Leakage diagnosis method, device and electronic equipment for supercritical water reactor

Abstract

The application relates to the technical field of nuclear reactors, in particular to a leakage diagnosis method and device for a supercritical water reactor and electronic equipment. The method comprises the following steps: acquiring an acoustic time domain signal in a to-be-measured region of a supercritical water reactor and physical parameter information in the to-be-measured region which is synchronous with the acoustic time domain signal, and determining an original time-frequency matrix corresponding to the acoustic time domain signal; inputting the original time-frequency matrix into a reconstruction model constrained by physical information to obtain a reconstruction time-frequency matrix output by the reconstruction model and physical parameter prediction information; determining whether the acoustic time domain signal is abnormal based on the reconstruction time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information and the physical parameter information; if the acoustic time domain signal is abnormal, determining whether there is leakage in the to-be-measured region based on the original time-frequency matrix and the physical parameter information; and if there is leakage in the to-be-measured region, determining positioning information of a leakage source in the to-be-measured region based on the acoustic time domain signal and Fermat's principle.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of nuclear reactors, in particular to a leakage diagnosis method and device for a supercritical water reactor and an electronic device. BACKGROUND

[0002] As a fourth-generation nuclear reactor with great potential, the supercritical water-cooled reactor has the core advantages of high thermal efficiency and simplified system design. The realization of these advantages depends on the extreme operating conditions in the supercritical region. Under such conditions, the supercritical water coolant exhibits unique physical properties. It is neither liquid nor gas, and its key physical parameters such as density, viscosity, specific heat, and sound speed change dramatically and highly nonlinearly with small changes in temperature and pressure, especially near the "pseudo-critical point".

[0003] However, this unique physical environment poses a severe challenge to the structural materials of the supercritical water reactor, and also makes the traditional leakage monitoring technology almost completely ineffective. Therefore, there is an urgent need for a leakage diagnosis technology for the supercritical water reactor to ensure the safe operation of the supercritical water reactor. SUMMARY

[0004] The purpose of the present application is to provide a leakage diagnosis method and device for a supercritical water reactor and an electronic device, which combines the physical information and acoustic data of the supercritical water reactor to accurately diagnose and locate the leakage of the supercritical water reactor, thereby ensuring the safe operation of the nuclear reactor.

[0005] In a first aspect, the present application provides a leakage diagnosis method for a supercritical water reactor, comprising:

[0006] Obtaining an acoustic time-domain signal in a to-be-measured region of the supercritical water reactor, and physical parameter information in the to-be-measured region synchronized with the acoustic time-domain signal, and determining an original time-frequency matrix corresponding to the acoustic time-domain signal;

[0007] Inputting the original time-frequency matrix into a reconstruction model constrained by physical information to obtain a reconstructed time-frequency matrix and physical parameter prediction information output by the reconstruction model; the reconstruction model has learned the ability to reconstruct normal acoustic signals and the ability to predict physical parameters based on normal acoustic signals;

[0008] Based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information, and the physical parameter information, determining whether the acoustic time-domain signal is abnormal;

[0009] If the acoustic time-domain signal is abnormal, determining whether there is a leakage in the to-be-measured region based on the original time-frequency matrix and the physical parameter information.

[0010] If there is a leakage in the to-be-tested region, based on the acoustic time domain signal and Fermat's principle, the positioning information of the leakage source in the to-be-tested region is determined.

[0011] In some embodiments, the determination of the original time-frequency matrix corresponding to the acoustic time domain signal comprises:

[0012] The acoustic time domain signal is subjected to continuous wavelet transform or short-time Fourier transform to obtain the original time-frequency matrix.

[0013] In some embodiments, the determination of whether the acoustic time domain signal is abnormal based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information and the physical parameter information comprises:

[0014] A first anomaly score between the original time-frequency matrix and the reconstructed time-frequency matrix is determined; the first anomaly score is used to represent the difference between the original time-frequency matrix and the reconstructed time-frequency matrix;

[0015] A second anomaly score between the physical parameter information and the physical parameter prediction information is determined; the second anomaly score is used to represent the difference between the physical parameter information and the physical parameter prediction information;

[0016] Based on the first anomaly score and the second anomaly score, a comprehensive anomaly score is determined.

[0017] If the comprehensive anomaly score is greater than or equal to a threshold value, it is determined that the acoustic time domain signal is abnormal.

[0018] As a possible implementation, the determination of the comprehensive anomaly score based on the first anomaly score and the second anomaly score comprises:

[0019] The first anomaly score and the second anomaly score are subjected to weighted summation operation, and the operation result is determined as the comprehensive anomaly score.

[0020] In some embodiments, the determination of whether there is a leakage in the to-be-tested region based on the original time-frequency matrix and the physical parameter information comprises:

[0021] Based on the original time-frequency matrix and the physical parameter information, an acoustic input tensor is determined.

[0022] The acoustic input tensor is input into a leakage prediction model to obtain a leakage prediction result output by the leakage prediction model; the leakage prediction model is trained based on training samples comprising acoustic input tensor samples and their corresponding leakage labels.

[0023] Based on the leakage prediction result, it is determined whether there is leakage in the to-be-tested region.

[0024] As a possible implementation manner, the physical parameter information includes temperature information and pressure information; and the determining of the acoustic input tensor based on the original time-frequency matrix and the physical parameter information includes:

[0025] Based on the temperature information and the pressure information, a temperature matrix and a pressure matrix with the same dimension as the original time-frequency matrix are respectively constructed;

[0026] Based on the temperature information and the pressure information, a density gradient matrix is determined; the density gradient matrix has the same dimension as the original time-frequency matrix;

[0027] Based on the original time-frequency matrix, the temperature matrix, the pressure matrix and the density gradient matrix, splicing is performed along the channel dimension to obtain the acoustic input tensor.

[0028] In some embodiments, the acoustic time-domain signal is acquired based on an acoustic emission sensor array; the acoustic emission sensor array includes N acoustic emission sensors, N being an integer greater than 1; and the determining of the positioning information of the leakage source in the to-be-tested region based on the acoustic time-domain signal and Fermat's principle includes:

[0029] A non-uniform sound speed field in the to-be-tested region is determined.

[0030] Based on a cross-correlation algorithm, an actual time delay difference between the acoustic time-domain signals respectively acquired by any two acoustic emission sensors in the N acoustic emission sensors is determined.

[0031] For any point in the to-be-tested region as a source point, based on the non-uniform sound speed field and the sound wave transmission paths between the source point and each acoustic emission sensor, a theoretical time delay difference of sound waves reaching each acoustic emission sensor is determined.

[0032] Based on the difference between the theoretical time delay difference and the actual time delay difference, the position of the source point is constantly updated until the difference meets a preset condition, and the position information of the latest source point is determined as the positioning information of the leakage source.

[0033] As a possible implementation manner, the determining of the non-uniform sound speed field in the to-be-tested region includes:

[0034] A three-dimensional temperature field and a three-dimensional pressure field of the to-be-tested region are determined through fluid dynamics simulation or finite element interpolation algorithm.

[0035] Based on the three-dimensional temperature field and the three-dimensional pressure field, the non-uniform sound speed field is constructed.

[0036] In a second aspect, the present application provides a leakage diagnosis device of a supercritical water reactor, comprising:

[0037] An acquisition module is configured to acquire an acoustic time-domain signal in a to-be-tested region of a supercritical water reactor and physical parameter information in the to-be-tested region synchronized with the acoustic time-domain signal, and determine an original time-frequency matrix corresponding to the acoustic time-domain signal;

[0038] A reconstruction module is configured to input the original time-frequency matrix into a reconstruction model with physical information constraint, and obtain a reconstructed time-frequency matrix and physical parameter prediction information output by the reconstruction model; the reconstruction model has learned the ability to reconstruct normal acoustic signals and the ability to predict physical parameters based on normal acoustic signals;

[0039] A determination module is configured to determine whether the acoustic time-domain signal is abnormal based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information and the physical parameter information;

[0040] A diagnosis module is configured to, if the acoustic time-domain signal is abnormal, determine whether there is leakage in the to-be-tested region based on the original time-frequency matrix and the physical parameter information;

[0041] A positioning module is configured to, if there is leakage in the to-be-tested region, determine positioning information of a leakage source in the to-be-tested region based on the acoustic time-domain signal and Fermat's principle.

[0042] In a third aspect, the present application provides an electronic device comprising a processor and a memory storing a computer program, wherein the processor implements the leakage diagnosis method of the supercritical water reactor according to the first aspect when executing the program.

[0043] The leakage diagnosis method, device and electronic device of the supercritical water reactor provided by the present application have the following beneficial effects:

[0044] The acoustic time domain signal in the to-be-tested region of the supercritical water reactor is acquired, and the physical parameter information in the to-be-tested region synchronized with the acoustic time domain signal is acquired, and an original time-frequency matrix corresponding to the acoustic time domain signal is determined; the original time-frequency matrix is input into a reconstruction model constrained by physical information to obtain a reconstructed time-frequency matrix and physical parameter prediction information output by the reconstruction model; the reconstruction model has learned the ability of reconstructing normal acoustic signals and the ability of predicting physical parameters based on the normal acoustic signals; whether the acoustic time domain signal is abnormal is determined based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information and the physical parameter information; if the acoustic time domain signal is abnormal, whether there is leakage in the to-be-tested region is determined based on the original time-frequency matrix and the physical parameter information; if there is leakage in the to-be-tested region, the positioning information of the leakage source in the to-be-tested region is determined based on the acoustic time domain signal and Fermat's principle. The present application combines acoustic data and physical parameter information to realize a complete closed loop from three stages of abnormal identification, leakage diagnosis and leakage source positioning, realizes leakage diagnosis of the supercritical water reactor, and improves the accuracy of leakage diagnosis. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0046] Figure 1 A flowchart of a leakage diagnosis method of a supercritical water reactor provided by the embodiment of the present application;

[0047] Figure 2 A flowchart of another leakage diagnosis method of a supercritical water reactor provided by the embodiment of the present application;

[0048] Figure 3 A structural schematic diagram of a leakage diagnosis device of a supercritical water reactor provided by the embodiment of the present application;

[0049] Figure 4 A structural schematic diagram of an electronic device provided by the embodiment of the present application. DETAILED DESCRIPTION

[0050] The technical solutions of the present application will be described in detail below with reference to the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0051] It should be noted that the macroscopic parameter monitoring method is a common traditional leakage monitoring method, which relies on monitoring pressure, temperature, liquid level and the like to determine leakage, but it is completely inapplicable in a supercritical water environment. Because the supercritical water system itself will have a huge change in physical property parameters during normal power regulation or start-stop process, the normal fluctuations will completely mask the macroscopic parameter changes caused by a small leakage (such as a leakage of g / s level), resulting in that the leakage diagnosis cannot be performed through the parameter monitoring method.

[0052] The acoustic monitoring method is also a common way to monitor leakage, but the supercritical water flowing at a high speed itself will produce intense flow noise, combined with the noise of pumps, valves and the like, forming a strong noise environment with an energy several orders of magnitude higher than the early micro-leakage signal. More fatally, the dramatic change in the physical properties of supercritical water will cause the spectrum and intensity of the background noise to also dynamically change with the working condition. The traditional acoustic method based on a fixed threshold or conventional filtering cannot effectively extract the weak leakage signal in such a complex, dynamic and strong background noise.

[0053] For other leakage monitoring methods, such as the chemical sampling analysis method, although accurate, the response time is long and cannot meet the real-time requirement of safety monitoring. The method based on radioactive monitoring also has the problems of response lag and insufficient sensitivity.

[0054] To solve the above problems, the present application provides a leakage diagnosis method, device and electronic equipment for a supercritical water reactor.

[0055] Figure 1 A flowchart of a leakage diagnosis method for a supercritical water reactor provided by an embodiment of the present application is shown in FIG. 1. As shown in the figure, the method comprises the following steps: Figure 1

[0056] In step 101, an acoustic time-domain signal in a to-be-measured region of a supercritical water reactor is acquired, and physical parameter information synchronous with the acoustic time-domain signal in the to-be-measured region is acquired, and an original time-frequency matrix corresponding to the acoustic time-domain signal is determined.

[0057] In some embodiments, the to-be-measured region is a region in the supercritical water reactor that needs to be subjected to leakage diagnosis. As an example, the to-be-measured region can be a pressure-bearing device in the supercritical water reactor, such as a heat transfer pipe of an intermediate heat exchanger, a main pipeline and the like.

[0058] In some embodiments, the acoustic time-domain signal in the to-be-measured region can be acquired by a plurality of acoustic emission sensors arranged on the outer wall of the to-be-measured device. As an example, an acoustic emission sensor array can be arranged on the outer wall of the monitoring device, the acoustic emission sensor array comprising a plurality of high-temperature-resistant and radiation-resistant acoustic emission sensors, and the plurality of acoustic emission sensors in the acoustic emission sensor array being arranged in a specific geometric configuration, such as a spiral shape, a ring shape and the like.​

[0059] That is, the acoustic time-domain signal in the to-be-tested region can include acoustic time-domain signals collected by multiple acoustic emission sensors respectively.

[0060] In some embodiments, the physical parameter information can include pressure, temperature, etc., which can be collected by sensors arranged in the to-be-tested region, such as multiple temperature sensors and pressure sensors arranged at different positions in the to-be-tested region, so as to obtain the physical parameter information synchronized with the acoustic time-domain signal.

[0061] In some embodiments, the acoustic time-domain signal and the physical parameter information can both be data in a period of time, such as the current execution time, and the acoustic time-domain signal and the physical parameter information including the current execution time and a continuous period of time before the current execution time can be obtained for leakage diagnosis.

[0062] In some embodiments, in order to extract the frequency domain information of the acoustic signal, the acoustic time-domain signal is converted by a time-frequency analysis method to obtain an original time-frequency matrix. The original time-frequency matrix is a two-dimensional matrix, where the rows of the original time-frequency matrix represent frequencies, and the columns represent times. The element in the i-th row and the j-th column of the matrix represents the energy density, phase information, etc. at the time point j and the frequency point i.

[0063] As an example, the implementation process of determining the original time-frequency matrix corresponding to the acoustic time-domain signal includes: performing continuous wavelet transform or short-time Fourier transform on the acoustic time-domain signal to obtain the original time-frequency matrix.

[0064] In step 102, the original time-frequency matrix is input into a physical information constrained reconstruction model to obtain a reconstructed time-frequency matrix output by the reconstruction model and physical parameter prediction information. The reconstruction model has learned the ability to reconstruct normal acoustic signals and the ability to predict physical parameters based on normal acoustic signals.

[0065] In some embodiments, the physical information constrained reconstruction model has the ability to reconstruct normal acoustic information in the to-be-tested region and the ability to predict physical parameters in the to-be-tested region based on normal acoustic signals. The physical information constrained reconstruction model can be a neural network model, such as one constructed based on a variational autoencoder VAE. The normal acoustic signal refers to an acoustic signal collected when the supercritical water reactor is in normal operation and has not failed.

[0066] The standard VAE includes an encoder (for mapping the input x to the latent variable z) and a decoder (for mapping z back to the reconstructed x). To meet the needs of physical parameter prediction, a decoding branch for physical parameter prediction can be added to the standard VAE to predict the physical parameter p from the latent variable z. Through training, the latent variable can not only reconstruct the input, but also predict the physical parameter related to the input, so as to obtain a reconstruction model constrained by physical information. That is, the reconstruction model constrained by physical information can include an encoder and two decoders, one of which is used to reconstruct the acoustic signal, and the other is used to predict the physical parameter.

[0067] As a possible implementation, the acoustic time-domain signal of the supercritical water reactivity in the normal state can be collected based on the acoustic emitter array, and the physical parameter information in the synchronous state can be collected based on the pressure and temperature sensors; the training sample is constructed based on the acoustic time-domain signal in the normal state, the original time-frequency matrix sample converted based on the acoustic time-domain signal, and the physical parameter label; the training sample is input to the initial reconstruction model to obtain the reconstruction time-frequency matrix sample and the physical parameter prediction sample output by the initial reconstruction model; the loss value is calculated based on the reconstruction time-frequency matrix sample, the original time-frequency matrix sample, the physical parameter label and the physical parameter prediction sample; the model parameters are adjusted based on the loss value to obtain the trained reconstruction model.

[0068] As an example, if the reconstruction model constrained by the physical parameter is constructed based on the VAE model, a decoding branch for physical parameter prediction is added to the standard VAE, and the loss value in the training process of the reconstruction model includes the reconstruction loss, the physical parameter prediction loss and the KL divergence loss. Among them, the reconstruction loss is used to represent the difference between the reconstruction time-frequency matrix sample and the original time-frequency matrix sample; the physical parameter prediction loss is used to represent the difference between the physical parameter prediction sample and the physical parameter label; the physical parameter prediction loss and the reconstruction loss can be calculated by means of mean square error or binary cross entropy; the KL divergence loss is used to represent the difference between the distribution of the latent variable and the standard normal distribution, and the KL divergence loss plays a role of regularization, so that the distribution of the latent variable is more regular, continuous and prevents the model from overfitting.

[0069] For example, the loss function in the training process of the reconstruction model constrained by the physical parameter is as follows:

[0070] (1);

[0071] (2);

[0072] (3);

[0073] (4);

[0074] wherein, is a total loss value; is a reconstruction loss; is a KL divergence loss; is a physical information prediction loss; is a sample size; is an i-th original time-frequency matrix sample; is a reconstructed time-frequency matrix sample corresponding to the i-th original time-frequency matrix sample; is a dimension of the latent space; is a variance of a j-th dimension of the encoder output; is a mean of a j-th dimension of the encoder output; is a physical information label; is a physical information prediction sample; is a physical parameter weight.

[0075] At step 103, based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information and the physical parameter information, it is determined whether the acoustic time-domain signal is abnormal.

[0076] It can be understood that the reconstruction model constrained by the physical information can realize reconstruction of the normal acoustic signal and prediction of the physical parameter information, but when the supercritical water reactor leaks, the acoustic data in the reactor will be abnormal, either the acoustic feature cannot be well reconstructed by the reconstruction model or the physical environment predicted by the acoustic feature does not match the real physical environment. Therefore, based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information and the physical parameter information, it can be directly determined whether the acoustic time-domain signal is abnormal.

[0077] In some embodiments, based on the difference between the original time-frequency matrix and the reconstructed time-frequency matrix, and the difference between the physical parameter prediction information and the physical parameter information, an abnormal score can be calculated, and when the abnormal score is greater than a threshold value, it is determined that the acoustic time-domain signal is abnormal.

[0078] As a possible implementation manner, the implementation process of step 103 includes the following steps:

[0079] At step S1, a first abnormal score between the original time-frequency matrix and the reconstructed time-frequency matrix is determined; the first abnormal score is used to represent the difference between the original time-frequency matrix and the reconstructed time-frequency matrix.

[0080] In some embodiments, the difference between the original time-frequency matrix and the reconstructed time-frequency matrix can be determined as the first abnormal score.

[0081] As an example, the first abnormality score can be calculated by the following formula (5):

[0082] A1= (5);

[0083] wherein A1 is the first abnormality score; is the original time-frequency matrix; is the reconstructed time-frequency matrix; denotes the Frobenius norm.

[0084] Step S2, determining a second abnormality score between the physical parameter information and the physical parameter prediction information; the second abnormality score is used to represent the difference between the physical parameter information and the physical parameter prediction information.

[0085] In some embodiments, the second abnormality score can be determined based on the difference between the physical parameter information and the physical parameter prediction information.

[0086] As an example, the second abnormality score can be calculated by the following formula (6).

[0087] A2= (6);

[0088] wherein A2 is the second abnormality score; is the physical parameter information; is the physical parameter prediction information; denotes the L2 norm.

[0089] Step S3, determining a comprehensive abnormality score based on the first abnormality score and the second abnormality score.

[0090] In some embodiments, the first abnormality score and the second abnormality score can be summed, and the sum is taken as the comprehensive abnormality score.

[0091] In some embodiments, the average of the first abnormality score and the second abnormality score can be taken as the comprehensive abnormality score.

[0092] In some embodiments, the first abnormality score and the second abnormality score can be weighted and summed, and the operation result is determined as the comprehensive abnormality score.

[0093] Step S4, if the comprehensive abnormality score is greater than or equal to a threshold value, determining that the acoustic time-domain signal is abnormal.

[0094] That is, a threshold value can be preset, when the comprehensive abnormality score is greater than or equal to the threshold value, it is considered that the acoustic time-domain signal is abnormal; when the comprehensive abnormality score is less than the threshold value, it is considered that the acoustic time-domain signal is not abnormal. If the acoustic time-domain signal is abnormal, step 104 needs to be continued to determine whether the acoustic signal abnormality is caused by leakage.

[0095] Step 104, if the acoustic time-domain signal is abnormal, based on the original time-frequency matrix and the physical parameter information, it is determined whether there is leakage in the to-be-detected region.

[0096] In some embodiments, a leakage prediction model can be constructed in advance, the original time-frequency matrix and the physical parameter information are input into the leakage prediction model to obtain a probability of leakage in the to-be-detected region; if the probability of leakage in the to-be-detected region is greater than a corresponding threshold value, it is determined that there is leakage in the to-be-detected region.

[0097] In some embodiments, based on the original time-frequency matrix and the physical parameter information, whether there is leakage in the to-be-detected region can be realized by the following steps:

[0098] Step S5, based on the original time-frequency matrix and the physical parameter information, an acoustic input tensor is determined.

[0099] In some embodiments, the physical parameter information includes temperature information and pressure information. Based on the original time-frequency matrix and the physical parameter information, the process of determining the acoustic input tensor includes: based on the temperature information and the pressure information, a temperature matrix and a pressure matrix with the same dimension as the original time-frequency matrix are constructed respectively; based on the temperature information and the pressure information, a density gradient matrix is determined; the dimension of the density gradient matrix is the same as that of the original time-frequency matrix; based on the original time-frequency matrix, the temperature matrix, the pressure matrix and the density gradient matrix, splicing is performed along the channel dimension to obtain the acoustic input tensor.

[0100] Wherein, the temperature information can include temperature values at each time in a continuous period of time at different positions, or can be a temperature scalar value in a continuous period of time at different positions. The pressure information can include pressure values at each time in a continuous period of time at different positions, or can be a pressure scalar value in a continuous period of time at different positions. However, the temperature information and the pressure information are different from the dimension of the original time-frequency matrix, which can be broadcasted to a temperature matrix and a pressure matrix with the same dimension as the original time-frequency matrix through expansion.

[0101] In some embodiments, according to the temperature information and the pressure information, the IAPWS-IF97 property database can be consulted to calculate the partial derivative of density with respect to temperature, and also broadcasted into a matrix with the same dimension as the original time-frequency matrix to obtain the density gradient matrix. The partial derivative of density with respect to temperature reaches a peak value in the pseudo-critical region, which is a unique key physical indicator of supercritical water.

[0102] As an example, the expression of the acoustic input tensor is shown in the following formula (7).

[0103] (7);

[0104] wherein, is the acoustic input tensor; is the original time-frequency matrix; is the temperature matrix; is the pressure matrix; is the density gradient matrix; is the splicing processing along the channel dimension.

[0105] That is, each pixel point in the acoustic input tensor includes acoustic features, as well as corresponding temperature, pressure, and density gradient features.

[0106] Step S6, inputting the acoustic input tensor into a leakage prediction model to obtain a leakage prediction result output by the leakage prediction model; the leakage prediction model is trained based on training samples including acoustic input tensor samples and corresponding leakage labels.

[0107] In some embodiments, the leakage prediction model can be a classifier constructed based on a convolutional neural network, which can output a probability value of occurrence of a leakage fault. Since the acoustic input tensor is a tensor containing acoustic, temperature, pressure, density gradient, and other multi-modal information, the leakage prediction model can extract acoustic features and physical environment features based on the acoustic tensor, thereby greatly improving the accuracy of leakage identification. In addition, the acoustic signal anomaly detection and leakage detection are combined in the embodiments of the present application, which can effectively distinguish the noise fluctuation caused by the normal working condition change of the supercritical water reactor and the real leakage signal, and fundamentally eliminate the false alarm caused by normal operation transient.

[0108] Step S7, determining whether there is leakage in the to-be-tested region based on the leakage prediction result.

[0109] In some embodiments, if the probability value of occurrence of leakage in the leakage prediction result is greater than a corresponding probability threshold value, it is determined that there is leakage in the to-be-tested region, otherwise, it is determined that there is no leakage in the to-be-tested region.

[0110] In some other embodiments, the leakage confidence index can also be accumulated in time sequence to determine whether there is leakage in the to-be-tested region. As an example, the leakage fault probability values in a continuous period of time can be subjected to sliding average, exponential weighted moving average, or other operations to obtain an accumulated leakage confidence; if the accumulated leakage confidence is greater than a corresponding threshold value, it is determined that there is leakage in the to-be-tested region.

[0111] Step 105, if there is a leakage in the to-be-tested region, based on the acoustic time-domain signal and Fermat's principle, the positioning information of the leakage source in the to-be-tested region is determined.

[0112] It should be noted that, in the supercritical water reactor, due to the extremely uneven distribution of temperature and pressure, the sound speed is different everywhere in space. That is to say, in the supercritical water reactor, the sound speed is no longer a constant, and the sound wave no longer propagates along a straight line, but along the path with the shortest propagation time. It can be seen that, in the non-uniform medium scenario in the supercritical water reactor, the traditional acoustic wave triangulation positioning method has already been unable to adapt.

[0113] Fermat's principle means that the path of light (or any wave) between two points is the one with the shortest time consumption. Based on the path optimization algorithm of Fermat's principle, the transmission path of the sound wave can be optimized and calculated, and based on the signal delay between each acoustic emission sensor in the acoustic time-domain signal and the theoretical delay determined by the transmission path of the sound wave, the positioning information of the leakage source can be accurately determined through continuous iteration. That is to say, the leakage positioning calculation is no longer a simple geometric problem, but an optimization problem. The goal is to find a hypothetical leakage point in the three-dimensional space of the to-be-tested region of the reactor, and according to the theoretical time difference of the sound wave reaching each sensor calculated based on Fermat's principle, the theoretical time difference is most consistent with the actual measured time difference, and the point where the theory and the actual are most consistent is determined as the final leakage position. This finding process is usually completed through an iterative algorithm (such as simulated annealing, particle swarm optimization, etc.).

[0114] It should also be noted that, in order to implement the leakage diagnosis method of the supercritical water reactor in the embodiment of the present application, the required hardware configuration includes: (1) a sensor array, a plurality of high-temperature-resistant (working temperature > 500°C) and radiation-resistant acoustic emission sensors are arranged on the outer wall of the monitored equipment (such as a heat exchanger shell) in a specific geometric configuration (such as a spiral or a ring); (2) a physical parameter synchronous interface for synchronously collecting real-time physical parameters corresponding to the positions of each sensor from the reactor DCS (Distributed Control System) at a high speed and high precision; (3) a data acquisition unit for amplifying and filtering the acoustic signal, and then performing analog-to-digital conversion at a sampling rate of not less than 5MHz to ensure that the signal is not distorted, that is, the acoustic time-domain signal in the embodiment of the present application is collected based on the data acquisition unit; (4) an intelligent diagnosis core host, which can be a "CPU+GPU" heterogeneous computing platform, the CPU is responsible for executing steps 101 to 105, and the GPU is responsible for parallel computing of deep learning models, such as training of a physical information constrained reconstruction model and a leakage prediction model.

[0115] The application provides a leakage diagnosis method of a supercritical water reactor. The method comprises the following steps: acquiring an acoustic time domain signal in a to-be-tested region of the supercritical water reactor and physical parameter information in the to-be-tested region that is synchronous with the acoustic time domain signal, and determining an original time-frequency matrix corresponding to the acoustic time domain signal; inputting the original time-frequency matrix into a reconstruction model with physical information constraint to obtain a reconstruction time-frequency matrix and physical parameter prediction information output by the reconstruction model; the reconstruction model has learned the ability to reconstruct a normal acoustic signal and the ability to predict a physical parameter based on the normal acoustic signal; determining whether the acoustic time domain signal is abnormal based on the reconstruction time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information and the physical parameter information; if the acoustic time domain signal is abnormal, determining whether there is leakage in the to-be-tested region based on the original time-frequency matrix and the physical parameter information; and if there is leakage in the to-be-tested region, determining positioning information of a leakage source in the to-be-tested region based on the acoustic time domain signal and Fermat's principle. The method combines acoustic data and physical parameter information, realizes a complete closed loop from three stages of abnormality identification, leakage diagnosis and leakage source positioning, realizes leakage diagnosis of the supercritical water reactor, and improves the accuracy of leakage diagnosis.

[0116] Next, the process of determining the positioning information of the leakage source in the to-be-tested region based on the acoustic time domain signal and Fermat's principle will be described in detail.

[0117] Figure 2 Another flowchart of a leakage diagnosis method of a supercritical water reactor is provided for the embodiments of the application. In some embodiments of the application, the acoustic time domain signal is acquired based on an acoustic emission sensor array, and the acoustic emission sensor array comprises N acoustic emission sensors, wherein N is an integer greater than 1. As shown in the figure, the implementation process of step 105 in the above embodiment can comprise the following steps: Figure 2 Figure 1

[0118] Step 201: determining a non-uniform sound speed field in the to-be-tested region.

[0119] Since the sound speed is different at different positions in the supercritical water reactor, the sound speed at each position needs to be determined for subsequent positioning calculation of the leakage source. The non-uniform sound speed field in the to-be-tested region refers to the sound speed distribution in the three-dimensional space of the to-be-tested region.

[0120] In some embodiments, the process of determining the non-uniform sound speed field in the to-be-tested region comprises: determining a three-dimensional temperature field and a three-dimensional pressure field of the to-be-tested region through fluid dynamics simulation or finite element interpolation algorithm; and constructing the non-uniform sound speed field based on the three-dimensional temperature field and the three-dimensional pressure field.

[0121] ​​Among them, the three-dimensional temperature field refers to the temperature distribution field at different spatial locations in the area to be measured, and the three-dimensional pressure field refers to the pressure distribution field at different spatial locations in the area to be measured.

[0122] As an example, the physical parameter information obtained in step 101 includes the temperature collected by temperature sensors at different locations and the pressure collected by pressure sensors at different locations. Based on the spatial location of the sensors, the temperature and pressure at different spatial locations can be determined. Based on the temperature and pressure at different spatial locations, fluid dynamics simulation or finite element interpolation is performed to obtain the three-dimensional temperature field and three-dimensional pressure field of the area to be measured.

[0123] As one possible implementation, a non-uniform sound velocity field can be constructed based on a three-dimensional temperature field and a three-dimensional pressure field. This can be achieved by calling the IAPWS-IF97 standard library to calculate the sound velocity at the corresponding spatial location based on the temperature and pressure at different spatial locations. As an example, it is shown in the following equation (8):

[0124] (8);

[0125] in, , which is the three-dimensional spatial coordinate vector of the area to be measured; Spatial location The speed of sound at that location; It is a three-dimensional temperature field, i.e., spatial location. Temperature at that location; Spatial location Pressure at the location; This is a function that calls the IAPWS-IF97 standard library to calculate the speed of sound.

[0126] Step 202: Based on the cross-correlation algorithm, determine the actual time delay difference between the acoustic time-domain signals collected by any two acoustic emission sensors among the N acoustic emission sensors.

[0127] In some embodiments, the process of determining the actual time delay difference between the acoustic time-domain signals acquired by any two acoustic emission sensors among N acoustic emission sensors, based on the cross-correlation algorithm, includes: for the first acoustic time-domain signal of any one of the N acoustic emission sensors, shifting the acoustic time-domain signal on the time axis by t (t is the duration, which can be preset), and then calculating the integral of the product of the shifted acoustic time-domain signal with the other N-1 remaining second acoustic time-domain signals. This integral is the cross-correlation function. For each cross-correlation function, the cross-correlation function will reach a peak at a certain time. The time when the peak is reached is the optimal matching delay time between the first acoustic time-domain signal and the corresponding second acoustic time-domain signal, i.e., the actual time delay difference. By performing the above process once for the acoustic time-domain signals corresponding to all sensors, the actual time delay difference between the acoustic time-domain signals of any two sensors can be obtained.

[0128] Step 203: Taking any point in the area to be measured as the source point, based on the non-uniform sound velocity field and the sound wave transmission path from the source point to each acoustic emission sensor, determine the theoretical time delay difference between the sound waves reaching each acoustic emission sensor.

[0129] In some embodiments, any point within the area to be measured As a source point, based on the non-uniform sound velocity field and the sound wave transmission path from the source point to each acoustic emission sensor, the theoretical time delay difference between the sound waves arriving at each acoustic emission sensor is determined by: determining all possible paths for the sound waves to travel from the source point to each acoustic emission sensor; for each acoustic emission sensor, determining the duration of the shortest path among all possible paths; and obtaining the theoretical time delay difference between the sound waves arriving at each acoustic emission sensor based on the duration of the sound waves traveling from the source point to each acoustic emission sensor along the shortest path.

[0130] As an example, step 203 can be achieved by the following equation (9):

[0131] (9);

[0132] in, Source point; This refers to the point from the source. The theoretical time delay difference between the sound wave signal propagating outward and the sound wave signal reaching the i-th acoustic emission sensor and the sound wave signal reaching the j-th acoustic emission sensor; To start from the source up to the i-th acoustic emission sensor The set of all possible paths between them; To start from the source Up to the j-th acoustic emission sensor The set of all possible paths between them; For sound waves from the source point up to the i-th acoustic emission sensor The propagation along the shortest path takes time; For sound waves from the source point Up to the j-th acoustic emission sensor The propagation process using the shortest time path requires time.

[0133] Step 204: Based on the difference between the theoretical delay difference and the actual delay difference, continuously update the location of the source point until the difference meets the preset conditions, and determine the latest source point location information as the location information of the leakage source.

[0134] Among them, the preset conditions can be threshold conditions or optimization target conditions, that is, the difference between theoretical delay difference and actual delay difference is minimized.

[0135] In some embodiments, all actual time delay differences between the acoustic time-domain signals acquired by the two acoustic emission sensors can be used to form an actual time delay difference vector; all theoretical time delay differences corresponding to the current source point can be used to form a theoretical time delay difference vector; the deviation between the actual time delay difference vector and the theoretical time delay difference vector can be calculated, and the source point corresponding to the smallest deviation can be determined as the location information of the leakage source.

[0136] As an example, step 204 can be achieved by solving the objective function of equation (10) as follows:

[0137] (10);

[0138] in, This provides information on the location of the leak source. This is the actual time delay difference vector; As the source The corresponding theoretical time delay difference vector; for and Mahalanobis distance between them; Σ is the inverse of the covariance matrix. The covariance matrix Σ is a diagonal matrix representing the reliability of the actual time delay difference between the acoustic time-domain signals acquired by the two acoustic emission sensors. The diagonal elements of the covariance matrix are dynamically calculated based on the signal quality using a cross-correlation algorithm. If the normalized cross-correlation peak of the two signals is high (good signal-to-noise ratio), then the corresponding variance is small (high reliability); if the peak is low (high noise), the variance is large (low reliability). The diagonal elements of the covariance matrix are the variances calculated using the cross-correlation algorithm. In equation (10), It plays a role of automatic weighting, reuses credible measurement data, and ignores noise data that is not credible.

[0139] According to the leakage diagnosis method of the supercritical water reactor, by constructing a non-uniform sound speed field, based on the path optimization algorithm of Fermat principle, the actual time delay difference is compared with the theoretical time delay difference, the source point in the to-be-measured region is continuously updated until the difference between the actual time delay difference and the theoretical time delay difference meets a preset condition, and the position information of the latest source point is determined as the positioning information of the leakage source, so that accurate positioning of the leakage source can be realized.

[0140] In order to realize the above-mentioned embodiments, the application further provides a leakage diagnosis device of a supercritical water reactor.

[0141] Figure 3 A structural schematic diagram of a leakage diagnosis device of a supercritical water reactor provided by an embodiment of the application is shown in the figure. Figure 3 As shown in the figure, the device can include an acquisition module 310, a reconstruction module 320, a determination module 330, a diagnosis module 340 and a positioning module 350.

[0142] The acquisition module 310 is configured to acquire an acoustic time domain signal in a to-be-measured region of a supercritical water reactor, and physical parameter information in the to-be-measured region that is synchronized with the acoustic time domain signal, and determine an original time-frequency matrix corresponding to the acoustic time domain signal.

[0143] The reconstruction module 320 is configured to input the original time-frequency matrix into a reconstruction model constrained by physical information, and obtain a reconstructed time-frequency matrix output by the reconstruction model and physical parameter prediction information.

[0144] The determination module 330 is configured to determine whether the acoustic time domain signal is abnormal based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information and the physical parameter information.

[0145] The diagnosis module 340 is configured to, if the acoustic time domain signal is abnormal, determine whether there is a leakage in the to-be-measured region based on the original time-frequency matrix and the physical parameter information.

[0146] The positioning module 350 is configured to, if there is a leakage in the to-be-measured region, determine positioning information of a leakage source in the to-be-measured region based on the acoustic time domain signal and Fermat principle.

[0147] In some embodiments, the acquisition module 310 is specifically configured to:

[0148] Perform continuous wavelet transform or short-time Fourier transform on the acoustic time domain signal to obtain the original time-frequency matrix.

[0149] In some embodiments, the determining module 330 is specifically configured to:

[0150] determine a first anomaly score between the original time-frequency matrix and the reconstructed time-frequency matrix; the first anomaly score is used to represent a difference degree between the original time-frequency matrix and the reconstructed time-frequency matrix;

[0151] determine a second anomaly score between the physical parameter information and the physical parameter prediction information; the second anomaly score is used to represent a difference degree between the physical parameter information and the physical parameter prediction information;

[0152] determine a comprehensive anomaly score based on the first anomaly score and the second anomaly score;

[0153] if the comprehensive anomaly score is greater than or equal to a threshold value, determine that the acoustic time-domain signal is abnormal.

[0154] As a possible implementation manner, the determining module 330 is further configured to:

[0155] perform a weighted sum operation on the first anomaly score and the second anomaly score, and determine an operation result as the comprehensive anomaly score.

[0156] In some embodiments, the diagnosing module 340 is specifically configured to:

[0157] determine an acoustic input tensor based on the original time-frequency matrix and the physical parameter information;

[0158] input the acoustic input tensor into a leakage prediction model to obtain a leakage prediction result output by the leakage prediction model; the leakage prediction model is trained based on training samples including acoustic input tensor samples and corresponding leakage labels;

[0159] determine whether there is leakage in the to-be-tested region based on the leakage prediction result.

[0160] In some embodiments, the physical parameter information includes temperature information and pressure information; the diagnosing module 340 is further configured to:

[0161] construct a temperature matrix and a pressure matrix with the same dimension as the original time-frequency matrix based on the temperature information and the pressure information, respectively;

[0162] determine a density gradient matrix based on the temperature information and the pressure information; the density gradient matrix has the same dimension as the original time-frequency matrix;

[0163] based on the original time-frequency matrix, the temperature matrix, the pressure matrix and the density gradient matrix, perform splicing along a channel dimension to obtain the acoustic input tensor.

[0164] In some embodiments, the acoustic time-domain signal is acquired based on an acoustic emission sensor array; the acoustic emission sensor array includes N acoustic emission sensors, where N is an integer greater than 1; the positioning module 350 is specifically used for:

[0165] Determine the non-uniform sound velocity field within the region to be measured;

[0166] Based on the cross-correlation algorithm, the actual time delay difference between the acoustic time domain signals collected by any two acoustic emission sensors among N acoustic emission sensors is determined.

[0167] Taking any point in the area to be measured as the source point, based on the non-uniform sound velocity field and the sound wave transmission path from the source point to each acoustic emission sensor, the theoretical time delay difference between the sound waves arriving at each acoustic emission sensor is determined.

[0168] Based on the difference between the theoretical delay difference and the actual delay difference, the location of the source point is continuously updated until the difference meets the preset conditions, and the latest source point location information is determined as the location information of the leakage source.

[0169] The positioning module 350 is also used for:

[0170] The three-dimensional temperature field and three-dimensional pressure field of the region to be measured are determined by fluid dynamics simulation or finite element interpolation algorithm;

[0171] A non-uniform sound velocity field is constructed based on a three-dimensional temperature field and a three-dimensional pressure field.

[0172] According to the embodiment of the present invention, the leakage diagnosis device for supercritical water reactors combines acoustic data and physical parameter information to realize a complete closed loop from three stages: anomaly identification, leakage diagnosis, and leakage source location. This enables leakage diagnosis of supercritical water reactors and improves the accuracy of leakage diagnosis.

[0173] It should be noted that the explanations and descriptions in the above embodiments regarding the leakage diagnosis method for supercritical water reactors can also be applied to leakage diagnosis devices for supercritical water reactors, and will not be repeated here.

[0174] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include a processor 410, a communication interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communication interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call the computer program in the memory 430 to execute the steps of the leakage diagnosis method for supercritical water reactors provided in the above embodiments.

[0175] Moreover, the logic instructions in the memory 430 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0176] In another aspect, the embodiments of the present application also provide a computer program product, which includes a computer program that can be stored on a computer readable storage medium, and when the computer program is executed by a processor, a computer can execute the steps of the leakage diagnosis method of the supercritical water reactor provided by the above embodiments.

[0177] In another aspect, the embodiments of the present application also provide a non-transitory computer readable storage medium, which stores a computer program for causing a processor to execute the leakage diagnosis method of the supercritical water reactor provided by the above embodiments.

[0178] The non-transitory computer readable storage medium can be any available medium or data storage device that the processor can access, including but not limited to a magnetic storage (such as a floppy disk, a hard disk, a magnetic tape, a magneto-optical disk (MO), etc.), an optical storage (such as a CD, a DVD, a BD, a HVD, etc.), and a semiconductor memory (such as a ROM, an EPROM, an EEPROM, a NAND FLASH, a solid state disk (SSD)), etc.

[0179] The device embodiments described above are only schematic, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0180] Those skilled in the art can clearly understand the technical solutions of the embodiments from the above description of the embodiments, and the embodiments can be implemented by means of software with the necessary general hardware platform, and of course, can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in the sense of contribution to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, or an optical disc, and includes a plurality of instructions to cause a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the methods.

[0181] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for diagnosing leaks in a supercritical water reactor, characterized in that, include: Acquire the acoustic time-domain signal within the test area of ​​the supercritical water reactor, as well as the physical parameter information within the test area that is synchronized with the acoustic time-domain signal, and determine the original time-frequency matrix corresponding to the acoustic time-domain signal; The original time-frequency matrix is ​​input into the reconstruction model constrained by physical information to obtain the reconstructed time-frequency matrix and physical parameter prediction information output by the reconstruction model; the reconstruction model has learned the ability to reconstruct normal acoustic signals and the ability to predict physical parameters based on normal acoustic signals. Based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information, and the physical parameter information, it is determined whether there is an anomaly in the acoustic time-domain signal; If the acoustic time-domain signal is abnormal, based on the original time-frequency matrix and the physical parameter information, determine whether there is a leak in the area to be tested; If a leak exists in the area to be tested, the location information of the leak source in the area to be tested is determined based on the acoustic time-domain signal and Fermat's principle. The step of determining whether a leak exists in the area to be tested based on the original time-frequency matrix and the physical parameter information includes: Based on the original time-frequency matrix and the physical parameter information, the acoustic input tensor is determined; The acoustic input tensor is input into the leakage prediction model to obtain the leakage prediction result output by the leakage prediction model; the leakage prediction model is trained based on training samples including acoustic input tensor samples and their corresponding leakage labels; Based on the leakage prediction results, it is determined whether a leakage exists in the area to be tested; The physical parameter information includes temperature information and pressure information; determining the acoustic input tensor based on the original time-frequency matrix and the physical parameter information includes: Based on the temperature information and the pressure information, a temperature matrix and a pressure matrix with the same dimensions as the original time-frequency matrix are constructed respectively. Based on the temperature information and the pressure information, a density gradient matrix is ​​determined; the dimension of the density gradient matrix is ​​the same as the dimension of the original time-frequency matrix. Based on the original time-frequency matrix, the temperature matrix, the pressure matrix, and the density gradient matrix, the acoustic input tensor is obtained by splicing them along the channel dimension. The acoustic time-domain signal is acquired based on an acoustic emission sensor array; the acoustic emission sensor array includes N acoustic emission sensors, where N is an integer greater than 1; the determination of the location information of the leakage source in the area to be measured based on the acoustic time-domain signal and Fermat's principle includes: Determine the non-uniform sound velocity field within the region to be measured; Based on the cross-correlation algorithm, the actual time delay difference between the acoustic time-domain signals collected by any two acoustic emission sensors among the N acoustic emission sensors is determined. Taking any point in the area to be measured as the source point, based on the non-uniform sound velocity field and the sound wave transmission path from the source point to each of the acoustic emission sensors, the theoretical time delay difference between the sound waves arriving at each of the acoustic emission sensors is determined. Based on the difference between the theoretical delay difference and the actual delay difference, the location of the source point is continuously updated until the difference meets the preset conditions, and the latest source point location information is determined as the location information of the leakage source.

2. The method according to claim 1, characterized in that, Determining the original time-frequency matrix corresponding to the acoustic time-domain signal includes: The original time-frequency matrix is ​​obtained by performing continuous wavelet transform or short-time Fourier transform on the acoustic time-domain signal.

3. The method according to claim 1, characterized in that, The step of determining whether there is an anomaly in the acoustic time-domain signal based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information, and the physical parameter information includes: A first anomaly score is determined between the original time-frequency matrix and the reconstructed time-frequency matrix; the first anomaly score is used to characterize the degree of difference between the original time-frequency matrix and the reconstructed time-frequency matrix; A second anomaly score is determined between the physical parameter information and the physical parameter prediction information; the second anomaly score is used to characterize the degree of difference between the physical parameter information and the physical parameter prediction information. Based on the first anomaly score and the second anomaly score, a comprehensive anomaly score is determined; If the comprehensive anomaly score is greater than or equal to the threshold, it is determined that the acoustic time-domain signal is abnormal.

4. The method according to claim 3, characterized in that, The determination of the comprehensive anomaly score based on the first anomaly score and the second anomaly score includes: The first abnormal score and the second abnormal score are weighted and summed, and the result is determined as the comprehensive abnormal score.

5. The method according to claim 1, characterized in that, Determining the non-uniform sound velocity field within the region to be measured includes: The three-dimensional temperature field and three-dimensional pressure field of the region to be measured are determined by fluid dynamics simulation or finite element interpolation algorithm; Based on the three-dimensional temperature field and the three-dimensional pressure field, the non-uniform sound velocity field is constructed.

6. A leakage diagnosis device for a supercritical water reactor, characterized in that, include: The acquisition module is used to acquire the acoustic time-domain signal in the test area of ​​the supercritical water reactor, as well as the physical parameter information of the test area that is synchronized with the acoustic time-domain signal, and to determine the original time-frequency matrix corresponding to the acoustic time-domain signal. The reconstruction module is used to input the original time-frequency matrix into the reconstruction model constrained by physical information, and obtain the reconstructed time-frequency matrix and physical parameter prediction information output by the reconstruction model; the reconstruction model has learned the ability to reconstruct normal acoustic signals and the ability to predict physical parameters based on normal acoustic signals; The determination module is used to determine whether there is an anomaly in the acoustic time-domain signal based on the reconstructed time-frequency matrix, the original time-frequency matrix, the physical parameter prediction information, and the physical parameter information; The diagnostic module is used to determine whether there is a leak in the area to be tested, based on the original time-frequency matrix and the physical parameter information, if there is an anomaly in the acoustic time-domain signal. The positioning module is used to determine the positioning information of the leakage source in the area to be tested based on the acoustic time-domain signal and Fermat's principle if a leakage exists in the area to be tested. The step of determining whether a leak exists in the area to be tested based on the original time-frequency matrix and the physical parameter information includes: Based on the original time-frequency matrix and the physical parameter information, the acoustic input tensor is determined; The acoustic input tensor is input into the leakage prediction model to obtain the leakage prediction result output by the leakage prediction model; the leakage prediction model is trained based on training samples including acoustic input tensor samples and their corresponding leakage labels; Based on the leakage prediction results, it is determined whether a leakage exists in the area to be tested; The physical parameter information includes temperature information and pressure information; determining the acoustic input tensor based on the original time-frequency matrix and the physical parameter information includes: Based on the temperature information and the pressure information, a temperature matrix and a pressure matrix with the same dimensions as the original time-frequency matrix are constructed respectively. Based on the temperature information and the pressure information, a density gradient matrix is ​​determined; the dimension of the density gradient matrix is ​​the same as the dimension of the original time-frequency matrix. Based on the original time-frequency matrix, the temperature matrix, the pressure matrix, and the density gradient matrix, the acoustic input tensor is obtained by splicing them along the channel dimension. The acoustic time-domain signal is acquired based on an acoustic emission sensor array; the acoustic emission sensor array includes N acoustic emission sensors, where N is an integer greater than 1; the determination of the location information of the leakage source in the area to be measured based on the acoustic time-domain signal and Fermat's principle includes: Determine the non-uniform sound velocity field within the region to be measured; Based on the cross-correlation algorithm, the actual time delay difference between the acoustic time-domain signals collected by any two acoustic emission sensors among the N acoustic emission sensors is determined. Taking any point in the area to be measured as the source point, based on the non-uniform sound velocity field and the sound wave transmission path from the source point to each of the acoustic emission sensors, the theoretical time delay difference between the sound waves arriving at each of the acoustic emission sensors is determined. Based on the difference between the theoretical delay difference and the actual delay difference, the location of the source point is continuously updated until the difference meets the preset conditions, and the latest source point location information is determined as the location information of the leakage source.

7. An electronic device, characterized in that, It includes a processor and a memory storing a computer program, wherein the processor executes the program to implement the leakage diagnosis method for a supercritical water reactor as described in any one of claims 1-5.

Images