to the input subdivision level (preset division level value) 'subdivisionLevel'.

[0110] Step S300, based on the angle threshold between the face element normal vector and the reference direction vector, extract the parameter information and spectral energy value of the triangular face element of the target surface, define 'threshold = cos(pi / 3)', use the angle threshold range between each face element normal vector and the reference direction vector to extract the test surface (target surface) of the blade to be tested, further extract the parameter information and spectral energy value of the triangular face element of the target surface, input the moving blade, guide vane and its adjacent blade model, use the angle relationship between the reference vector and the face element normal vector on the blade model surface (the angle between the normal vector and the reference vector is within 60°) to extract the test surface of the blade to be tested as the target surface;

[0111] Calculate the energy gradient of each triangular face element on the target surface, adjust the triangular face element once based on the energy gradient to obtain the adjusted triangular face element, and extract the parameter information and spectral energy value of the adjusted triangular face element; wherein the parameter information of the triangular face element includes the area, side length and angle of the triangular face element;

[0112] After the triangular facet division is completed, the system will measure the energy gradient of all facets in the entire area of ​​the blade to be tested. From the existing conclusions, it can be seen that in the same band, the higher the spectral energy value, the higher the temperature. This is achieved by analyzing the multi-channel spectral energy data collected from the spectrometer. The calculation of the energy gradient is used to evaluate the severity of the temperature change inside each facet; the energy gradient of each triangular facet is calculated based on formula (3);

[0113] △E=[ max(E)- min(E)] / S Formula (3);

[0114] In formula (3), max(E) and min(E) represent the highest spectral energy value and the lowest spectral energy value inside the triangular surface element, respectively, and S is the area of ​​the triangular surface element;

[0115] According to the measured energy gradient, the system will dynamically adjust the size of the facet. In areas with high energy gradient, that is, areas with drastic temperature changes, the temperature measurement accuracy will be improved by reducing the facet size; the facets are subdivided using the temperature and reflected radiation distribution. According to the temperature distribution of the blade, the surface to be measured of the turbine blade is first divided into triangular facets to ensure that each area can be fully covered. Facing the middle part of the blade, usually because it has a higher temperature and significant reflected radiation influence relative to the edge area, the triangular facets in this area are further subdivided to capture more detailed temperature changes;

[0116] Using formula (4) based on the energy gradient of the triangular face element, the area size of the triangular face element is adjusted;

[0117] S i =αexp(-β△E i ) Formula (4);

[0118] In formula (4), S i represents the area of ​​the i-th triangle face after one adjustment, △E i is the energy gradient at the center of the triangular facet, α and β are adjustment parameters, the α value mainly controls the initial ratio of the facet size adjustment, and its value determines the size of the facet source when there is no radiation energy gradient or the radiation energy gradient value is very small, and the initial value of α is set to 0.02; the β value is used to control the sensitivity of the facet size to the energy gradient, and the initial value of β is set to 0.05; each time the values ​​of α and β are adjusted subsequently, the numerical adjustment is performed with a step size of 0.005, and all facets are traversed at the same time to obtain the corresponding area parameters. When the area of ​​the facet among all facets is greater than 0.1% of the overall area of ​​the measured surface, the numerical adjustment of α and β is stopped, and the values ​​at this time are taken as the optimal values;

[0119] Step S400, based on the neural network model technology, the triangular face element after the primary adjustment is adjusted for a secondary time, parameter information of the triangular face element after the secondary adjustment is obtained, and the spectral energy value of the triangular face element after the secondary adjustment is extracted; specifically includes:

[0120] Step S401, using the spectral energy value and parameter information of the triangular facet before adjustment as input values ​​and the parameter information of the triangular facet after adjustment as output values, constructing a neural network model 1; using the historical sample data of the turbine blades, training, testing and verifying the neural network model 1;

[0121] Step S402: inputting the parameter information and spectral energy value of the triangular facet after the primary adjustment in the historical sample data of the turbine blade into the neural network model 1 as input values, and obtaining the parameter information of the triangular facet after the secondary adjustment corresponding to the historical sample data;

[0122] Step S403: using the parameter information and spectral energy value of the triangular facet after the primary adjustment as input values ​​and the parameter information of the triangular facet after the secondary adjustment as output values, constructing a second neural network model, and using the historical sample data of the turbine blades to train, test and verify the second neural network model;

[0123] Step S404: input the parameter information and spectral energy value of the triangular face element after the primary adjustment of the target surface as input values ​​into the trained neural network model 2 to obtain the parameter information of the triangular face element after the secondary adjustment.

[0124] The deep learning network is then used to optimize the shape of the triangular facets after the first adjustment to ensure that each facet better fits the actual geometric shape of the blade. The network automatically adjusts the geometric configuration of the current facet by learning the relationship between the facet shape and the temperature measurement effect in the historical data, thereby reducing the measurement error caused by unreasonable facet shape.

[0125] Initialize the dataset of the machine learning neural network model and define the feature set 'Features', including the spectral data of each facet, where the spectral data is the energy value obtained from the spectrometer and denoised, covering the wavelength range of 1100nm to 1450nm. Define the target variable 'Labels' to store the shape parameters of each facet, such as side length and angle, which will be optimized during training. Select the data of the feature set 1100nm-1400nm at each temperature point as the training set, and the data of 1400nm-1450nm as the test set.

[0126] Set the machine learning neural network model structure. The neural network model adopts a hybrid architecture of convolutional neural network (CNN) combined with fully connected layers (Dense Layers). The input layer receives the normalized triangular face element parameter information and spectral energy value data; the convolution layer contains three layers of convolution, each convolution layer is followed by batch normalization and ReLU activation function. The convolution layer extracts different features of the spectral energy value data by applying multiple filters, such as energy peak, displacement and other features; batch normalization stabilizes network training and speeds up learning, while ReLU introduces nonlinearity. Each convolution layer is followed by a maximum pooling layer to reduce data dimensions and prevent overfitting; after that, two fully connected layers are connected, with the number of nodes being 128 and 64 respectively, and ReLU activation is used; the last layer is a fully connected layer, which is the output layer, used to output the adjusted face element parameter information, such as area, side length, angle (the angle of the three angles of the triangular face element) and other parameters, and multiple nodes are set to correspond to each parameter one by one.

[0127] The formula of the neural network model is: Shape new =f(Shape old ,τ(Shapeold ,Temperature))

[0128] Among them, Shape new and Shape old They represent the shape of the face element before and after adjustment, f is the optimization function, and τ is the loss function, which is used to evaluate the deviation between the face element shape and the actual temperature measurement effect. The entire network loss function τ uses the mean square error (MSE) to calculate the difference between the predicted shape parameters and the actual parameters. Temperature is the actual measured temperature of the face element; the optimizer uses the Adam optimizer that can automatically adjust the learning rate.

[0129] The batch size of the neural network model is set to 32, the number of iterations is set to 50 rounds, the initial learning rate is set to 0.001, and K-fold cross validation (K=5) is used to evaluate the generalization ability of the model. The final shape optimization quality is evaluated by the average mean square error (MSE) on the test set.

[0130] Step S500: Calculate the weight W of the triangle face after secondary adjustment based on formula (1): i ; Specifically include:

[0131] Step S501, perform quality assessment on the collected spectral data, using the signal-to-noise ratio (SNR) as the assessment index, which is calculated based on formula (7);

[0132] Formula (7);

[0133] In formula (7), SNR i is the signal-to-noise ratio of the i-th bin, P signal,i and P noise,i Represent the signal power and noise power of the i-th bin respectively;

[0134] Step 502: Introduce an adjustment factor R i To take into account the influence of reflected radiation, especially in high temperature areas. Adjustment factor R i It reflects the influence of reflected radiation on signal quality and is calculated by formula (8):

[0135] Formula (8);

[0136] In formula (8), T iis the temperature of the i-th face element; θ is the evaluation factor, and its initial value is set to 0.1. The evaluation factor θ is used to measure the sensitivity of temperature to reflected radiation. Later, according to the optical path of the constructed system, the material properties of the turbine blades, etc., the degree of closeness between the temperature measurement output result and the actual data can be observed when θ takes different values; the temperature in the middle area of ​​the blade surface to be tested is relatively high, and the reflected radiation has a great influence, while the temperature in the edge area is relatively low, and the reflected radiation has a small influence; the entire surface to be tested is vertically projected to obtain a rectangular outline, with the midpoint of the long side and the short side as the center, and 20% of the length of the long side and the short side as the interval length. Another sub-rectangle is taken inside the rectangular outline. The face elements of the turbine blade surface to be tested projected in this area are greatly affected by reflected radiation, and the setting range of the evaluation factor is 0.05~0.1; the face elements of other surfaces to be tested outside this area are less affected by reflected radiation, and the setting range of the evaluation factor is 0.1~0.2;

[0137] Step S503: Combine the signal-to-noise ratio and the reflected radiation adjustment factor, and devalue the data in the reflected radiation sensitive area by considering the high signal-to-noise ratio. This makes it possible that in the high temperature area with strong reflected radiation, even if the signal-to-noise ratio is high, the data quality index will be reduced due to the influence of reflected radiation. The data quality index x is calculated using formula (5): i , aggregate the data quality scores of all facets and calculate the average value, and use formula (6) to calculate the average value x of the data quality scores of all triangular facets mean ;

[0138] x i =SNR i ×(1-R i ) Formula (5);

[0139] Formula (6);

[0140] Where, N in formula (6) is the total number of facets;

[0141] Step S540: Calculate the weight of each facet based on the difference between the data quality score of each facet and the average value, and use formula (1) to calculate the weight W i

[0142] Formula (1);

[0143] Where, k in formula (1) is the adjustment coefficient, x i is the data quality score of the ith triangle face, x mean is the average of the data quality scores of all triangular facets;

[0144] Finally, the priority and detail level of each facet are weighted and adjusted according to the data collected from different spectral channels. In this process, the weight distribution depends on the quality of the data in each channel and its contribution to the temperature prediction, ensuring that spectral data with high signal-to-noise ratio has a greater impact on temperature prediction.

[0145] Step 600: Utilize the parameter information, spectral energy value and weight value of the triangular facet obtained after the secondary adjustment in step S404, and calculate the radiation temperature of the turbine blade to be measured based on the emissivity model and the temperature inversion algorithm;

[0146] Specifically, a multi-spectral temperature measurement algorithm is used to process the parameter information and spectral energy value of each face element after secondary adjustment, and Planck's law is applied to calculate the radiation brightness temperature of each face element in different bands. These radiation brightness temperatures reflect the temperature conditions of each face element in different bands. After calculating the radiation brightness temperature of each face element in each band, the radiation brightness temperature is weighted using the face element weight value after secondary adjustment obtained in step S404. The radiation brightness temperature of each face element in different bands is multiplied by its corresponding weight to obtain the weighted radiation brightness temperature. The weighted temperature values ​​of all face elements on the side to be measured are combined (added) and divided by the total number of face elements to obtain the final weighted average temperature of the entire surface to be measured, which is the radiation temperature measurement value of the surface to be measured.

[0147] The multi-spectral temperature measurement method based on the facet optimization processing technology provided by the present invention selects the band of 1100nm-1450nm and performs measurement at 1200K. The experiment is divided into two groups. In the first group, the surface to be measured of the turbine blade is roughly divided into triangular facets using only the midpoint method. In the second group, the surface to be measured of the turbine blade is finely divided based on multiple factors such as energy and temperature using the method of the present invention. The subsequent temperature inversion steps are the same. Using the traditional division method, the temperature measured at 1200K is 1224.41K, with a relative error of 2.034%; using the division method of the present invention, the temperature measured at 1200K is 1218.44K, with a relative error of 1.5368%.

[0148] Beneficial effects of this application:

[0149] 1. The present invention can provide higher spatial resolution in key hotspot areas by finely dividing the turbine blade model into triangular facets and dynamically adjusting it according to the actual impact of temperature gradient and reflected radiation, allowing the system to more accurately map and analyze the temperature distribution of high-temperature turbine blades;

[0150] 2. The present invention uses deep learning technology, combined with multi-spectral temperature measurement data and loss function, to optimize the shape of triangular surface elements so that they better fit the actual geometric shape of the blade, avoiding the use of vertically mapped rectangular surfaces for approximate measurement in traditional temperature measurement, avoiding data errors caused by unreasonable shapes, and improving the accuracy of temperature measurement data;

[0151] 3. The present invention adaptively adjusts the facet shape to avoid the occurrence of multiple irregular and distorted triangular facets in the facet division process due to the influence of curvature, which affects data processing;

[0152] 4. The present invention analyzes the quality of spectral data in each band and adaptively adjusts the data weight to ensure that high-quality spectral data has a greater impact on temperature prediction. This weight adjustment mechanism allows the system to automatically optimize the temperature inversion algorithm based on real-time data quality, improves the reliability and stability of the temperature measurement process, and is suitable for temperature measurement environments with a large impact of reflected radiation.

[0153] The number of devices and processing scales described here are used to simplify the description of the present invention. Applications, modifications and variations of the present invention will be obvious to those skilled in the art.

[0154] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A multi-spectral temperature measurement method based on facet optimization processing technology, characterized in that: The following steps are involved: Step 1: Obtain spectral information of the turbine blade to be tested, and convert the spectral information into spectral energy value; Step 2: Obtain a 3D model of the turbine blade to be tested, and divide each curved surface of the turbine blade into a plurality of triangular face elements; Step 3: based on the angle threshold between the face element normal vector and the reference direction vector, extract the parameter information and spectral energy value of the triangular face element of the target surface, calculate the energy gradient of each triangular face element on the target surface, adjust the triangular face element once based on the energy gradient, obtain the triangular face element after the adjustment, and extract the parameter information and spectral energy value of the triangular face element after the adjustment; wherein the parameter information of the triangular face element includes the area, side length and angle of the triangular face element; Step 4: Based on the neural network model technology, the triangular surface element after the primary adjustment is adjusted for a second time to obtain parameter information of the triangular surface element after the secondary adjustment, and the spectral energy value of the triangular surface element after the secondary adjustment is extracted; Step 5: Calculate the weight W of the triangular face after secondary adjustment based on formula (1) i ; Formula (1); Where, k in formula (1) is the adjustment coefficient, x i is the data quality score of the ith triangle face, x mean is the average of the data quality scores of all triangle faces; In formula (1), x i is obtained by calculating formula (5), x mean It is obtained by calculating formula (6); x i =SNR i *(1 - R i ) Equation (5); Formula (6); Among them, SNR in formula (5) i is the signal-to-noise ratio of the i-th bin, which is calculated by formula (7); is the adjustment factor, which is calculated by formula (8); in formula (6), N is the total number of bins; Formula (7); Formula (8); Among them, P in formula (7) signal,i and P noise,i Represent the signal power and noise power of the i-th bin respectively; in formula (8), T i is the temperature of the ith facet; θ is the evaluation factor, and its initial value is set to 0.1; In step 3, the energy gradient of each triangular face is calculated based on formula (3); △E=[ max(E)- min(E)] / S Formula (3); In formula (3), max(E) and min(E) represent the highest spectral energy value and the lowest spectral energy value inside the triangular surface element, respectively, and S is the area of ​​the triangular surface element; Using formula (4) based on the energy gradient of the triangular face element, the area size of the triangular face element is adjusted; S i =αexp(-β△E i ) Formula (4); In formula (4), S i represents the area of ​​the i-th triangle face after one adjustment, △E i is the energy gradient at the center of the triangular face, α and β are adjustment parameters, the initial value of α is 0.02, and the initial value of β is 0.05; Step 4 specifically includes: Step S4.1, using the spectral energy value and parameter information of the triangular face element before adjustment as input values ​​and the parameter information of the triangular face element after adjustment as output values, constructing a neural network model 1; using the historical sample data of the turbine blades, training, testing and verifying the neural network model 1; Step S4.2, inputting the parameter information and spectral energy value of the triangular facet after the primary adjustment in the historical sample data of the turbine blade into the neural network model 1 as input values, and obtaining the parameter information of the triangular facet after the secondary adjustment corresponding to the historical sample data; Step S4.3, using the parameter information and spectral energy value of the triangular face element after the primary adjustment as input values, and the parameter information of the triangular face element after the secondary adjustment as output values, constructing a second neural network model, and using the historical sample data of the turbine blades to train, test and verify the second neural network model; Step S4.4, inputting the parameter information and spectral energy value of the triangular surface element after the primary adjustment of the target surface into the trained neural network model 2 as input values, to obtain the parameter information of the triangular surface element after the secondary adjustment; Step 6: Using the obtained parameter information of the triangular facet after secondary adjustment, the spectral energy value and the weight value of the triangular facet, based on the emissivity model and the temperature inversion algorithm, the radiation temperature of the turbine blade to be measured is calculated.

2. The multi-spectral temperature measurement method based on the facet optimization processing technology according to claim 1, characterized in that: In step 1, the spectral information is converted into spectral energy value based on formula (2); P=10*log 10 L formula (2); In formula (2), P is the spectral information, the unit is dBm, and L is the spectral energy value, the unit is mW.

3. The multi-spectral temperature measurement method based on the facet optimization processing technology as claimed in claim 2, characterized in that: Step 2 is as follows: S2.1 Take the midpoint of each side length of the curved surface of the 3D model of the turbine blade to be tested, connect adjacent midpoints in sequence, and divide the curved surface into a plurality of primary triangular face elements; S2.2, according to step S2.1, divide each primary triangular face into multiple secondary triangular face elements by taking the midpoint of the side length; S2.3, repeat step S2.2 to perform multiple face element divisions until the number of divisions is equal to the preset division level value, thereby obtaining multiple triangular face elements.

4. The multi-spectral temperature measurement method based on the facet optimization processing technology as claimed in claim 3 is characterized in that: In step 4, the neural network model is a hybrid architecture formed by a convolutional neural network combined with a fully connected layer; the convolutional layer includes three layers of convolution, each convolution layer is followed by batch normalization and ReLU activation function, and each convolution layer is followed by a pooling layer; the fully connected layer includes two fully connected layers with 128 and 64 nodes respectively; The batch size of the neural network model is 32, the number of iterations is 50, and the learning rate is initially set to 0.

001. K-fold cross validation is used to evaluate the generalization ability of the neural network model, and the quality of the output results is evaluated by the average mean square error on the test set.

5. A storage medium having a computer program stored therein, characterized in that: When the computer program is loaded and executed by a processor, the method according to any one of claims 1 to 4 is implemented.

6. An electronic device, characterized in that include: A processor and a memory; wherein the memory is used to store a computer program; and the processor is used to load and execute the computer program so that the electronic device executes the method according to any one of claims 1 to 4.

Images