Sample construction method and system for uncertainty analysis and storage medium
By constructing the Latin hypercube sampling matrix and mapping it to a multivariate Gaussian distribution, combining Shannon diversity indicators and the minimum sample distance, the problem of uneven distribution of input parameters is solved, and the accuracy of uncertainty analysis and the reliability of the proxy model are improved.
Patent Information
- Application Number
- CN202410989034.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-12
AI Technical Summary
In uncertainty analysis, the distribution of input parameters is difficult to accurately determine, especially affected by incomplete data and subjectivity of expert opinions, resulting in uneven distribution of parameters, affecting the reliability and accuracy of the proxy model.
By constructing the Latin hypercube sampling matrix, map the sample data to a multivariate Gaussian distribution based on the average value matrix and covariance matrix of the measured data, calculate the Shannon diversity indicator and the minimum distance of the sample, and determine the input samples of the proxy model to ensure the input parameters Even distribution and diversity.
The uniformity of sampling and the diversity of sample data are improved, the distribution of input parameters is accurately determined, the input samples of the proxy model are representative and diverse, and the accuracy of uncertainty analysis is improved.
Smart Images

Figure CN118940506A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of nuclear reactor simulation, and in particular to a sample construction method, system and storage medium for uncertainty analysis. Background Art
[0002] Numerical reactor multi-physics coupling simulation is an important research direction in the field of nuclear energy, which involves the coupling solution of multiple physical processes such as nuclear reaction, heat conduction, fluid mechanics, etc. Multi-physics coupling simulation usually requires a large number of input parameters (input parameters include material properties, geometry, boundary conditions, etc.) to build a proxy model, so as to use the proxy model to simulate multiple physical fields of the reactor.
[0003] Uncertainty analysis is an important task in numerical reactor research. Uncertainty analysis aims to evaluate the uncertainty of the simulation results of the proxy model in order to determine the reliability and accuracy of the proxy model. Currently, when performing uncertainty analysis, multiple samples are generally obtained from the distribution of input parameters through random sampling techniques, and these samples are input into the proxy model as input to obtain the corresponding output results, and then uncertainty analysis is performed based on these output results to evaluate the uncertainty of the proxy model. However, in practical applications, the distribution of input parameters may be affected by factors such as incomplete data, subjectivity or inconsistency of expert opinions, and it is often difficult to accurately determine the distribution of input parameters, and sometimes the input parameters are unevenly distributed. Summary of the invention
[0004] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, the present application proposes a sample construction method, system and storage medium for uncertainty analysis, which can accurately determine the distribution of input parameters while making the input parameters uniformly distributed.
[0005] In order to solve the above technical problems, the present invention proposes the following technical solutions:
[0006] The first aspect of the present application provides a sample construction method for uncertainty analysis, including:
[0007] Determining, based on measurement data of various input parameters in a numerical reactor, a mean value matrix and a covariance matrix corresponding to the measurement data, the input parameters being associated with a proxy model of the numerical reactor;
[0008] Constructing a Latin hypercube sampling matrix, and mapping the Latin hypercube sampling matrix to a multivariate Gaussian distribution based on the mean value matrix and the covariance matrix to obtain multivariate Gaussian sampling data;
[0009] Determine the Shannon diversity index data corresponding to the multivariate Gaussian sampling data, and determine the minimum sample distance corresponding to the multivariate Gaussian sampling data;
[0010] The input samples of the proxy model are determined based on the Shannon diversity index data and the sample minimum distance.
[0011] According to the sample construction method of the uncertainty analysis of the embodiment of the first aspect of the present application, there are at least the following beneficial effects: the present application performs uniform sampling by constructing a Latin hypercube sampling matrix so that the sample data is evenly distributed in each dimension, thereby improving the uniformity of sampling. Moreover, after obtaining the Latin hypercube sampling matrix, the sample data in the Latin hypercube sampling matrix will be mapped to a multivariate Gaussian distribution based on the mean value matrix and covariance matrix corresponding to the measured data to obtain multivariate Gaussian sampling data, which can retain the mean characteristics and covariance characteristics of the measured data and accurately determine the distribution of the input parameters. Finally, the Shannon diversity index data and the sample minimum distance corresponding to the multivariate Gaussian sampling data will also be calculated to evaluate the diversity and representativeness of the multivariate Gaussian sampling data through the Shannon diversity index data, and to evaluate the correlation of the multivariate Gaussian sampling data through the sample minimum distance. The present application determines the input samples of the proxy model based on the Shannon diversity index data and the sample minimum distance, thereby ensuring the diversity and representativeness of the input samples of the proxy model.
[0012] According to some embodiments of the first aspect of the present application, the step of determining the input sample of the proxy model based on the Shannon diversity index data and the sample minimum distance includes:
[0013] Get a first weight and a second weight;
[0014] Based on the first weight and the second weight, calculating a weighted sum of the Shannon diversity index data and the sample minimum distance as a feature score of the multivariate Gaussian sampling data;
[0015] Determine the input samples of the proxy model based on the feature scores of the multivariate Gaussian sampling data.
[0016] According to some embodiments of the first aspect of the present application, the determining the input sample of the proxy model based on the feature score of the multivariate Gaussian sampling data includes:
[0017] If the feature score of the multivariate Gaussian sampling data reaches a first threshold, determining the multivariate Gaussian sampling data as an input sample of the proxy model;
[0018] The method further comprises:
[0019] If the characteristic score of the multivariate Gaussian sampling data is less than the first threshold, return to the step of constructing the Latin hypercube sampling matrix until the characteristic score of the multivariate Gaussian sampling data reaches the first threshold.
[0020] According to some embodiments of the first aspect of the present application, the determining the input sample of the proxy model based on the feature score of the multivariate Gaussian sampling data includes:
[0021] Get historical sampling data;
[0022] If the feature score of the multivariate Gaussian sampling data reaches a second threshold, adding the sample data in the multivariate Gaussian sampling data to the historical sampling data to obtain expanded sampling data;
[0023] Determining a feature score of the expanded sampling data;
[0024] An input sample of the proxy model is determined based on the feature scores of the augmented sampling data.
[0025] According to some embodiments of the first aspect of the present application, determining the input sample of the proxy model based on the feature score of the expanded sampling data includes:
[0026] If the feature score of the expanded sampling data reaches a third threshold, using the expanded sampling data as updated historical sampling data;
[0027] Returning to the step of constructing the Latin hypercube sampling matrix, until the number of sample data in the updated historical sampling data meets the second threshold, the updated historical sampling data is determined as the input sample of the proxy model;
[0028] The method further comprises:
[0029] If the feature score of the expanded sampling data is less than the third threshold, return to the step of constructing the Latin hypercube sampling matrix until the number of sample data in the updated historical sampling data meets the second threshold, and determine the updated historical sampling data as the input sample of the proxy model.
[0030] According to some embodiments of the first aspect of the present application, constructing a Latin hypercube sampling matrix includes:
[0031] Determine the number of columns of the Latin hypercube sampling matrix based on the number of types of input parameters, and determine the number of rows of the Latin hypercube sampling matrix based on the number of measurement data corresponding to each input parameter;
[0032] For multiple sample data in each column of the Latin hypercube sampling matrix, equally divide the preset first value range based on the number of measurement data corresponding to each input parameter, determine the value range corresponding to each sample data in the Latin hypercube sampling matrix, and determine the value of each sample data based on the value range;
[0033] A Latin hypercube sampling matrix is constructed based on the number of rows of the Latin hypercube sampling matrix, the number of columns of the Latin hypercube sampling matrix and the value of sample data in each column of the Latin hypercube sampling matrix.
[0034] According to some embodiments of the first aspect of the present application, determining the Shannon diversity index data corresponding to the multivariate Gaussian sampling data includes:
[0035] Performing cluster analysis on the multivariate Gaussian sampling data based on a preset clustering algorithm to obtain the number of clusters and the number of samples in each cluster;
[0036] Shannon diversity index data is determined based on the number of clusters, the number of samples in each cluster, and the total number of samples included in the multivariate Gaussian sampling data.
[0037] According to some embodiments of the first aspect of the present application, determining the minimum sample distance corresponding to the multivariate Gaussian sampling data includes:
[0038] For any two sample data in the multivariate Gaussian sampling data, the distance between the two sample data is calculated based on a preset distance measurement algorithm;
[0039] The minimum value of the distance between any two sample data in the multivariate Gaussian sampling data is determined as the sample minimum distance.
[0040] The second aspect of the present application provides a sample construction system for uncertainty analysis, including:
[0041] at least one memory;
[0042] at least one processor;
[0043] at least one program;
[0044] The program is stored in the memory, and the processor executes at least one of the programs to implement:
[0045] A sample construction method for uncertainty analysis as described in any one of the first aspects of the present application.
[0046] A third aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer-executable signal, and the computer-executable signal is used to execute:
[0047] A sample construction method for uncertainty analysis as described in any one of the first aspects of the present application.
[0048] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] The accompanying drawings are used to provide further understanding of the technical solution of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the technical solution of the present disclosure and do not constitute a limitation on the technical solution of the present disclosure.
[0050] Figure 1 is a system architecture diagram of a system to which a sample construction method for uncertainty analysis according to an embodiment of the present disclosure is applied;
[0051] Figure 2 is a main flow chart of a sample construction method for uncertainty analysis according to an embodiment of the present disclosure;
[0052] Figure 3 is a flow chart of constructing a Latin hypercube sampling matrix according to an embodiment of the present disclosure;
[0053] Figure 4 is a flow chart of determining Shannon diversity index data corresponding to multivariate Gaussian sampling data according to an embodiment of the present disclosure;
[0054] Figure 5 is a flow chart for determining the minimum distance of samples corresponding to multivariate Gaussian sampling data according to an embodiment of the present disclosure;
[0055] Figure 6 is a flowchart of determining input samples of a proxy model based on Shannon diversity index data and sample minimum distance according to an embodiment of the present disclosure;
[0056] Figure 7 is a flow chart of determining input samples of a proxy model based on feature scores of multivariate Gaussian sampling data according to an embodiment of the present disclosure;
[0057] Figure 8 4 is a block diagram of a sample construction system for uncertainty analysis according to an embodiment of the present disclosure. DETAILED DESCRIPTION
[0058] In order to make the purpose, technical solution and advantages of the present disclosure more clear, the present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present disclosure and are not used to limit the present disclosure.
[0059] Before further describing the embodiments of the present disclosure in detail, the nouns and terms involved in the embodiments of the present disclosure are described. The nouns and terms involved in the embodiments of the present disclosure are subject to the following interpretations:
[0060] Uncertainty analysis: Uncertainty analysis is an estimation and study of the changes and impacts of various external factors that cannot be controlled in advance during production and operation. Uncertain factors in economic development are ubiquitous, such as whether the investment exceeds the limit, whether the construction period is delayed, whether the price of raw materials rises, whether the production capacity can meet the design requirements, etc. In order to make correct decisions, it is necessary to conduct a comprehensive technical and economic evaluation, calculate the probability of occurrence of each factor and its impact on the decision-making plan, and select the best plan. Its basic analysis methods include: profit and loss analysis, sensitivity analysis, and probability analysis. The main calculations are the profit and loss value, regret value, and expected value of the plan.
[0061] Numerical reactor: Numerical reactor (also known as virtual reactor, or numerical reactor for short) is a computer simulation software used for CAD modeling, operation simulation, material performance analysis, etc. of real nuclear reactors. Numerical reactors can be used for reactor core physics analysis, fuel design and performance optimization, extreme accident analysis and prediction, etc. The research scope covers the entire process of nuclear fuel prospecting, purification, fuel preparation, nuclear power plant operation, secondary refining of nuclear waste, and burial of high-level radioactive waste. Numerical reactors require the establishment of a precise three-dimensional simulation platform to form a complete high-resolution numerical simulation software system.
[0062] Numerical reactor multi-physics coupling simulation is an important research direction in the field of nuclear energy, which involves the coupling solution of multiple physical processes such as nuclear reaction, heat conduction, fluid mechanics, etc. Multi-physics coupling simulation usually requires a large number of input parameters (input parameters include material properties, geometry, boundary conditions, etc.) to build a proxy model, so as to use the proxy model to simulate multiple physical fields of the reactor.
[0063] Uncertainty analysis is an important task in numerical reactor research. Uncertainty analysis aims to evaluate the uncertainty of the simulation results of the proxy model in order to determine the reliability and accuracy of the proxy model. Currently, when performing uncertainty analysis, multiple samples are generally obtained from the distribution of input parameters through random sampling techniques, and these samples are input into the proxy model as input to obtain the corresponding output results, and then uncertainty analysis is performed based on these output results to evaluate the uncertainty of the proxy model. However, in practical applications, the distribution of input parameters may be affected by factors such as incomplete data, subjectivity or inconsistency of expert opinions, and it is often difficult to accurately determine the distribution of input parameters, and sometimes the input parameters are unevenly distributed.
[0064] Based on this, the embodiments of the present disclosure provide a sample construction method, system and storage medium for uncertainty analysis. It can perform uniform sampling by constructing a Latin hypercube sampling matrix so that the sample data is evenly distributed in each dimension, thereby improving the uniformity of sampling. Moreover, after determining the Latin hypercube sampling matrix, the present application also maps the sample data in the Latin hypercube sampling matrix to a multivariate Gaussian distribution to obtain multivariate Gaussian sampling data, which can retain the mean characteristics and covariance characteristics of the measurement data, accurately determine the distribution of the input parameters, and thereby achieve uniform distribution of the input parameters while accurately determining the distribution of the input parameters.
[0065] System architecture and scenario description of the application of the embodiments of the present disclosure
[0066] Figure 1 1 is a system architecture diagram used in the sample construction method of uncertainty analysis according to the embodiment of the present disclosure, which includes a terminal 140, an Internet 130, a gateway 120, a server 110, and the like.
[0067] The terminal 140 includes various forms such as desktop computers, laptop computers, PDAs (personal digital assistants), mobile phones, vehicle-mounted terminals, home theater terminals, and dedicated terminals. In addition, it can be a single device or a collection of multiple devices. The terminal 140 can communicate with the Internet 130 in a wired or wireless manner to exchange data. Moreover, the sample construction method of the uncertainty analysis of the embodiment of the present disclosure can be implemented in the terminal 140, and the terminal 140 can show the calculation process of the sample construction method of the uncertainty analysis to the user.
[0068] The server 110 refers to a computer system that can provide certain services to the terminal 140. The sample construction method of the uncertainty analysis of the embodiment of the present disclosure can also be implemented on the server 110. Compared with the ordinary terminal 140, the server 110 has higher requirements in terms of stability, security, performance, etc. The server 110 can be a high-performance computer in a network platform, a cluster of multiple high-performance computers, a part of a high-performance computer (such as a virtual machine), a combination of parts of multiple high-performance computers (such as virtual machines), etc.
[0069] The gateway 120 is also called an internetwork connector or a protocol converter. The gateway realizes network interconnection at the transport layer and is a computer system or device that acts as a converter. The gateway is a translator between two systems that use different communication protocols, data formats or languages, or even completely different architectures. At the same time, the gateway can also provide filtering and security functions. The message sent by the terminal 140 to the server 110 must be sent to the corresponding server 110 through the gateway 120. The message sent by the server 110 to the terminal 140 must also be sent to the corresponding terminal 140 through the gateway 120.
[0070] The embodiments of the present disclosure can be applied in fields such as uncertainty analysis of proxy models. Here, the application scenario of building a numerical reactor is used as an example for explanation: Assuming that the embodiments of the present disclosure are implemented on the terminal 140, the user first needs to confirm the input parameters of the numerical reactor, where the input parameters can be material properties, geometric shapes, boundary conditions, etc. Then, the measurement data of each input parameter is collected and input into the terminal 140, so that the terminal 140 executes the sample construction method of the uncertainty analysis of the embodiment of the present disclosure based on the measurement data of each input parameter input by the user, and determines the input sample of the proxy model of the numerical reactor. Furthermore, after determining the input sample of the proxy model of the numerical reactor, the terminal 140 can also input the determined input sample into the constructed proxy model of the numerical reactor to perform uncertainty analysis on the proxy model of the numerical reactor.
[0071] It should be understood that the above content only illustrates some application scenarios of the present disclosure. The business scenarios to which the present disclosure can be applied may include but are not limited to the specific embodiments listed above.
[0072] General description of the disclosed embodiments
[0073] It should be emphasized that the embodiments of the present disclosure can be applied to a variety of application scenarios, such as uncertainty analysis of proxy models. Uncertainty analysis is an important task in numerical reactor research. Uncertainty analysis aims to evaluate the uncertainty of the simulation results of the proxy model to determine the reliability and accuracy of the proxy model. At present, when performing uncertainty analysis, multiple samples are generally obtained from the distribution of input parameters through random sampling technology, and these samples are input into the proxy model as input to obtain corresponding output results, so as to perform uncertainty analysis based on these output results to evaluate the uncertainty of the proxy model. However, in practical applications, the distribution of input parameters may be affected by factors such as incomplete data, subjectivity or inconsistency of expert opinions, and it is often difficult to accurately determine the distribution of input parameters, and sometimes there will be uneven distribution of input parameters. Some embodiments of the present disclosure provide a sample construction method, system and storage medium for uncertainty analysis, which can perform uniform sampling by constructing a Latin hypercube sampling matrix so that the sample data is evenly distributed in each dimension, thereby improving the uniformity of sampling. Moreover, after determining the Latin hypercube sampling matrix, the present application also maps the sample data in the Latin hypercube sampling matrix to a multivariate Gaussian distribution to obtain multivariate Gaussian sampling data. The multivariate Gaussian sampling data can retain the mean characteristics and covariance characteristics of the measured data and accurately determine the distribution of the input parameters, thereby achieving a uniform distribution of the input parameters while accurately determining the distribution of the input parameters.
[0074] The sample construction method for uncertainty analysis in the embodiment of the present disclosure may be executed on the terminal 140 or on the server 110 .
[0075] like Figure 2 As shown, according to one embodiment of the present disclosure, a sample construction method for uncertainty analysis includes:
[0076] Step 210, based on the measurement data of each input parameter in the numerical reactor, determining the mean value matrix and the covariance matrix corresponding to the measurement data, and associating the input parameters with the proxy model of the numerical reactor;
[0077] Step 220: construct a Latin hypercube sampling matrix, and map the Latin hypercube sampling matrix to a multivariate Gaussian distribution based on the mean value matrix and the covariance matrix to obtain multivariate Gaussian sampling data;
[0078] Step 230, determining the Shannon diversity index data corresponding to the multivariate Gaussian sampling data, and determining the minimum sample distance corresponding to the multivariate Gaussian sampling data;
[0079] Step 240: Determine the input samples of the proxy model based on the Shannon diversity index data and the minimum distance of the samples.
[0080] The above steps 210 - 240 are briefly described below.
[0081] In step 210, based on the measurement data of each input parameter in the numerical reactor, a mean value matrix and a covariance matrix corresponding to the measurement data are determined, and the input parameters are associated with a proxy model of the numerical reactor.
[0082] According to an embodiment of the present disclosure, input parameters refer to parameters of the proxy model of the input numerical reactor, and the input parameters include the material properties, geometric shape, state variables detected by the temperature detector, state variables detected by the pressure detector, and boundary conditions of the nuclear reactor.
[0083] It should be noted that for the numerical reactor multi-physics coupling analysis problem, the general mathematical expression is as follows:
[0084] Y = F(X,ε);
[0085] Among them, X represents the variable parameter vector input by the user; ε represents some variable parameter variables of the proxy model; F represents the proxy model (i.e., the theoretical model of the numerical reactor, such as a set of differential equations); Y represents the output of the proxy model, i.e., the simulation results corresponding to the multi-physical field.
[0086] Based on this, when performing uncertainty analysis on the proxy model of the numerical reactor, it is necessary to obtain the uncertainty distribution of the input parameters (X, ε) in order to use these input parameters to infer the uncertainty of the output results of the proxy model. In order to obtain the uncertainty distribution of the input parameters (X, ε), this application first needs to collect the measurement data of the input parameters. Taking the input parameter x as an example, it is necessary to collect the measurement data related to the input parameter x. The measurement data related to the input parameter x can be expressed as x1, x2..., x n Furthermore, since the proxy model has multiple input parameters, the present application also needs to obtain measurement data corresponding to different input parameters.
[0087] According to an embodiment of the present disclosure, the average value matrix corresponding to the measurement data is a matrix composed of the average values of the measurement data corresponding to different input parameters; the covariance matrix corresponding to the measurement data is a matrix composed of the covariances of the measurement data of different input parameters.
[0088] It can be understood that, after determining the measured data x1, x2, . . . , x n Afterwards, the measured data x1, x2, ..., x of each input parameter x can be used to calculate the nThe data matrix X is constructed, wherein each column of the data matrix X corresponds to a type of input parameter, and each row of the data matrix X corresponds to a measurement data corresponding to an input parameter. Assuming that the dimension of the data matrix X is m×n, it can be determined that the number of types of input parameters in the data matrix X is n, and the number of measurement data corresponding to each input parameter in the data matrix X is m.
[0089] Further, after determining to obtain the data matrix X, the measured data of each input parameter in the data matrix X can be averaged to determine to obtain the average matrix M, and the dimension of the average matrix M is 1×n. Then, the covariance of the measured data of various input parameters in the data matrix X can also be calculated to determine to obtain the covariance matrix C, and the dimension of the covariance matrix C is n×n. When the numpy open source database is used to execute the sample construction method of the uncertainty analysis of the present application, the covariance of the measured data of each input parameter in the data matrix X can also be calculated by the function numpy.cov(X) to obtain the covariance matrix C.
[0090] According to another embodiment of the present disclosure, for unmeasurable input parameters, they can also be given through user experience, thereby determining the mean value matrix and covariance matrix corresponding to the measurement data.
[0091] In step 220, a Latin hypercube sampling matrix is constructed, and the Latin hypercube sampling matrix is mapped to a multivariate Gaussian distribution based on a mean value matrix and a covariance matrix to obtain multivariate Gaussian sampling data.
[0092] According to an embodiment of the present disclosure, Latin hypercube sampling is a method for approximating random sampling from a multivariate parameter distribution, and a Latin hypercube sampling matrix refers to a matrix constructed by Latin hypercube sampling. Multivariate Gaussian distribution is also called multivariate Gaussian distribution, which is a generalization of a single-dimensional normal distribution to multiple dimensions.
[0093] It can be understood that the Latin hypercube sampling matrix is mapped to a multivariate Gaussian distribution based on the mean value matrix and the covariance matrix to obtain multivariate Gaussian sampling data, including: decomposing the covariance matrix using the square root method to obtain a covariance decomposition matrix; based on the mean value matrix and the covariance decomposition matrix, the Latin hypercube sampling matrix is mapped to a multivariate Gaussian distribution to obtain multivariate Gaussian sampling data.
[0094] According to an embodiment of the present disclosure, the square root method refers to the decomposition of a symmetric positive definite covariance matrix as a product of a lower triangular matrix and its transpose, wherein the lower triangular matrix refers to the covariance decomposition matrix. Here, the method of decomposing the covariance matrix using the square root method is described in combination with the formula:
[0095] C=LL T ;
[0096] Where C represents the covariance matrix; L represents the covariance decomposition matrix; L T Represents the transposed matrix of the covariance decomposition matrix L.
[0097] It should be noted that when the numpy open source database is used to execute the sample construction method of the uncertainty analysis of the present application, the covariance matrix C can also be decomposed by the function numpy.linalg.cholesky(C) to obtain the covariance decomposition matrix L.
[0098] Furthermore, the present application maps the Latin hypercube sampling matrix to a multivariate Gaussian distribution based on the mean value matrix and the covariance decomposition matrix, so that the multivariate Gaussian sampling data finally determined retains the mean value characteristics and covariance characteristics corresponding to the measured data, that is, it ensures that the generated multivariate Gaussian sampling data has statistical characteristics similar to the measured data of each input parameter, and accurately determines the distribution of the input parameters. Here, the step of "mapping the Latin hypercube sampling matrix to a multivariate Gaussian distribution based on the mean value matrix and the covariance decomposition matrix to obtain the multivariate Gaussian sampling data" is explained in combination with a mathematical formula:
[0099] S=LU T +M
[0100] Among them, S represents multivariate Gaussian sampling data; L represents the covariance decomposition matrix; U T represents the transposed matrix of the Latin hypercube sampling matrix; M represents the mean matrix.
[0101] In step 230, the Shannon diversity index data corresponding to the multivariate Gaussian sampling data is determined, and the sample minimum distance corresponding to the multivariate Gaussian sampling data is determined.
[0102] According to an embodiment of the present disclosure, Shannon diversity index data is used to indicate the diversity of sample data in multivariate Gaussian sampling data. The more complex the diversity of sample data, the more likely it is to represent the characteristics of the global variable space. The present application can measure the diversity of multivariate Gaussian sampling data by Shannon diversity index data. The higher the Shannon diversity index data, the stronger the diversity of multivariate Gaussian sampling data.
[0103] According to an embodiment of the present disclosure, the sample minimum distance refers to the minimum distance between any two sample data in the multivariate Gaussian sampling data, and the sample minimum distance is used to indicate the correlation of the sample data in the multivariate Gaussian sampling data. The present application can measure the correlation of the sample data of the multivariate Gaussian sampling data by the sample minimum distance. The larger the sample minimum distance, the smaller the correlation of the sample data of the multivariate Gaussian sampling data.
[0104] In step 240 , input samples of the proxy model are determined based on the Shannon diversity index data and the sample minimum distance.
[0105] According to an embodiment of the present disclosure, after the Shannon diversity index data and the minimum sample distance are determined, the diversity and correlation of the multivariate Gaussian sampling data can be measured based on the Shannon diversity index data and the minimum sample distance to ensure that the input samples of the final determined proxy model have stronger diversity and smaller correlation.
[0106] The above steps 210 to 240 perform uniform sampling by constructing a Latin hypercube sampling matrix so that the sample data is evenly distributed in each dimension, thereby improving the uniformity of sampling. Moreover, after obtaining the Latin hypercube sampling matrix, the sample data in the Latin hypercube sampling matrix will be mapped to a multivariate Gaussian distribution based on the mean value matrix and covariance matrix corresponding to the measured data to obtain multivariate Gaussian sampling data. The multivariate Gaussian sampling data can retain the mean characteristics and covariance characteristics of the measured data, and accurately determine the distribution of the input parameters. Finally, the Shannon diversity index data and the sample minimum distance corresponding to the multivariate Gaussian sampling data will also be calculated to evaluate the diversity and representativeness of the multivariate Gaussian sampling data through the Shannon diversity index data, and to evaluate the correlation of the multivariate Gaussian sampling data through the sample minimum distance. The present application determines the input samples of the proxy model based on the Shannon diversity index data and the sample minimum distance, thereby ensuring the diversity and representativeness of the input samples of the proxy model.
[0107] The above is a general description of step 210 to step 240. Since step 210 has been described in sufficient detail above, only the specific implementation process of step 220, step 230 and step 240 will be described in detail below.
[0108] Detailed description of step 220
[0109] In step 220, a Latin hypercube sampling matrix is constructed, and the Latin hypercube sampling matrix is mapped to a multivariate Gaussian distribution based on a mean value matrix and a covariance matrix to obtain multivariate Gaussian sampling data.
[0110] In one embodiment, referring to Figure 3 , step 220 comprises:
[0111] Step 310: Determine the number of columns of the Latin hypercube sampling matrix based on the number of types of input parameters, and determine the number of rows of the Latin hypercube sampling matrix based on the number of measurement data corresponding to each input parameter;
[0112] Step 320: for the multiple sample data in each column of the Latin hypercube sampling matrix, equally divide the preset first value range based on the number of measurement data corresponding to each input parameter, determine the value range corresponding to each sample data in the Latin hypercube sampling matrix, and determine the value of each sample data based on the value range;
[0113] Step 330: construct a Latin hypercube sampling matrix based on the number of rows of the Latin hypercube sampling matrix, the number of columns of the Latin hypercube sampling matrix, and the value of the sample data in each column of the Latin hypercube sampling matrix.
[0114] Steps 310 to 330 are described in detail below.
[0115] According to the embodiments of the present disclosure, the steps of constructing a Latin hypercube sampling matrix are described here in combination with a mathematical formula: first, the number of columns of the Latin hypercube sampling matrix is determined based on the number of types of input parameters, and the number of rows of the Latin hypercube sampling matrix is determined based on the number of measurement data corresponding to each input parameter. Assuming that the number of measurement data corresponding to each input parameter is m, and the number of types of input parameters is n, the number of rows of the Latin hypercube sampling matrix can be determined to be m, and the number of columns of the Latin hypercube sampling matrix can be determined to be n. Based on this, it can be determined that the dimension of a Latin hypercube sampling matrix is m×n.
[0116] Furthermore, each column of the Latin hypercube sampling matrix corresponds to a sample data corresponding to an input parameter, and each row corresponds to a set of Latin hypercube samples. Here, the Latin hypercube sampling matrix is denoted as U. Assuming that the dimension of the Latin hypercube sampling matrix is m×n, for multiple sample data in the i-th column of U, the preset first numerical range (0,1) is divided into m equal parts to determine the value range corresponding to each sample data in the i-th column. Specifically, the value range of the sample data in the i-th column and the j-th row is
[0117] ((j-1) / m, j / m), it can be determined that the specific value of the sample data in the i-th column and j-th row is obtained by randomly selecting a value in the value range ((j-1) / m, j / m). When the numpy open source database is used to execute the sample construction method of uncertainty analysis of the present application, the function U[j,i]=numpy.random.uniform((j-1) / m,j / m) can also be used to randomly select a value in the value range ((j-1) / m,j / m) to determine the specific value of the sample data in the i-th column and j-th row.
[0118] After determining that m sample data in the i-th column of U are obtained, it is also necessary to randomly shuffle all sample data in the i-th column of U. When using the numpy open source database to execute the sample construction method of uncertainty analysis of the present application, all sample data in the i-th column of U can also be randomly shuffled by the function U[:,i]=numpy.random.shuffle(U[:,i]).
[0119] Finally, after traversing all the columns of the Latin hypercube sampling matrix U, the value of the sample data in each column of the Latin hypercube sampling matrix U can be determined. At this point, the construction of the Latin hypercube sampling matrix is completed.
[0120] This application can ensure that the sample data is evenly distributed in each dimension by constructing a Latin hypercube sampling matrix to better cover the entire sampling space, reduce the overlap between the sampled data, and make the input parameters evenly distributed. At the same time, Latin hypercube sampling can reduce the correlation between the sampled data and better capture the variability of the sample data by dividing the sampling space into multiple sub-intervals and randomly sampling.
[0121] Detailed description of step 230
[0122] In step 230, the Shannon diversity index data corresponding to the multivariate Gaussian sampling data is determined, and the minimum sample distance corresponding to the multivariate Gaussian sampling data is determined.
[0123] In one embodiment, referring to Figure 4 , step 230 comprises:
[0124] Step 410: Perform cluster analysis on the multivariate Gaussian sampling data based on a preset clustering algorithm to obtain the number of clusters and the number of samples in each cluster;
[0125] Step 420: Determine Shannon diversity index data based on the number of clusters, the number of samples in each cluster, and the total number of samples included in the multivariate Gaussian sampling data.
[0126] Steps 410 to 420 are described in detail below.
[0127] According to an embodiment of the present disclosure, after mapping the Latin hypercube sampling matrix to a multivariate Gaussian distribution and obtaining multivariate Gaussian sampling data, it is first necessary to select a suitable clustering algorithm to perform cluster analysis on the multivariate Gaussian sampling data. Clustering algorithms include parameter-free hierarchical clustering methods, etc. (such as the AgglomerativeClustering method of the cluster module in the open source machine learning library sklearn, etc.). In practical applications, the specific type of the clustering algorithm needs to be determined according to the specific type of the input parameter. After determining that the clustering algorithm is obtained, the multivariate Gaussian sampling data S is input into the clustering algorithm, and the multivariate Gaussian sampling data S is divided into K clusters based on the clustering algorithm, and the number of samples in each cluster is determined. Then, the ratio p of the number of samples in the i-th cluster and the total number of samples contained in the multivariate Gaussian sampling data S is determined by the number of samples in each cluster and the total number of samples contained in the multivariate Gaussian sampling data S. i Based on this, we can calculate the ratio p of the number of samples in the ith cluster and the total number of samples contained in the multivariate Gaussian sampling data S. i Determine the Shannon diversity index data. The formula for determining the Shannon diversity index data can be expressed as:
[0128]
[0129] Among them, H represents the Shannon diversity index data; K represents the number of clusters; p i It represents the ratio of the number of samples in the i-th cluster to the total number of samples contained in the multivariate Gaussian sampling data S.
[0130] It should be noted that the larger the Shannon diversity index data H is, the stronger the diversity of the multivariate Gaussian sampling data S is.
[0131] In one embodiment, referring to Figure 5 , step 230 comprises:
[0132] Step 510: for any two sample data in the multivariate Gaussian sampling data, the distance between the two sample data is calculated based on a preset distance measurement algorithm;
[0133] Step 520: Determine the minimum value of the distance between any two sample data in the multivariate Gaussian sampling data as the sample minimum distance.
[0134] Steps 510 to 520 are described in detail below.
[0135] According to the embodiment of the present disclosure, after obtaining the Shannon diversity index data H, it is also necessary to select a suitable distance measurement algorithm to calculate the distance between any two sample data in the multivariate Gaussian sampling data S. The distance measurement algorithm includes the Euclidean distance method, the Manhattan distance method, or the correlation coefficient method, and the specific type of the distance measurement algorithm needs to be determined according to the specific type of the input parameter. The distance measurement algorithm for determining the distance between any two sample data in the multivariate Gaussian sampling data S is:
[0136]
[0137] Among them, x i and i represents any two sample data in the multivariate Gaussian sampling data S; d ij represents the distance between any two sample data in the multivariate Gaussian sampling data S; t represents a constant, and the value of t is determined based on the distance measurement algorithm.
[0138] In determining the distance d between any two sample data in the multivariate Gaussian sampling data S ij After that, we need to calculate the distance d between any two sample data in the multivariate Gaussian sampling data S ij The minimum value of is determined as the minimum distance of the sample. The formula for determining the minimum distance of the sample can be expressed as:
[0139]
[0140] Among them, x i and i represents any two sample data in the multivariate Gaussian sampling data S; d(x i ,y i ) represents the distance between any two sample data in the multivariate Gaussian sampling data S; d min Indicates the minimum distance of samples.
[0141] It should be noted that the minimum sample distance d min The larger it is, the smaller the correlation of the multivariate Gaussian sampling data S is.
[0142] Detailed description of step 240
[0143] In step 240, the input samples of the proxy model are determined based on the Shannon diversity index data and the minimum distance of the samples.
[0144] In one embodiment, referring to Figure 6 , step 240 comprises:
[0145] Step 610: Obtain a first weight and a second weight;
[0146] Step 620: Calculate the weighted sum of the Shannon diversity index data and the minimum distance of the sample based on the first weight and the second weight as the feature score of the multivariate Gaussian sampling data;
[0147] Step 630: Determine input samples of the proxy model based on the feature scores of the multivariate Gaussian sampling data.
[0148] Steps 610 to 630 are described in detail below.
[0149] According to an embodiment of the present disclosure, the first weight refers to the minimum distance d of the samples. min The weight indicates the minimum distance d of the sample min The importance of the first weight is in the range of (0,1). The specific value of the first weight is determined by the user based on experience. The second weight refers to the weight of the Shannon diversity index data H, which indicates the importance of the Shannon diversity index data H. The value of the second weight is also in the range of (0,1). The specific value of the second weight is determined by the user based on experience. The feature score of multivariate Gaussian sampling data can comprehensively reflect the diversity and correlation of multivariate Gaussian sampling data. The higher the feature score of multivariate Gaussian sampling data, the stronger the diversity of the multivariate Gaussian sampling data and the smaller the correlation. The formula for determining the feature score of multivariate Gaussian sampling data can be expressed as:
[0150] e=w1×d min +w2×H;
[0151] Where e represents the characteristic score of multivariate Gaussian sampling data; w1 represents the first weight; w2 represents the second weight; d min represents the minimum distance of samples; H represents the Shannon diversity index data.
[0152] The specific method of "determining the input samples of the proxy model based on the feature scores of multivariate Gaussian sampling data" will be described in detail below.
[0153] In one embodiment, determining the input sample of the proxy model based on the feature score of the multivariate Gaussian sampling data includes: if the feature score of the multivariate Gaussian sampling data reaches a first threshold, determining the multivariate Gaussian sampling data as the input sample of the proxy model;
[0154] The method also includes: if the feature score of the multivariate Gaussian sampling data is less than a first threshold, returning to the step of constructing a Latin hypercube sampling matrix until the feature score of the multivariate Gaussian sampling data reaches the first threshold.
[0155] It can be understood that the present application has two methods for determining the input samples of the proxy model based on the feature scores of the multivariate Gaussian sampling data. Here, the first method for determining the input samples of the proxy model based on the feature scores of the multivariate Gaussian sampling data is described: Assuming that an input sample containing 1000 sample data needs to be constructed, the operations of steps 220 to 240 need to be repeated to repeatedly construct the multivariate Gaussian sampling data containing 1000 sample data. If the feature score of the multivariate Gaussian sampling data currently constructed containing 1000 sample data reaches the first threshold, it means that the diversity and correlation of the multivariate Gaussian sampling data have met the preset requirements. At this time, the multivariate Gaussian sampling data can be determined as the input sample of the proxy model, and the operations of repeating steps 220 to 240 are stopped. Among them, the first threshold is pre-set by the staff. On the other hand, if the feature score of the multivariate Gaussian sampling data is less than the first threshold, it means that the currently constructed multivariate Gaussian sampling data does not meet the preset requirements. At this time, it is necessary to return to step 220, reconstruct the Latin hypercube sampling matrix, and re-determine a multivariate Gaussian sampling data containing 1000 sample data until the feature score of this multivariate Gaussian sampling data reaches the first threshold. Then, the sampling operation is stopped, and the multivariate Gaussian sampling data with a feature score higher than the first threshold is determined as the input sample of the proxy model.
[0156] In one embodiment, referring to Figure 7 , step 630 further includes:
[0157] Step 710: Obtain historical sampling data;
[0158] Step 720: If the feature score of the multivariate Gaussian sampling data reaches a second threshold, the sample data in the multivariate Gaussian sampling data is added to the historical sampling data to obtain expanded sampling data;
[0159] Step 730, determining the feature score of the expanded sampling data;
[0160] Step 740: Determine input samples of the proxy model based on the feature scores of the expanded sampling data.
[0161] Steps 710 to 740 are described in detail below.
[0162] According to an embodiment of the present disclosure, a second method for determining the input sample of the proxy model based on the feature score of multivariate Gaussian sampling data is described here: Assuming that an input sample containing 1,000 sample data needs to be constructed, steps 220 to 240 need to be repeated each time to construct a multivariate Gaussian sampling data containing 100 sample data. Assuming that the historical sampling data already includes 400 sample data, if the feature score of the currently constructed multivariate Gaussian sampling data containing 100 sample data reaches the second threshold, it means that the diversity and correlation of the multivariate Gaussian sampling data have met the preset requirements. At this time, the 100 sample data of the multivariate Gaussian sampling data can be added to the historical sampling data to obtain expanded sampling data, and the expanded sampling data includes 500 sample data. Among them, the second threshold is pre-set by the staff.
[0163] Furthermore, after the expanded sampling data is obtained, the weighted sum of the Shannon diversity index data and the sample minimum distance of the expanded sampling data will be calculated to determine the feature score of the expanded sampling data and measure the diversity and relevance of the expanded sampling data.
[0164] In one embodiment, the input sample of the proxy model is determined based on the feature score of the expanded sampled data, including: if the feature score of the expanded sampled data reaches a third threshold, the expanded sampled data is used as the updated historical sampled data; returning to the step of constructing the Latin hypercube sampling matrix until the number of sample data in the updated historical sampled data meets the second threshold, and the updated historical sampled data is determined as the input sample of the proxy model;
[0165] The method also includes: if the feature score of the expanded sampling data is less than a third threshold, returning to the step of constructing a Latin hypercube sampling matrix until the number of sample data in the updated historical sampling data meets a second threshold, and determining the updated historical sampling data as the input sample of the proxy model.
[0166] It is understandable that, assuming that an input sample containing 1000 sample data needs to be constructed, steps 220 to 240 need to be repeated each time to construct a multivariate Gaussian sampling data containing 100 sample data. After adding 100 sample data of the multivariate Gaussian sampling data to the historical sampling data to obtain the expanded sampling data containing 500 sample data, if the feature score of the expanded sampling data obtained by adding the determined feature score reaches the third threshold, it means that the diversity and relevance of the expanded data have met the preset requirements, and the expanded sampling data can be used as the updated historical sampling data. Among them, the third threshold is pre-set by the staff. Further, since the updated historical sampling data contains 500 sample data at this time, and does not reach the target 1000 sample data, it is necessary to return to step 220 and construct a multivariate Gaussian sampling data containing 100 sample data until the number of sample data contained in the updated historical sampling data reaches the target 1000 sample data, and the updated historical sampling data containing 1000 sample data is determined as the input sample of the proxy model.
[0167] On the other hand, if the feature score of the expanded sampling data is less than the third threshold, it means that the currently constructed expanded sampling data does not meet the preset requirements. At this time, the currently constructed expanded sampling data and multivariate Gaussian sampling data need to be discarded, and there is no need to update the historical sampling data. After discarding the currently constructed expanded sampling data and multivariate Gaussian sampling data, return to step 220, and reconstruct a multivariate Gaussian sampling data containing 100 sample data until the number of sample data contained in the updated historical sampling data reaches the target 1000 sample data, and the updated historical sampling data containing 1000 sample data is determined as the input sample of the proxy model.
[0168] Description of a sample construction system and a computer-readable storage medium for uncertainty analysis of the present disclosure
[0169] Second, refer to Figure 8 , the embodiment of the present application provides a sample construction system for uncertainty analysis, including:
[0170] at least one memory 200;
[0171] at least one processor 100;
[0172] at least one program;
[0173] The program is stored in the memory 200, and the processor 100 executes at least one program to implement:
[0174] A sample construction method for uncertainty analysis as in any embodiment of the first aspect of the present application.
[0175] The processor 100 and the memory 200 may be connected via a bus or in other ways.
[0176] The memory 200, as a non-transitory readable storage medium, can be used to store non-transitory software instructions and non-transitory instructables. In addition, the memory 200 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one disk storage device, a flash memory device, or other non-transitory solid-state storage device. It is understood that the memory 200 may optionally include a memory 200 remotely arranged relative to the processor 100, and these remote memories 200 may be connected to the processor 100 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0177] The processor 100 executes non-transitory software instructions, commands and signals stored in the memory 200, thereby performing various functional applications and data processing, that is, implementing the sample construction method for uncertainty analysis of the first aspect embodiment described above.
[0178] The non-transient software instructions and instructions required to implement the sample construction method of the uncertainty analysis of the above embodiment are stored in the memory 200. When executed by the processor 100, the sample construction method of the uncertainty analysis of the first aspect of the present application is executed, for example, the sample construction method of the uncertainty analysis described above is executed. Figure 2 Steps 210 to 240, Figure 3 Steps 310 to 330, Figure 4 Steps 410 to 420, Figure 5 Steps 510 to 520, Figure 6 Steps 610 to 630, Figure 7 Steps 710 to 740 in .
[0179] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer-executable signal, and the computer-executable signal is used to execute:
[0180] A sample construction method for uncertainty analysis according to any embodiment of the first aspect of the application.
[0181] For example, executing the above description Figure 2 Steps 210 to 240, Figure 3 Steps 310 to 330, Figure 4 Steps 410 to 420, Figure 5 Steps 510 to 520 in Figure 6 Steps 610 to 630, Figure 7 Steps 710 to 740 in .
[0182] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., may be located in one place, or may be distributed over multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0183] Through the description of the above embodiments, it will be appreciated by those skilled in the art that all or some steps and systems in the disclosed methods above can be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a readable medium, and the readable medium can include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as a computer-readable signal, a data structure, an instruction module or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, magnetic tape, disk storage or other magnetic storage device, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media generally include computer readable signals, data structures, instruction modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.
[0184] The embodiments of the present application are described in detail above in conjunction with the accompanying drawings, but the present application is not limited to the above embodiments, and various changes can be made within the knowledge scope of ordinary technicians in the relevant technical field without departing from the purpose of the present application.
Claims
1. A sample construction method for uncertainty analysis, characterized in that: include: Determining, based on measurement data of various input parameters in a numerical reactor, a mean value matrix and a covariance matrix corresponding to the measurement data, the input parameters being associated with a proxy model of the numerical reactor; Constructing a Latin hypercube sampling matrix, and mapping the Latin hypercube sampling matrix to a multivariate Gaussian distribution based on the mean value matrix and the covariance matrix to obtain multivariate Gaussian sampling data; Determine the Shannon diversity index data corresponding to the multivariate Gaussian sampling data, and determine the minimum sample distance corresponding to the multivariate Gaussian sampling data; The input samples of the proxy model are determined based on the Shannon diversity index data and the sample minimum distance.
2. The sample construction method for uncertainty analysis according to claim 1, characterized in that: The step of determining the input sample of the proxy model based on the Shannon diversity index data and the sample minimum distance comprises: Get a first weight and a second weight; Based on the first weight and the second weight, calculating a weighted sum of the Shannon diversity index data and the sample minimum distance as a feature score of the multivariate Gaussian sampling data; An input sample of the proxy model is determined based on the feature scores of the multivariate Gaussian sampling data.
3. The sample construction method for uncertainty analysis according to claim 2, characterized in that: The step of determining the input sample of the proxy model based on the feature score of the multivariate Gaussian sampling data comprises: If the feature score of the multivariate Gaussian sampling data reaches a first threshold, determining the multivariate Gaussian sampling data as an input sample of the proxy model; The method further comprises: If the characteristic score of the multivariate Gaussian sampling data is less than the first threshold, return to the step of constructing the Latin hypercube sampling matrix until the characteristic score of the multivariate Gaussian sampling data reaches the first threshold.
4. The sample construction method for uncertainty analysis according to claim 2, characterized in that: The step of determining the input sample of the proxy model based on the feature score of the multivariate Gaussian sampling data comprises: Get historical sampling data; If the feature score of the multivariate Gaussian sampling data reaches a second threshold, adding the sample data in the multivariate Gaussian sampling data to the historical sampling data to obtain expanded sampling data; Determining a feature score of the expanded sampling data; An input sample of the proxy model is determined based on the feature scores of the augmented sampling data.
5. The sample construction method for uncertainty analysis according to claim 4, characterized in that: The step of determining the input sample of the proxy model based on the feature score of the expanded sampling data comprises: If the feature score of the expanded sampling data reaches a third threshold, using the expanded sampling data as updated historical sampling data; Returning to the step of constructing the Latin hypercube sampling matrix, until the number of sample data in the updated historical sampling data meets the second threshold, the updated historical sampling data is determined as the input sample of the proxy model; The method further comprises: If the feature score of the expanded sampling data is less than the third threshold, return to the step of constructing the Latin hypercube sampling matrix until the number of sample data in the updated historical sampling data meets the second threshold, and determine the updated historical sampling data as the input sample of the proxy model.
6. The sample construction method for uncertainty analysis according to claim 1, characterized in that: The step of constructing a Latin hypercube sampling matrix comprises: Determine the number of columns of the Latin hypercube sampling matrix based on the number of types of input parameters, and determine the number of rows of the Latin hypercube sampling matrix based on the number of measurement data corresponding to each input parameter; For multiple sample data in each column of the Latin hypercube sampling matrix, equally divide the preset first value range based on the number of measurement data corresponding to each input parameter, determine the value range corresponding to each sample data in the Latin hypercube sampling matrix, and determine the value of each sample data based on the value range; A Latin hypercube sampling matrix is constructed based on the number of rows of the Latin hypercube sampling matrix, the number of columns of the Latin hypercube sampling matrix and the value of sample data in each column of the Latin hypercube sampling matrix.
7. The sample construction method for uncertainty analysis according to claim 1, characterized in that: The determining of the Shannon diversity index data corresponding to the multivariate Gaussian sampling data includes: Performing cluster analysis on the multivariate Gaussian sampling data based on a preset clustering algorithm to obtain the number of clusters and the number of samples in each cluster; Shannon diversity index data is determined based on the number of clusters, the number of samples in each cluster and the total number of samples included in the multivariate Gaussian sampling data.
8. The sample construction method for uncertainty analysis according to claim 1, characterized in that: The determining of the minimum sample distance corresponding to the multivariate Gaussian sampling data comprises: For any two sample data in the multivariate Gaussian sampling data, the distance between the two sample data is calculated based on a preset distance measurement algorithm; The minimum value of the distance between any two sample data in the multivariate Gaussian sampling data is determined as the sample minimum distance.
9. A sample construction system for uncertainty analysis, characterized in that: include: at least one memory; at least one processor; at least one program; The program is stored in the memory, and the processor executes at least one of the programs to implement: A sample construction method for uncertainty analysis according to any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer-executable signal, wherein the computer-executable signal is used to execute: A sample construction method for uncertainty analysis according to any one of claims 1 to 8.
Citation Information
Cited By
Sensitivity analysis and uncertainty quantification framework for nuclear reactor design
WO2026217613A1
