Temperature field uncertainty quantification reconstruction method and system

Abstract

The invention discloses a temperature field uncertainty quantitative reconstruction method and system. The method comprises the following steps: constructing physical field prior distribution of a to-be-processed component under a current working condition; determining corrected temperature field prior distribution based on an introduced constraint equation and the physical field prior distribution; and based on the currently acquired temperature field measurement data and the corrected temperature field prior distribution, reconstructing the temperature field of the to-be-processed component to obtain a reconstructed temperature field and an uncertainty distribution diagram of the reconstructed temperature field. According to the method, high-precision reconstruction of the temperature field and quantitative prediction of uncertainty distribution are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of aero-engine thermodynamic monitoring, and particularly relates to a temperature field uncertainty quantification reconstruction method and system. BACKGROUND

[0002] Turbine blade temperature field monitoring is a key technology for safe operation of an aero-engine. Traditional methods mainly include numerical simulation and experimental measurement, but both have significant limitations. Numerical simulation methods are represented by computational fluid dynamics (CFD), and model simplification errors and boundary condition uncertainties limit their accuracy. For example, when using the Reynolds-averaged Navier-Stokes (RANS) model, the error can be as high as 15%; a 5% deviation in the heat transfer coefficient can result in a 50K temperature error. Experimental measurement methods have low spatial resolution and limited high-temperature applicability.

[0003] In recent years, the introduction of artificial intelligence technology has revolutionized this field. Data-driven methods (such as neural networks) can establish a mapping relationship between input parameters and temperature fields, but generally lack strict uncertainty quantification capabilities, and the lack of physical consistency restricts the improvement of prediction accuracy. Therefore, existing temperature field reconstruction methods have significant shortcomings in uncertainty quantification and prediction accuracy, which severely restricts the temperature field monitoring of complex surfaces represented by turbine blades. SUMMARY

[0004] To address the above shortcomings of the prior art, the present application solves the technical problem of providing a turbine blade temperature field uncertainty quantification reconstruction method and system that combines efficient computation, strict uncertainty quantification, and physical constraints.

[0005] To solve the above technical problems, one technical solution adopted by the present application is to provide a turbine blade temperature field uncertainty quantification reconstruction method, comprising the following steps: Constructing a physical field prior distribution under the current operating condition of the component to be processed; Determining a corrected temperature field prior distribution based on the introduced constraint equation and the physical field prior distribution; Reconstructing the temperature field of the component to be processed based on the currently obtained temperature field measurement data and the corrected temperature field prior distribution, to obtain a reconstructed temperature field and an uncertainty distribution map of the reconstructed temperature field.

[0006] Further, the method for constructing a physical field prior distribution under the current operating condition of the component to be processed includes but is not limited to: Calculating statistical characteristic parameters of the physical field data based on the physical field data, and constructing a physical field prior probability distribution based on the statistical characteristic parameters, wherein the statistical characteristic parameters include a mean field and a covariance matrix; Based on the operating parameters and physical field data under different operating conditions, a probabilistic mapping model is performed to obtain a prediction model to predict the prior distribution of the physical field under the current operating condition.

[0007] Furthermore, the step of probabilistic mapping modeling based on operating parameters and physical field data under different operating conditions to obtain a prediction model for predicting the prior distribution of the physical field under the current operating condition includes the following sub-steps: Dimensionality reduction is performed by using intrinsic orthogonal decomposition to reduce the dimensionality of the physical field data, extracting the dominant modes, and calculating the modal coefficients after dimensionality reduction. A model is constructed using the aforementioned operating condition parameters as input and the reduced modal coefficients as output to build a predictive model for predicting the prior distribution of the physical field. Predict the prior distribution, obtain the current operating parameters, and use the prediction model to predict the prior distribution of the physical field under the current operating conditions.

[0008] Furthermore, the step of determining the corrected prior temperature field distribution based on the introduced constraint equations and the prior physical field distribution specifically includes: if the prior physical field distribution is a prior heat source term, then the heat conduction equation is introduced for correction to obtain the corrected prior temperature field distribution; if the prior physical field distribution is a prior temperature field distribution, then the energy conservation equation is introduced for correction to obtain the corrected prior temperature field distribution.

[0009] Furthermore, when the prior distribution of the physical field is a prior distribution of heat source terms, the heat conduction equation introduced is as follows: (one) In formula (1), Represents the heat conduction operator. Represents the temperature field vector. Represents the vector of heat source terms; When the heat conduction operator When reversible, the prior distribution of the temperature field is as follows: When the heat conduction operator When irreversible, the prior distribution of the temperature field is as follows: ,in, Represents the heat conduction operator The false rebellion, Represents the identity matrix. This represents the regularization hyperparameter. The term represents the mean of the prior distribution of the heat source term. This represents the covariance of the prior distribution of the heat source term.

[0010] Further, in the step of reconstructing the temperature field of the to-be-processed component based on the current acquired temperature field measurement data and the corrected temperature field prior distribution to obtain a reconstructed temperature field and an uncertainty distribution map of the reconstructed temperature field, the step specifically comprises: calculating a temperature field posterior distribution under a Bayesian framework based on the current acquired temperature field measurement data and the corrected temperature field prior distribution; extracting a mean value of the temperature field posterior distribution as a point estimation of the reconstructed temperature field to obtain the reconstructed temperature field; calculating a standard deviation distribution map of temperature values at each position of the reconstructed temperature field based on a covariance matrix of the temperature field posterior distribution to generate an uncertainty quantification distribution map.

[0011] Further, in the step of calculating the temperature field posterior distribution, the step specifically comprises: establishing an observation model according to the temperature field measurement data, introducing an observation operator and a covariance matrix of observation noise; under a Bayesian framework, obtaining a temperature field posterior distribution based on the observation operator, the covariance matrix of observation noise and the temperature field prior distribution a calculation formula of a covariance matrix of the posterior distribution is as follows: (II) In formula (II), denotes the observation operator, denotes the covariance matrix of observation noise, denotes the covariance matrix of the temperature field prior distribution; a calculation formula of a mean value of the posterior distribution is as follows: (III) In formula (III), denotes a mean value of the temperature field prior distribution, denotes the temperature field measurement data.

[0012] Further, when the physical field prior distribution is a heat source term prior distribution, the physical field data is heat source term data, and in the step of dimension reduction processing, the following sub-steps are specifically included: calculating a mean field and a fluctuation component matrix of the heat source term data; calculating eigenvalues of a correlation matrix composed of the fluctuation component matrix; arranging the eigenvalues in descending order, reserving the first m dominant modes according to a cumulative energy contribution rate of the eigenvalues and calculating mode coefficients.

[0013] Further, in the step of calculating the modal coefficient, specifically comprising: based on the eigenvalue and eigencomponent of the correlation matrix and the fluctuation component matrix, calculating a dominant modal; for the heat source fluctuation data of any working condition, calculating its modal coefficient under the dominant modal through projection.

[0014] To solve the above technical problems, another technical solution adopted by the present application is to provide a temperature field uncertainty quantification reconstruction system, comprising: A prior distribution construction module is configured to construct a physical field prior distribution of a component under a current working condition. A correction module is configured to determine a corrected temperature field prior distribution based on the introduced constraint equation and the physical field prior distribution. A temperature field reconstruction module is configured to reconstruct a temperature field of the component based on the currently acquired temperature field measurement data and the corrected temperature field prior distribution, to obtain a reconstructed temperature field and an uncertainty distribution map of the reconstructed temperature field.

[0015] The turbine blade temperature field uncertainty quantification reconstruction method and system have at least the following beneficial effects: the present application effectively improves the accuracy of temperature field reconstruction by constructing a physical field prior distribution and combining sparse measurement data for temperature field reconstruction under a Bayesian framework; the accuracy of the reconstruction result is improved by introducing a constraint equation for correcting the prior distribution; in the entire reconstruction process, not only the point estimate of the temperature field is output, but also the uncertainty of the temperature value at each position is quantified; the intrinsic orthogonal decomposition is used to reduce the dimension of high-dimensional physical field data, and the multi-output Gaussian process regression modeling is combined to significantly reduce the computational complexity and realize fast probabilistic prediction; the present application is suitable for components with complex surfaces such as turbine blades, and provides reliable technical support for the thermodynamic monitoring and safety evaluation of high-temperature components such as aircraft engines. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which are included to provide a further understanding of the application, constitute a part of this application and illustrate embodiments of the present application and its description, which are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings: Figure 1 The flowchart of an embodiment of the temperature field uncertainty quantification reconstruction method of the present application.

[0017] Figure 2 The flowchart of step S100 in the embodiment. Figure 1 The flowchart of step S110 in the embodiment.

[0018] Figure 3 The flowchart of step S110 in the embodiment. Figure 2 The flowchart of step S110 in the embodiment.

[0019] Figure 4 The flowchart of step S110 in the embodiment. Figure 1The flow chart of step S300.

[0020] Figure 5 The system block diagram of an embodiment of the temperature field uncertainty quantification reconstruction system of the present application. DETAILED DESCRIPTION The present application will be further described below with reference to the accompanying drawings.

[0021] The present application proposes a temperature field uncertainty quantification reconstruction method, which realizes high-precision reconstruction of the temperature field and quantitative prediction of the uncertainty distribution of the temperature field by constructing a physical field prior distribution and introducing a constraint equation for correction in combination with a Bayesian framework, and takes into account both the improvement of the reconstruction accuracy of the temperature field and the quantification of the reliability of the reconstruction result.

[0022] The temperature field monitoring of turbine blades is a key technology for the safe operation of an aero-engine, and the present application is completed by the inventors in the work of the temperature field monitoring of turbine blades. It is found after the fact that the present application can also be used for the temperature field uncertainty quantification reconstruction of other components with complex curved surfaces, and of course, is also applicable to the temperature field uncertainty quantification reconstruction of general planar components.

[0023] In order to better illustrate the working principle of the present application, the present embodiment mainly takes turbine blades as an example to illustrate the specific content of the temperature field uncertainty quantification reconstruction method.

[0024] Please refer to Figure 1 The flow chart of an embodiment of the temperature field uncertainty quantification reconstruction method of the present application. The present embodiment includes the following steps: S100, constructing a prior distribution.

[0025] Specifically, the physical field prior distribution under the current working condition of the component to be processed is constructed. The specific method includes but is not limited to: ① calculating statistical characteristic parameters of the physical field data based on the physical field data, and constructing a physical field prior probability distribution based on the statistical characteristic parameters, wherein the statistical characteristic parameters include a mean field and a covariance matrix; ② performing probabilistic mapping modeling based on the working condition parameters and the physical field data under different working conditions to obtain a prediction model to predict the physical field prior distribution under the current working condition.

[0026] The second method will be described in detail below. The present embodiment adopts a multi-output Gaussian process regression (MOGPR) model as the prediction model. Although the Gaussian process regression model has a natural probability inference capability and can quantify the prediction uncertainty through a covariance function, its calculation complexity increases by the third power with the data size, and it is difficult to directly process the high-dimensional temperature field (grid nodes are more than 10 7). While the proper orthogonal decomposition (POD) can significantly reduce the data computation and improve the computational efficiency by extracting the key modes and retaining the main energy features, the present embodiment adopts a method combining the Gaussian process and the proper orthogonal decomposition to realize the construction of the prior distribution of the physical field of the turbine blade.

[0027] Referring to Figure 2 , the step S100 includes the following sub-steps: S110, dimensionality reduction processing.

[0028] Specifically, the proper orthogonal decomposition (POD) is adopted to perform the dimensionality reduction processing on the physical field data, extract the dominant modes, and calculate the mode coefficients after the dimensionality reduction. The physical field data in the high-dimensional space is mapped to the low-dimensional feature space through the dimensionality reduction processing, and the complexity of the subsequent calculation is significantly reduced.

[0029] Before this step, the physical field data and the working condition parameters of the turbine blade (i.e., the component to be processed) under different working conditions are acquired first. The physical field data includes the temperature field data and the heat source term data. The temperature field refers to the continuous distribution state of the temperature values of all points distributed in the entire three-dimensional space volume of the turbine blade. The temperature field data refers to the specific numerical set of the temperature field obtained through measurement or simulation, such as sensor readings or CFD simulation results. The heat source term refers to the heat generated or absorbed per unit time and per unit volume in the heat conduction process of the turbine blade, which describes the influence of the internal heat source of the system on the temperature field. The heat source term data refers to the set of heat source term values discretely distributed on all nodes of the turbine blade under different working conditions, which can be obtained through CFD numerical simulation. In the present embodiment, one working condition is defined by a set of working condition parameters, which are used to describe one complete and stable working state of the turbine blade. The working condition parameters can be selected according to the actual situation, and generally include the inlet total pressure, turbine speed, and other parameters.

[0030] For example, when the constructed prior distribution of the physical field is the prior distribution of the heat source term, in actual operation, the snapshot data set of the heat source term under different working conditions is obtained based on CFD numerical simulation. The snapshot corresponds to one working condition, and refers to the complete set of heat source term data covering all grid nodes of the turbine blade calculated through CFD numerical simulation under a specific working condition. It is assumed that there are M snapshots, each of which contains the heat source term data of N grid nodes, forming a data matrix .

[0031] Referring to Figure 3 , the step S110 includes the following sub-steps: S111, calculating the mean field and the fluctuation component matrix of the heat source term data.

[0032] Specifically, first, the average field of the heat source term data under all working conditions is calculated, which represents the common part or the reference state of the spatial distribution of the heat source term under all working conditions. The average field of the heat source term data is obtained by the following formula:

[0033] wherein, represents the average heat source field matrix, with a dimension of N x 1, and M represents the number of working conditions, each working condition corresponding to a group of heat source term data, represents the heat source term data of the t-th group (N grid node data).

[0034] Then, the fluctuation component of each working condition is obtained by subtracting the average field from the original heat source term data of each working condition, and the fluctuation components of all working conditions constitute a fluctuation component matrix. The fluctuation component matrix is obtained by the following formula:

[0035] wherein, represents the fluctuation component matrix of the global heat source data, with a dimension of N x M, represents the M-th group of fluctuation components, .

[0036] S112, calculate the eigenvalue of the correlation matrix.

[0037] Specifically, first, the correlation matrix is constructed, which is obtained by multiplying the fluctuation component matrix and its transpose matrix. The correlation matrix is obtained by the following formula:

[0038] wherein, C represents the matrix of the mutual relationship of the heat source data under different working conditions, i.e. the correlation matrix, with a dimension of M x M, represents the transpose matrix of the fluctuation component matrix .

[0039] Then, the eigenvalue decomposition is performed on the correlation matrix, and the correlation matrix is decomposed into eigenvalue and eigencomponent. The eigenvalue decomposition of the correlation matrix is performed by the following formula:

[0040] wherein, represents the eigencomponent of the correlation matrix, i = 1, 2,..., N, represents the eigenvalue of the correlation matrix, and the eigenvalue represents the energy of the corresponding POD mode, i.e. its importance.

[0041] S113, retain the dominant mode and calculate the modal coefficient.

[0042] Specifically, in descending order of eigenvalues, the first m dominant modes are retained according to the cumulative energy contribution rate of the eigenvalues, and the modal coefficients are calculated. The first m modes are referred to as dominant modes. This step determines the number of dominant modes to be retained according to the eigenvalues, and calculates the projection coefficients of the heat source term data on these dominant modes, completing the dimension reduction. The judgment condition for retaining the first m dominant modes is that the cumulative energy contribution rate exceeds the set energy retention ratio threshold. In this embodiment, the set energy retention ratio threshold is 99.9%. The cumulative energy contribution rate is obtained by the following formula:

[0043] wherein, indicates the cumulative energy contribution rate. By setting a high threshold, such as 99.9%, it can be ensured that the discarded modes only contain insignificant noise or secondary information, thereby achieving optimal dimension reduction.

[0044] For the heat source fluctuation data of any working condition, its representation in the low-dimensional space, i.e., the modal coefficient, can be calculated by projecting it onto the retained dominant modes. The modal coefficient is subsequently used as the output target of the multi-output Gaussian process regression model to establish a probabilistic mapping relationship from the working condition parameters to the heat source term distribution. Specifically, the method for calculating the modal coefficient includes: calculating the dominant modes based on the eigenvalues and eigenvectors of the correlation matrix and the fluctuation component matrix; and calculating the modal coefficient of the heat source fluctuation data of any working condition by projection. The dominant modes are obtained by the following formula:

[0045] wherein, indicates the i-th dominant mode, i = 1, 2, …, m.

[0046] The modal coefficient is obtained by the following formula:

[0047] wherein, indicates the heat source term fluctuation component of the t-th working condition, indicates the i-th modal coefficient of the heat source term data of the t-th working condition.

[0048] By retaining the first m dominant modes, the POD reconstruction formula can be used to approximate the heat source term distribution under the current working condition, realizing the reconstruction from the low-dimensional modal space to the high-dimensional physical field. The specific formula is as follows:

[0049] wherein, indicates the reconstructed heat source field matrix, indicates the average heat source field matrix, and m indicates the number of retained dominant modes.

[0050] S120, constructing a prediction model.

[0051] Specifically, a prediction model for predicting the prior distribution of the physical field is constructed with the working condition parameters as input and the reduced modal coefficients as output. In this embodiment, the prediction model is a MOGPR model, which is used to predict the prior distribution of the heat source term under new working condition parameters, and capture the nonlinear probabilistic mapping relationship between the working condition parameters and the reduced heat source modal coefficients. The MOGPR model can handle multiple related output tasks, and its essence is to define a Gaussian process prior for each modal coefficient and represent the correlation between different outputs through a covariance function. Since the training of the model is a mature technology, it will not be described here.

[0052] S130, predicting the prior distribution.

[0053] Specifically, the current working condition parameters are obtained, and the prediction model is used to predict the prior distribution of the physical field under the current working condition. For a new set of working condition parameters , the trained MOGPR model can predict it to obtain the probability distribution of all modal coefficients, which is a Gaussian distribution. Further, through the POD reconstruction formula, the probability distribution of the modal coefficient can be converted into the prior distribution of the heat source term. The prior distribution of the heat source term is also a Gaussian distribution, and its mean and covariance are determined by the prediction result of the MOGPR model and the POD modal. For new working condition parameters , the corresponding heat source term distribution is predicted, which completely quantifies the uncertainty of the predicted value of the heat source term under the new working condition, wherein represents the mean of the prior distribution of the heat source term, represents the covariance of the prior distribution of the heat source term.

[0054] S200, correcting the prior distribution.

[0055] Specifically, the corrected temperature field prior distribution is determined based on the introduced constraint equation and the prior distribution of the physical field. For example, if the prior distribution of the physical field is the prior distribution of the heat source term, the heat conduction equation is introduced for correction to obtain the corrected temperature field prior distribution; if the prior distribution of the physical field is the prior distribution of the temperature field, the energy conservation equation is introduced for correction to obtain the corrected temperature field prior distribution. In this embodiment, the constructed prior distribution of the physical field is the prior distribution of the heat source term, and the heat conduction equation is introduced for correction to ensure that the temperature field corresponding to the heat source term obtained by data-driven probabilistic prediction strictly obeys the physical law, and the consistency of the prediction result in the physical sense is constrained, thereby improving the prediction accuracy. The introduced heat conduction equation is as follows:

[0056] in, Represents the heat conduction operator. Represents the temperature field vector. Represents the vector of heat source terms; Because the prior distribution of the heat source term is modeled as a Gaussian distribution, and the physical constraint equations are linear, the prior distribution of the temperature field u is also a Gaussian distribution. When the heat conduction operator... When reversible, the prior distribution of the temperature field is as follows: When the heat conduction operator When irreversible, the prior distribution of the temperature field is as follows: ,in, Represents the heat conduction operator The false rebellion, This represents the regularization hyperparameter. Represents the identity matrix. The value represents the mean of the prior distribution of the heat source term. This represents the covariance of the prior distribution of the heat source term.

[0057] S300, reconstruct the temperature field and obtain the uncertainty distribution map.

[0058] Specifically, based on the currently acquired temperature field measurement data and the corrected prior temperature field distribution, the temperature field of the component to be processed is reconstructed to obtain the reconstructed temperature field and its uncertainty distribution map. Please refer to [link to relevant documentation]. Figure 4 This step 300 includes the following sub-steps: S310, Calculate the posterior distribution of the temperature field.

[0059] Specifically, the posterior distribution of the temperature field is calculated within a Bayesian framework based on the currently acquired temperature field measurement data and the corrected prior distribution of the temperature field.

[0060] First, measurement data is acquired using n temperature sensors arranged on the surface of the turbine blades. An observation model is established based on temperature field measurement data, incorporating observation operators and the covariance matrix of observation noise. The observation equations are as follows:

[0061] in, This represents the observation operator, whose elements are determined by the position of the sensor on the turbine blade. The observation noise is represented by a Gaussian distribution. .

[0062] Secondly, within the Bayesian framework, the posterior distribution of the temperature field is obtained based on the observation equation and the prior distribution of the temperature field. The covariance matrix of the posterior distribution The calculation formula of is as follows:

[0063] wherein, represents an observation operator, represents a covariance matrix of observation noise, represents a covariance of a temperature field prior distribution; a mean value of a posterior distribution The calculation formula of is as follows:

[0064] wherein, represents a mean value of a temperature field prior distribution, represents temperature field measurement data.

[0065] S320, reconstructing a temperature field.

[0066] Specifically, the mean value of the temperature field posterior distribution is extracted as a reconstructed temperature field point estimate, and a reconstructed temperature field is obtained. Because the posterior distribution is a Gaussian distribution, the mean value of the posterior distribution is a maximum a posteriori probability estimate, which represents the most possible and optimal estimate result of the temperature field under the condition of given prior knowledge and sparse measurement data, that is, the reconstructed temperature field point estimate.

[0067] S330, generating an uncertainty quantification distribution map.

[0068] Specifically, based on the covariance matrix of the temperature field posterior distribution, a standard deviation distribution map of the temperature value of each position of the reconstructed temperature field is calculated, and an uncertainty quantification distribution map is generated. The specific calculation process is a mature technology, and will not be described here.

[0069] Referring to Figure 5 is a system block diagram of an embodiment of a temperature field uncertainty quantification reconstruction system of the present application. The temperature field uncertainty quantification reconstruction system of the embodiment is used to implement the temperature field uncertainty quantification reconstruction method as described in the above embodiment. Specifically, the temperature field uncertainty quantification reconstruction system of the embodiment includes a prior distribution construction module 100, a correction module 200, and a temperature field reconstruction module 300. Wherein: The prior distribution construction module 100 is used to construct a physical field prior distribution under a current working condition of a component to be processed. In the embodiment, the prior distribution construction module 100 is used to perform probabilistic mapping modeling based on working condition parameters and physical field data under different working conditions, to obtain a prediction model to predict the physical field prior distribution under the current working condition. The physical field prior distribution can be a heat source term prior distribution or a temperature field prior distribution.

[0070] The correction module 200 is used to determine a corrected temperature field prior distribution based on the introduced constraint equation and the physical field prior distribution.

[0071] The temperature field reconstruction module 300 is used to reconstruct the temperature field of the component to be processed based on the current acquired temperature field measurement data and the corrected temperature field prior distribution, to obtain a reconstructed temperature field and an uncertainty distribution map of the reconstructed temperature field.

[0072] The present application effectively improves the accuracy of temperature field reconstruction by constructing a physical field prior distribution and combining sparse measurement data to reconstruct the temperature field in a Bayesian framework; the accuracy of the reconstruction result is improved by introducing constraint equations for correction and modifying the prior distribution; in the entire reconstruction process, not only the point estimate of the temperature field is output, but also the uncertainty of the temperature value at each position can be quantified; the intrinsic orthogonal decomposition is used to reduce the dimension of high-dimensional physical field data, and the multi-output Gaussian process regression modeling is combined to significantly reduce the computational complexity and realize fast probabilistic prediction; the present application is suitable for turbine blades and other components with complex surfaces, and provides reliable technical support for the thermodynamic monitoring and safety evaluation of high-temperature components such as aircraft engines.

[0073] The above only expresses the preferred embodiments of the present application, which are described in detail, but cannot be understood as limiting the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of the patent protection of the present application should be subject to the appended claims.

Claims

1. A temperature field uncertainty quantification reconstruction method, characterized in that, The method comprises the following steps: constructing a physical field prior distribution of a component to be processed under a current working condition; determining a corrected temperature field prior distribution based on the introduced constraint equation and the physical field prior distribution; reconstructing a temperature field of the component to be processed based on the current acquired temperature field measurement data and the corrected temperature field prior distribution, to obtain a reconstructed temperature field and an uncertainty distribution map of the reconstructed temperature field.

2. The temperature field uncertainty quantification reconstruction method of claim 1, wherein, The method for constructing a physical field prior distribution of a component to be processed under a current working condition comprises but is not limited to: calculating statistical characteristic parameters of the physical field data based on the physical field data, and constructing a physical field prior probability distribution based on the statistical characteristic parameters, wherein the statistical characteristic parameters comprise a mean field and a covariance matrix; performing probabilistic mapping modeling based on working condition parameters and physical field data under different working conditions, to obtain a prediction model for predicting a physical field prior distribution under a current working condition.

3. The temperature field uncertainty quantification reconstruction method of claim 2, wherein, In the step of performing probabilistic mapping modeling based on working condition parameters and physical field data under different working conditions, to obtain a prediction model for predicting a physical field prior distribution under a current working condition, the following sub-steps are included: dimension reduction processing, wherein intrinsic orthogonal decomposition is adopted to perform dimension reduction processing on the physical field data, to extract a dominant mode and calculate a dimension-reduced modal coefficient; constructing a model, wherein the working condition parameters are taken as inputs, and the dimension-reduced modal coefficient is taken as an output, to construct a prediction model for predicting a physical field prior distribution; predicting a prior distribution, wherein the current working condition parameters are acquired, and the prediction model is adopted to predict a physical field prior distribution under a current working condition.

4. The temperature field uncertainty quantification reconstruction method of claim 1, wherein, In the step of determining a corrected temperature field prior distribution based on the introduced constraint equation and the physical field prior distribution, the following is specifically included: if the physical field prior distribution is a heat source term prior distribution, a heat conduction equation is introduced for correction, to obtain a corrected temperature field prior distribution; if the physical field prior distribution is a temperature field prior distribution, an energy conservation equation is introduced for correction, to obtain a corrected temperature field prior distribution.

5. The temperature field uncertainty quantification reconstruction method of claim 4, wherein, When the physical field prior distribution is a heat source term prior distribution, the introduced heat conduction equation is as follows: (I) In formula (I), denotes the heat conduction operator, denotes the temperature field vector, denotes the heat source term vector; when the heat conduction operator is reversible, the temperature field prior distribution is ; when the heat conduction operator is irreversible, the temperature field prior distribution is where, denotes the pseudo-inverse of the heat conduction operator , denotes the identity matrix, denotes the regularization hyperparameter, denotes the mean of the heat source term prior distribution, and the denotes the covariance of the heat source term prior distribution.

6. The temperature field uncertainty quantification reconstruction method of claim 1, wherein, In the step of reconstructing a temperature field of the component to be processed based on the current acquired temperature field measurement data and the corrected temperature field prior distribution, to obtain a reconstructed temperature field and an uncertainty distribution map of the reconstructed temperature field, the following is specifically included: calculating a temperature field posterior distribution under a Bayesian framework based on the current acquired temperature field measurement data and the corrected temperature field prior distribution; extracting a mean value of the temperature field posterior distribution as a point estimation of the reconstructed temperature field, to obtain a reconstructed temperature field; calculating a standard deviation distribution map of temperature values at each position of the reconstructed temperature field based on a covariance matrix of the temperature field posterior distribution, to generate an uncertainty quantification distribution map.

7. The temperature field uncertainty quantification reconstruction method of claim 6, wherein, In the step of calculating a temperature field posterior distribution, the following is specifically included: establishing an observation model according to temperature field measurement data, and introducing an observation operator and a covariance matrix of observation noise; In the Bayesian framework, the posterior distribution of the temperature field is obtained based on the observation operator, the covariance matrix of the observation noise and the prior distribution of the temperature field The formula for calculating the covariance matrix of the posterior distribution is as follows ​ (two) In formula (two), represents an observation operator, represents a covariance matrix of observation noise, represents a covariance matrix of the temperature field prior distribution; The mean of the posterior distribution The formula for calculating the mean of the posterior distribution is as follows: (three) In formula (three), denotes the mean of the temperature field prior distribution, denotes the temperature field measurement data.

8. The temperature field uncertainty quantification reconstruction method of claim 3, wherein, when the physical field prior distribution is a heat source term prior distribution, the physical field data is heat source term data, and in the step of dimension reduction processing, the following sub-steps are specifically included: calculating a mean field and a fluctuation component matrix of the heat source term data; calculating eigenvalues of a correlation matrix composed of the fluctuation component matrix; According to descending order of eigenvalues, the first m dominant modes are reserved according to cumulative energy contribution rate of the eigenvalues, and the modal coefficients are calculated.

9. The temperature field uncertainty quantification reconstruction method of claim 8, wherein, In the step of calculating the modal coefficients, the dominant modes are calculated based on the eigenvalues and eigenvectors of the correlation matrix and the fluctuation component matrix; and for the heat source fluctuation data of any working condition, the modal coefficients of the heat source fluctuation data in the dominant modes are calculated through projection.

10. A temperature field uncertainty quantification reconstruction system, characterized in that, The method comprises the steps of: a prior distribution construction module, used for constructing a physical field prior distribution of a component under a current working condition; a correction module, used for determining a corrected temperature field prior distribution based on an introduced constraint equation and the physical field prior distribution; a temperature field reconstruction module, used for reconstructing a temperature field of the component based on currently acquired temperature field measurement data and the corrected temperature field prior distribution, to obtain a reconstructed temperature field and an uncertainty distribution map of the reconstructed temperature field.

Images