Multi-spectral radiation temperature measurement method and device based on MEA-LSSVM under high-temperature reflection background

By adopting the MEA-LSSVM method and OBKA algorithm in multispectral radiation temperature measurement technology, the problem of large errors in target emissivity model identification and temperature measurement under the background of high temperature reflection is solved, and higher recognition accuracy and temperature inversion accuracy are achieved .

CN119089627AActive Publication Date: 2024-12-06HARBIN ENG UNIV
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202410511436.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-04-26
Publication Date
2024-12-06
Estimated Expiration
2044-04-26

AI Technical Summary

Technical Problem

The existing multispectral radiation temperature measurement technology has problems such as inaccurate target emissivity model and large temperature measurement errors in the context of high temperature reflection, making it difficult to accurately identify target emissivity model and reduce inversion temperature errors.

Method used

Using a multispectral radiation temperature measurement method based on MEA-LSSVM, the emissivity model is constructed, the MIV algorithm is used to extract features, and the least squares support vector machine is optimized in combination with the thinking evolution algorithm, the target emissivity model is identified, and the improved OBKA algorithm is used to inverse it. Performance target temperature.

Benefits of technology

It improves the recognition accuracy of the target emissivity model and the accuracy of temperature inversion, reduces the inversion error, and is suitable for temperature measurement of unknown objects in the context of high-temperature environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089627A_ABST
    Figure CN119089627A_ABST
Patent Text Reader

Abstract

The invention discloses a multispectral radiation temperature measurement method and equipment based on MEA-LSSVM under a high-temperature reflection background, belongs to the technical field of radiation temperature measurement, and solves the problem of poor measurement accuracy of turbine blade radiation temperature measurement caused by the influence of reflection radiation of a complex environment on a turbine blade. The method comprises the following steps: firstly, generating spectrum samples within a specific wavelength range and at different temperatures; as the number of samples is large, in order to better extract feature information of multispectral radiation temperature measurement data, the method adopts an average influence value MIV algorithm to solve the feature importance of the spectral samples, then the features are put into a least square support vector machine MEA-LSSVM optimized by a mind evolutionary algorithm to be trained, the target spectral emissivity is recognized, and the target spectral emissivity is obtained. And finally, the target temperature is inversed by adopting an optimized Black-Wing Yuokuan algorithm OBKA. The method is suitable for temperature inversion of the unknown object under the high-temperature environment background.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of radiation temperature measurement, and in particular to multi-spectral radiation temperature measurement. Background Art

[0002] Aircraft engines and gas turbines represent a country's industrial level and are known as the "crown jewel" of the equipment manufacturing industry. Turbine blades are key components of engines. With the development of aerospace technology, the performance parameters of engines have gradually improved. At this time, the high temperature that turbine blades have to withstand during operation has increased significantly. It is of great significance to study the temperature measurement of turbine blades under the background of high temperature reflection.

[0003] Multispectral radiation temperature measurement is a non-contact temperature measurement method that can be performed in a high-temperature environment. It determines the surface temperature of an object by measuring the multispectral radiation energy emitted by the surface of the object. Compared with contact temperature measurement methods such as thin-film thermocouples and temperature-indicating paint, the multispectral radiation temperature measurement method can quickly obtain the temperature of the surface of an object, has higher temperature measurement accuracy, and is suitable for various scenarios such as high temperature and moving objects. However, multispectral radiation temperature measurement technology currently has problems such as inaccurate target emissivity model and large temperature measurement error. Therefore, how to accurately identify the target emissivity model and reduce the inversion temperature error is an important issue that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that the existing turbine blades are affected by the reflected radiation of the complex environment, resulting in poor real-time and accuracy of the radiation temperature measurement of the turbine blades. A multi-spectral radiation temperature measurement method and equipment based on MEA-LSSVM under a high-temperature reflection background is provided.

[0005] The present invention is implemented by the following technical solutions. On the one hand, the present invention provides a multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background, and the method comprises:

[0006] Step 1: construct an emissivity model and generate a sample spectrum using the emissivity model;

[0007] Step 2: Extract features from the sample spectrum and pyrometer measured radiation data, extract the features of the data through the MIV algorithm and arrange the feature importance to generate a feature matrix;

[0008] Step 3: Use the mind evolution algorithm to optimize the least squares support vector machine, obtain the optimized least squares support vector machine classifier MEA-LSSVM, and use MEA-LSSVM to identify the target emissivity model;

[0009] Step 4: After obtaining the target emissivity model, the target temperature is inverted using the OBKA algorithm, which is an improved BKA algorithm and includes:

[0010] The mathematical model of attack and migration is:

[0011]

[0012] in, and denote the position of the ith black-winged kite in the jth dimension and the (t+1)th iteration step, r is a random number between 0 and 1, p is a constant, represents the leading scorer of the black kite in the jth dimension at the tth iteration so far, F i represents the j-th dimension current position of any black-winged kite in the t-th iteration, F ri represents the fitness value of any black kite at a random position in the jth dimension in the tth iteration, C is a random number, and n is a natural number.

[0013] Furthermore, step 1 specifically includes:

[0014] A sample spectrum set is generated based on a wavelength of 1-2.5 μm with a total of 15 wavelengths at intervals of 0.1 μm, a temperature range of 500°C-1500°C, and a difference of 50°C between two adjacent temperature points;

[0015] The emissivity models include: linear rise and fall, exponential rise and fall, quadratic polynomial opening up and down, and sinusoidal emissivity models.

[0016] Furthermore, step 2 specifically includes:

[0017] Calculate several indicators for each set of data, including mean, variance, peak, kurtosis and effective value;

[0018] Through the MIV algorithm, multiple indicators with high feature importance are extracted.

[0019] Furthermore, in step 2, the generating of the feature matrix is ​​specifically:

[0020] The features with higher feature importance are fused. Assuming the features are Y1, Y2, Y3…, the feature matrix Y is generated by serial fusion:

[0021]

[0022] Further, in step 3, the step of obtaining the optimized least squares support vector machine classifier MEA-LSSVM specifically includes:

[0023] Step 3.1, input the feature matrix Y into the LSSVM classifier, set the thinking evolution algorithm parameters, generate the initial population, the winning sub-population and the temporary population;

[0024] Step 3.2: Perform convergence and alienation operations on each of the winning sub-groups and the temporary sub-groups to obtain mature winning sub-groups and mature temporary sub-groups. The convergence and alienation operations are as follows:

[0025] X(t+1)=X(t)+C*(MX(t))+D*(NX(t))

[0026] Among them, X(t) represents the number of individuals at time t, and X(t+1) represents the number of individuals at time t+1; C and D represent the convergence coefficient and alienation coefficient, which determine the degree of convergence and alienation of individuals to group behavior; M and N represent the average behavior level in the group and the degree of difference between individuals, respectively;

[0027] When C>0, individuals tend to converge with the average behavior level of the group, that is, the number of individuals will move closer to the average level; when C<0, individuals tend to alienate from the average behavior level of the group, that is, the number of individuals will move away from the average level;

[0028] When D>0, the degree of difference between individuals will increase, that is, the number of individuals will be more dispersed; when D<0, the degree of difference between individuals will decrease, that is, the number of individuals will be more concentrated;

[0029] Step 3.3: When the number of iterations is met, the optimal individual found is output as the kernel function and penalty function of the LSSVM model to obtain the MEA-LSSVM emission rate model classifier.

[0030] Further, step 4 includes:

[0031] After obtaining the target emissivity model, the multi-spectral radiation temperature measurement equation is constructed according to Planck's blackbody radiation law, and the target equation is solved using the OKBA algorithm to invert the target temperature.

[0032] Furthermore, the restriction conditions of the OKBA algorithm are:

[0033]

[0034] Where, ΔT min is the temperature difference, and F[ε(λ, T)] is the emissivity function.

[0035] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor runs the computer program stored in the memory, the steps of the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high-temperature reflection background as described above are executed.

[0036] In a third aspect, the present invention provides a computer-readable storage medium, in which a plurality of computer instructions are stored, and the plurality of computer instructions are used to enable a computer to execute a multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high-temperature reflection background as described above.

[0037] In a fourth aspect, the present invention provides an electronic device, comprising:

[0038] at least one processor; and,

[0039] a memory communicatively connected to the at least one processor; wherein,

[0040] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high-temperature reflection background as described above.

[0041] Beneficial effects of the present invention:

[0042] The present invention proposes a multi-spectral radiation temperature measurement method based on MEA-LSSVM considering the background of high temperature reflection. First, spectrum samples at different temperatures within a specific wavelength range are generated. Due to the large number of samples, in order to better extract the characteristic information of multi-spectral radiation temperature measurement data, the method adopts the mean influence value (MIV) algorithm to calculate the importance of the spectral sample features, and then puts the features into the least squares support vector machine (MEA-LSSVM) optimized by the thinking evolution algorithm for training, identifies the target spectral emissivity, and finally uses OBKA (optimized black kite algorithm) to invert the target temperature.

[0043] The present invention proposes a MEA-LSSVM method to identify a target spectral emissivity model. The training sample is a fusion matrix after the features are extracted using the MIV algorithm, and the target spectral emissivity is highly recognized with high accuracy.

[0044] The present invention improves the Black Kite Algorithm (BKA), and uses the improved OBKA multi-spectral radiation temperature measurement algorithm to perform temperature inversion on theoretically generated targets and thermal barrier coating samples with small errors, thereby improving the recognition accuracy of the target emissivity model and reducing the inversion temperature error.

[0045] The present invention is applicable to the temperature inversion of unknown objects in a high temperature environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the embodiments are briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0047] Figure 1 This is the flow chart of the multi-spectral radiation temperature measurement method based on MEA-LSSVM;

[0048] Figure 2 This is the flow chart of temperature inversion based on OBKA algorithm. DETAILED DESCRIPTION

[0049] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0050] Embodiment 1: A multi-spectral radiation temperature measurement method based on MEA-LSSVM under high temperature reflection background, the method comprising:

[0051] Step 1: construct an emissivity model and generate a sample spectrum using the emissivity model;

[0052] Step 2: Extract features from the sample spectrum and pyrometer measured radiation data, extract the features of the data through the MIV algorithm and arrange the feature importance to generate a feature matrix;

[0053] Step 3: Use the mind evolution algorithm to optimize the least squares support vector machine, obtain the optimized least squares support vector machine classifier MEA-LSSVM, and use MEA-LSSVM to identify the target emissivity model;

[0054] Step 4: After obtaining the target emissivity model, the target temperature is inverted using the OBKA algorithm, which is an improved BKA algorithm and includes:

[0055] The mathematical model of attack and migration is:

[0056]

[0057] in, and denote the position of the ith black-winged kite in the jth dimension and the (t+1)th iteration step, r is a random number between 0 and 1, p is a constant, represents the leading scorer of the black kite in the jth dimension at the tth iteration so far, F irepresents the j-th dimension current position of any black-winged kite in the t-th iteration, F ri represents the fitness value of any black kite at a random position in the jth dimension in the tth iteration, C is a random number, and n is a natural number.

[0058] In this embodiment, a MEA-LSSVM method is proposed to identify the target spectral emissivity model. The training sample is a fusion matrix after the features are extracted using the MIV algorithm, and the target spectral emissivity recognition accuracy is high. The improved algorithm OBKA multi-spectral radiation temperature measurement algorithm is used to perform temperature inversion on the theoretically generated target and thermal barrier coating samples, and the error is small. The accuracy of target emissivity model recognition is improved and the inversion temperature error is reduced.

[0059] It should be noted that this implementation combines the advantages of the Black Kite Optimization Algorithm and adaptively improves the Black Kite Algorithm BKA. The improved OBKA multi-spectral radiation temperature measurement algorithm is used to perform temperature inversion on the theoretically generated targets and thermal barrier coating samples with a small error, thereby improving the recognition accuracy of the target emissivity model and reducing the inversion temperature error.

[0060] Embodiment 2: This embodiment further defines the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background described in Embodiment 1. In this embodiment, step 1 is further defined, specifically including:

[0061] Step 1 specifically includes:

[0062] A sample spectrum set is generated based on a wavelength of 1-2.5 μm with a total of 15 wavelengths at intervals of 0.1 μm, a temperature range of 500°C-1500°C, and a difference of 50°C between two adjacent temperature points;

[0063] The emissivity models include: linear rise and fall, exponential rise and fall, quadratic polynomial opening up and down, and sinusoidal emissivity models.

[0064] In this implementation manner, the emissivity model may be a common emissivity model.

[0065] Implementation method three: This implementation method is a further limitation of the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background described in implementation method one. In this implementation method, step 2 is further limited, specifically including:

[0066] Step 2 specifically includes:

[0067] Calculate several indicators for each set of data, including mean, variance, peak, kurtosis and effective value;

[0068] Through the MIV algorithm, multiple indicators with high feature importance are extracted.

[0069] In this embodiment, the MIV algorithm is used to calculate the feature importance of the data to obtain the most important feature indicators. The mean impact value (MIV) is an indicator for evaluating the degree of influence of the input independent variable on the output variable. The positive or negative value of MIV indicates the direction of influence of the independent variable on the output variable, and its absolute value indicates the degree of influence. MIV constructs two new training samples by adding or subtracting 10% from the independent variable feature indicator value, calculates the impact change value (IV) on the output, and then averages the impact change value (IV) to obtain the MIV value of the independent variable. Finally, the above steps are repeated for each independent variable to obtain the MIV value of each independent variable, and the degree of influence of each variable is sorted by absolute value.

[0070] Several features with greater importance of the sample spectrum are fused to generate a feature matrix.

[0071] Embodiment 4: This embodiment further limits the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background described in Embodiment 3. In this embodiment, the generating feature matrix in step 2 is further limited, specifically including:

[0072] In step 2, the feature matrix is ​​generated as follows:

[0073] The features with higher feature importance are fused. Assuming the features are Y1, Y2, Y3…, the feature matrix Y is generated by serial fusion:

[0074] Y=[Y1,Y2…Y n ]n=1,2,3…。

[0075] In this implementation, a method for fusing several features of the sample spectrum with greater feature importance in step 2 is provided.

[0076] Embodiment 5: This embodiment further limits the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background described in Embodiment 1. In this embodiment, in step 3, the optimized least squares support vector machine classifier MEA-LSSVM is obtained, which is further limited, specifically including:

[0077] In step 3, the step of obtaining an optimized least squares support vector machine classifier MEA-LSSVM specifically includes:

[0078] Step 3.1, input the feature matrix Y into the LSSVM classifier, set the thinking evolution algorithm parameters, generate the initial population, the winning sub-population and the temporary population;

[0079] Step 3.2: Perform convergence and alienation operations on each of the winning sub-groups and the temporary sub-groups to obtain mature winning sub-groups and mature temporary sub-groups. The convergence and alienation operations are as follows:

[0080] X(t+1)=X(t)+C*(MX(t))+D*(NX(t))

[0081] Among them, X(t) represents the number of individuals at time t, and X(t+1) represents the number of individuals at time t+1; C and D represent the convergence coefficient and alienation coefficient, which determine the degree of convergence and alienation of individuals to group behavior; M and N represent the average behavior level in the group and the degree of difference between individuals, respectively;

[0082] When C>0, individuals tend to converge with the average behavior level of the group, that is, the number of individuals will move closer to the average level; when C<0, individuals tend to alienate from the average behavior level of the group, that is, the number of individuals will move away from the average level;

[0083] When D>0, the degree of difference between individuals will increase, that is, the number of individuals will be more dispersed; when D<0, the degree of difference between individuals will decrease, that is, the number of individuals will be more concentrated;

[0084] Step 3.3: When the number of iterations is met, the optimal individual found is output as the kernel function and penalty function of the LSSVM model to obtain the MEA-LSSVM emission rate model classifier.

[0085] In this implementation, LSSVM has achieved successful results in function estimation and nonlinear system identification, and has been widely used in practical applications. In order to better determine the kernel function and penalty parameters in the least squares support vector machine, the MEA algorithm is introduced. This algorithm draws on the evolution of human thinking, proposes two new "convergence" and "alienation" operation operators, and continuously iterates these operations to enable LSSVM to select the optimal kernel function and penalty parameters. The basic process of the algorithm is as follows Figure 2 As shown. Finally, the accuracy of LSSVM in identifying the firing rate model is improved.

[0086] Embodiment 6: This embodiment further defines the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background described in Embodiment 1. In this embodiment, step 4 is further defined, specifically including:

[0087] Step 4 includes:

[0088] After obtaining the target emissivity model, the multi-spectral radiation temperature measurement equation is constructed according to Planck's blackbody radiation law, and the target equation is solved using the OKBA algorithm to invert the target temperature.

[0089] In this implementation, the black kite optimization algorithm draws on the flight and predation behavior of black kites in nature, as well as their adaptability to environmental changes and target locations. Combining the advantages of the improved black kite optimization algorithm, a multi-spectral radiation temperature measurement equation is constructed based on Planck's blackbody radiation law to improve the effectiveness of the final inversion of the target temperature.

[0090] Embodiment 7: This embodiment further limits the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background described in Embodiment 1. In this embodiment, the restriction conditions of the OKBA algorithm are further limited, specifically including:

[0091] The constraints of the OKBA algorithm are:

[0092]

[0093] Where, ΔT min is the temperature difference, and F[ε(λ, T)] is the emissivity function.

[0094] In this implementation, the restriction conditions of the improved KBA algorithm in inverting the target temperature are given to achieve the purpose of multi-spectral radiation temperature measurement of the present application.

[0095] Embodiment 8: This embodiment is Embodiment 1 of the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background as described above, and specifically includes:

[0096] First, spectral samples at different temperatures within a specific wavelength range are generated. Due to the large number of samples, in order to better extract the characteristic information of multi-spectral radiation temperature measurement data, this method uses the mean influence value (MIV) algorithm to calculate the importance of spectral sample features, and then puts the features into the least squares support vector machine optimized by the thinking evolution algorithm (MEA-LSSVM) for training to identify the target spectral emissivity, and finally uses OOBKA (optimized black kite algorithm) to invert the target temperature. It mainly includes the following 4 steps:

[0097] Step 1: Construct a common emissivity model and theoretically generate a sample spectrum, that is, use the existing emissivity model to theoretically generate a sample spectrum.

[0098] Step 2: Extract features from the input sample spectrum set and pyrometer measured radiation data, extract the features of the data through the MIV algorithm and rank the feature importance.

[0099] Step 3: The least squares support vector machine optimized by the mind evolution algorithm (MEA-LSSVM) is used for training and model building. The algorithm uses the MEA algorithm to initialize the kernel function and penalty parameters of the least squares support vector machine, thereby improving the accuracy of the emissivity model classification.

[0100] Step 4: After obtaining the target emissivity model, use OBKA to invert the target temperature.

[0101] Furthermore, step 1 is specifically as follows: theoretically generate a spectral data sample set, add noise to the generated sample set to increase data authenticity, and improve the effectiveness and robustness of the method of the present invention. Preprocess the used data set and construct a common emissivity model.

[0102] Furthermore, step 2 is specifically as follows: calculating the mean, variance, peak value, kurtosis, effective value and other indicators of each group of data, and screening out several indicators with higher feature importance through the MIV algorithm, so as to extract the characteristics of samples and targets through the MIV algorithm.

[0103] Furthermore, step 3 is specifically as follows: LSSVM has achieved successful results in function estimation and nonlinear system identification and has been widely used in practical applications.

[0104] In order to better determine the kernel function and penalty parameters in the least squares support vector machine, the MEA algorithm is introduced. This algorithm draws on the evolution of human thinking and proposes two new "convergence" and "alienation" operation operators. Through continuous iteration of these operations, LSSVM can select the optimal kernel function and penalty parameters. The basic process of the algorithm is as follows Figure 2 As shown. Finally, the accuracy of LSSVM in identifying the firing rate model is improved.

[0105] Furthermore, step 4 is as follows: The black-winged kite optimization algorithm draws on the flight and predation behavior of black-winged kites in nature, as well as their adaptability to environmental changes and target positions. After obtaining the target emissivity model, a multi-spectral radiation temperature measurement equation is constructed according to Planck's blackbody radiation law, and the target equation is solved using the OKBA algorithm to invert the target temperature.

[0106] Implementation method nine: This implementation method is implementation method 2 of the multi-spectral radiation temperature measurement method based on MEA-LSSVM under a high temperature reflection background as described above, and specifically includes:

[0107] This embodiment provides a multi-spectral temperature measurement method based on MEA-LSSVM considering high temperature background, which is used to solve the problems of low recognition rate of emissivity model and large temperature difference in measurement. The basic steps are as follows: Figure 1 shown.

[0108] Step 1: Input the multi-spectral radiation data of the pyrometer measurement sample and construct common emissivity models (including linear, exponential, quadratic polynomial, and sine). Use the four emissivity models to establish a training sample set, and the spectral data of the pyrometer measurement sample is used as the test set.

[0109] Step 2: Process the input radiation data, calculate the mean, variance, peak, kurtosis, effective value and other indicators of each group of data, and screen out several indicators with higher feature importance through MIV calculation, and form a feature matrix from these indicators.

[0110] Step 3: The least squares support vector machine classifier was improved, the parameters of LSSVM were optimized through the MEA algorithm, and the MEA-LSSVM classifier was used to identify the target emissivity model. The subsequent application of it to the inversion temperature can make the calculated temperature results more accurate.

[0111] Step 4: After obtaining the target emissivity model, the target temperature is inverted using the OBKA algorithm.

[0112] Step 1 specifically includes:

[0113] The sample data used is the generated sample spectrum, with a wavelength of 1-2.5μm and a total of 15 wavelengths every 0.1μm, a temperature range of 500℃-1500℃, and a difference of 50℃ between two adjacent temperature points. After generation, a total of 7000*15 sample spectrum data sets are generated. The multispectral radiation data measured by the pyrometer is a 500*15 sample. The 7 common emissivity models are constructed: linear rise and fall, exponential rise and fall, quadratic polynomial opening up and down, and sine.

[0114] Step 2 specifically includes:

[0115] By calculating the mean, variance, peak, kurtosis, effective value and other indicators of each group of data, the MIV algorithm is used to calculate the feature importance of the data to obtain the most important feature indicators. The mean impact value (MIV) is an indicator for evaluating the degree of influence of the input independent variable on the output variable. The positive or negative value of MIV indicates the direction of the influence of the independent variable on the output variable, and its absolute value indicates the degree of influence. MIV constructs two new training samples by adding or subtracting 10% from the independent variable feature indicator value, calculates its impact change value (IV) on the output, and then averages the impact change value (IV) to obtain the MIV value of the independent variable. Finally, the above steps are repeated for each independent variable to obtain the MIV value of each independent variable, and the degree of influence of each variable is sorted by absolute value.

[0116] Fuse several features of the sample spectrum with greater importance. Assuming the features are Y1, Y2, Y3…, we can use the method of serial fusion, and the formula is:

[0117] Y=[Y1,Y2…Y n ]n=1,2,3…

[0118] The feature matrix Y will be used as a sample for spectral recognition.

[0119] Step 3 specifically includes:

[0120] The MEA algorithm is used to optimize the kernel function and penalty parameters of LSSVM, and the sample spectrum is input to train the classifier to identify the target spectrum.

[0121] Suppose there is a training data set {(x1,y1),(x2,y2),...,(xn,yn)}, where xn is the input feature vector and yn is the corresponding output value. For example, xn is the feature vector of spectral data and yn is the label corresponding to four emissivity models.

[0122] The optimization problem of the least squares support vector machine is transformed into:

[0123]

[0124] sty i (W·x i + b) = 1-e i , i=1,2,...,m

[0125] In the above formula, l is the regularization parameter, W is the hyperplane normal vector, b is a constant, and e is i is the error variable, λ is the penalty factor, and m is a natural number. When the training samples are nonlinearly separable, the optimization problem of the least squares support vector machine is a QP problem with equality constraints when the initial samples are mapped to a higher linearly separable space. In order to solve this problem, the Lagrangian function method of the optimization problem is listed:

[0126]

[0127] In the formula, α i Denotes the Lagrange multiplier corresponding to xi. After taking the derivatives of all its quantities using the Lagrange function method, their derivatives are set to 0:

[0128]

[0129] Solve for α=[α1, α2,…,α m ] T and b can be obtained by the following equation:

[0130]

[0131] In the formula I is the identity matrix, K is the m*m kernel matrix,

[0132]

[0133] In the formula, k is the kernel function and γ is the kernel parameter. This solution is simpler than solving the QP problem.

[0134] The model of the least squares support vector machine is obtained as follows:

[0135]

[0136] Among them, b is a constant and N is a natural number.

[0137] (1) Input the feature matrix Y extracted in step 2 into the LSSVM classifier, set the parameters of the thinking evolution algorithm, and generate the initial population, the winning sub-population and the temporary population.

[0138] (2) Performing convergence and alienation operations on each of the superior sub-groups and the temporary sub-groups respectively to obtain a mature superior sub-group and a mature temporary sub-group. The convergence and alienation operations are as follows:

[0139] X(t+1)=X(t)+C*(MX(t))+D*(NX(t))

[0140] Among them, X(t) represents the number of individuals at time t, and X(t+1) represents the number of individuals at time t+1. C and D represent the convergence coefficient and alienation coefficient, respectively, which determine the degree of convergence and alienation of individuals with respect to group behavior. M and N represent the average behavior level in the group and the degree of difference between individuals, respectively.

[0141] When C>0, individuals tend to converge with the average behavior level of the group, that is, the number of individuals will move closer to the average level. When C<0, individuals tend to alienate from the average behavior level of the group, that is, the number of individuals will move away from the average level.

[0142] When D>0, the degree of difference between individuals will increase, that is, the number of individuals will be more dispersed. When D<0, the degree of difference between individuals will decrease, that is, the number of individuals will be more concentrated.

[0143] (3) When the number of iterations is met, the optimal individual found is output as the kernel function and penalty function of the LSSVM model to obtain the MEA-LSSVM emission rate model classifier.

[0144] Step 4 specifically includes:

[0145] Since the target environment has high temperature background reflection and self-emissivity, the target radiation equation is expressed as:

[0146] M(λ,T)=ελ M(λ,T r )+(1-ε λ )M(λ,T b )

[0147] Where λ is the wavelength, T is the temperature calculated by the pyrometer measuring the radiation emittance received by the object; T r is the true temperature of the object; T b is the high temperature background temperature; M(λ,T) is the radiation emittance received by the detector, M(λ,T r ) is the black body radiation emittance at the same temperature as the object; M(λ,T b ) The radiation emittance from the high temperature background to the object; ε λ is the emissivity of the surface of the object being measured; 1-ε λ is the reflectivity of the object being measured.

[0148] For a multispectral radiation pyrometer with N spectral channels, the apparent radiation of the measured object at each wavelength can be expressed as:

[0149]

[0150] Further expressed as:

[0151]

[0152] Among them, λi is the wavelength and εi is the emissivity.

[0153] It can be seen from the equation that there are N equations in total, but there are N+1 unknown parameters, and the true temperature cannot be solved directly. Therefore, it is necessary to reasonably assume the spectral emissivity model for solution and convert it into the problem of solving the emissivity model coefficients.

[0154] The expression for the true temperature of an object can be obtained as follows:

[0155]

[0156] Among them, T r (i) is the temperature, c1 is the first radiation constant, and c2 is the second radiation constant.

[0157] And the object temperature solved should be equal in theory, and the range of emissivity is (0,1), as shown below:

[0158]

[0159] Where, ΔT min is the temperature difference, and F[ε(λ, T)] is the emissivity function.

[0160] This embodiment will use the above conditions to restrict the OBKA inversion temperature algorithm and ultimately solve the true temperature of the object.

[0161] The OBKA algorithm, with its unique bio-inspired features, not only captures the flight and predation behaviors of black kites in nature, but also deeply simulates the black kites’ high adaptability to environmental changes and target locations. The steps are as follows:

[0162] (1) Initialization of the initial population

[0163] To create a set of random solutions, the following matrix can be used to represent the position of each black kite:

[0164]

[0165] Where pop is the number of ways to solve the equation, dim is the dimension of the solution to the radiation equation, and Bi j is the jth dimension of the ith black kite. Randomly assign the position and velocity of each black kite.

[0166] v i =w×v i +c1×rand()×(pbest i -x i )+c2×rand()×(gbest i -x i )

[0167] x i =x i +v i i=1,2,…,N

[0168] Where N is the total number of black kites; vi is the speed of the black kite; xi is the current position of the black kite; rand() is a random number between (0,1); c1, c2 are learning factors, and w is a random number.

[0169] During the initialization process, OBKA randomly selects a black-winged kite as the leader, which is considered to be the optimal solution of the equation.

[0170] (2) Aggression and migration behavior

[0171] In the global exploration and search for the optimal solution of the equation, we learn from the attack and migration behavior of the black-winged kite to avoid local optimality. The following is the mathematical model of attack and migration:

[0172]

[0173] Here is the definition of the equation:

[0174] and denotes the position of the ith black-winged kite in the jth dimension and the (t+1)th iteration step, respectively. r is a random number between 0 and 1, and p is a constant of 0.9. F represents the leading scorer of black kites in the jth dimension at the tth iteration so far. i Fr represents the j-th dimension current position of any black-winged kite in the t-th iteration. i represents the fitness value of any black kite at a random position in the jth dimension in the tth iteration, C is a random number, and n is a natural number.

[0175] After obtaining the target emissivity model, the OBKA multi-spectral radiation temperature measurement algorithm is used to invert the target true temperature, such as Figure 2 shown.

Claims

1. A multi-spectral radiation temperature measurement method based on MEA-LSSVM under high temperature reflection background, characterized in that: The method comprises: Step 1: construct an emissivity model and generate a sample spectrum using the emissivity model; Step 2: Extract features from the sample spectrum and pyrometer measured radiation data, extract the features of the data through the MIV algorithm and arrange the feature importance to generate a feature matrix; Step 3: Use the mind evolution algorithm to optimize the least squares support vector machine, obtain the optimized least squares support vector machine classifier MEA-LSSVM, and use MEA-LSSVM to identify the target emissivity model; Step 4: After obtaining the target emissivity model, the target temperature is inverted using the OBKA algorithm, which is an improved BKA algorithm and includes: The mathematical model of attack and migration is: in, and denote the position of the ith black-winged kite in the jth dimension and the (t+1)th iteration step, r is a random number between 0 and 1, p is a constant, represents the leading scorer of the black kite in the jth dimension at the tth iteration so far, F i represents the j-th dimension current position of any black-winged kite in the t-th iteration, F ri represents the fitness value of any black kite at a random position in the jth dimension in the tth iteration, C is a random number, and n is a natural number.

2. According to the multi-spectral radiation temperature measurement method based on MEA-LSSVM under high temperature reflection background in claim 1, it is characterized in that: Step 1 specifically includes: A sample spectrum set is generated based on a wavelength of 1-2.5 μm with a total of 15 wavelengths at intervals of 0.1 μm, a temperature range of 500°C-1500°C, and a difference of 50°C between two adjacent temperature points; The emissivity models include: linear rise and fall, exponential rise and fall, quadratic polynomial opening up and down, and sinusoidal emissivity models.

3. The multi-spectral radiation temperature measurement method based on MEA-LSSVM under high temperature reflection background according to claim 1 is characterized in that: Step 2 specifically includes: Calculate several indicators for each set of data, including mean, variance, peak, kurtosis and effective value; Through the MIV algorithm, multiple indicators with high feature importance are extracted.

4. The multi-spectral radiation temperature measurement method based on MEA-LSSVM under high temperature reflection background according to claim 3 is characterized in that: In step 2, the feature matrix is ​​generated as follows: The features with higher feature importance are fused. Assuming the features are Y1, Y2, Y3…, the feature matrix Y is generated by serial fusion: Y=[Y1,Y2…Y n ]n=1,2,3….

5. The multi-spectral radiation temperature measurement method based on MEA-LSSVM under high temperature reflection background according to claim 1 is characterized in that: In step 3, the step of obtaining an optimized least squares support vector machine classifier MEA-LSSVM specifically includes: Step 3.1, input the feature matrix Y into the LSSVM classifier, set the thinking evolution algorithm parameters, generate the initial population, the winning sub-population and the temporary population; Step 3.2: Perform convergence and alienation operations on each of the winning sub-groups and the temporary sub-groups to obtain mature winning sub-groups and mature temporary sub-groups. The convergence and alienation operations are as follows: X(t+1)=X(t)+C*(MX(t))+D*(NX(t)) Among them, X(t) represents the number of individuals at time t, and X(t+1) represents the number of individuals at time t+1; C and D represent the convergence coefficient and alienation coefficient, which determine the degree of convergence and alienation of individuals to group behavior; M and N represent the average behavior level in the group and the degree of difference between individuals, respectively; When C>0, individuals tend to converge with the average behavior level of the group, that is, the number of individuals will move closer to the average level; when C<0, individuals tend to alienate from the average behavior level of the group, that is, the number of individuals will move away from the average level; When D>0, the degree of difference between individuals will increase, that is, the number of individuals will be more dispersed; when D<0, the degree of difference between individuals will decrease, that is, the number of individuals will be more concentrated; Step 3.3: When the number of iterations is met, the optimal individual found is output as the kernel function and penalty function of the LSSVM model to obtain the MEA-LSSVM emission rate model classifier.

6. The multi-spectral radiation temperature measurement method based on MEA-LSSVM under high temperature reflection background according to claim 1 is characterized in that: Step 4 includes: After obtaining the target emissivity model, the multi-spectral radiation temperature measurement equation is constructed according to Planck's blackbody radiation law, and the target equation is solved using the OKBA algorithm to invert the target temperature.

7. The multi-spectral radiation temperature measurement method based on MEA-LSSVM under high temperature reflection background according to claim 1 is characterized in that: The constraints of the OKBA algorithm are: Where, ΔT min is the temperature difference, and F[ε(λ, T)] is the emissivity function.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor runs the computer program stored in the memory, the steps of the method according to any one of claims 1 to 7 are performed.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of computer instructions, and the plurality of computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Metal surface oxide layer temperature measurement method

    CN112633292A

  • Multispectral radiation temperature measurement inversion calculation method based on generalized inverse-neural network, computer and storage medium

    CN113776675A

  • Radiation temperature measuring device for turbine blade of aero-engine and temperature measuring method thereof

    CN115493708A

  • Multi-spectral radiation temperature measurement method and system based on Alpha spectrum-LM algorithm in high-temperature variable environment

    CN116481650A

  • Multispectral radiation temperature measurement inversion calculation method based on ISA algorithm

    CN116625519A