CHF prediction model construction method, medium and electronic equipment

Abstract

The invention provides a CHF prediction model construction method, a medium and electronic equipment, and relates to the technical field of data processing, and the method comprises the steps: carrying out the edge working condition data recognition of a CHF test sample set, and screening an edge test sample set from the CHF test sample set; and calculating a prior CHF value corresponding to each edge grid by using the CHF prediction mechanism model in the edge grid region, constructing a virtual sample according to the working condition parameter combination corresponding to each edge grid and the prior CHF value, taking the virtual sample matched with the real CHF distribution as a target virtual sample, and constructing a CHF prediction model according to the target virtual sample and the CHF test sample set. On one hand, the edge working condition data is amplified under the physical constraint condition of the edge grid region, so that the amplified virtual sample better conforms to the physical law of the edge data. And on the other hand, the CHF prediction model is constructed based on the amplified virtual sample, so that the extrapolation capability and the prediction credibility of the model can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to a CHF prediction model construction method, a computer readable medium and an electronic device. BACKGROUND

[0002] Critical heat flux (CHF) is a key safety parameter in the operation of thermal equipment. When the heat flux density of the heating surface of the equipment reaches the CHF value, the heat exchange mechanism between the heating surface and the working medium will change suddenly, which can easily cause wall overheating, equipment burning and other safety accidents. Therefore, constructing a high-precision CHF prediction model to accurately predict the CHF value under different working conditions is a core requirement for ensuring the safe and stable operation of thermal equipment. SUMMARY

[0003] Therefore, the present application provides a CHF prediction model construction method, a computer readable medium and an electronic device, which can improve the extrapolation ability and prediction credibility of the CHF prediction model.

[0004] In a first aspect, the present application provides a CHF prediction model construction method, comprising: obtaining a CHF test sample set containing a plurality of CHF test samples, wherein each CHF test sample contains a working condition parameter combination and a CHF test value corresponding to the working condition parameter combination; performing edge working condition data identification on the CHF test sample set to filter out an edge test sample set from the CHF test sample set; determining an edge grid region containing a plurality of edge grids according to the working condition parameter combination corresponding to each edge test sample in the edge test sample set; calculating a prior CHF value corresponding to each edge grid in the edge grid region by using a CHF prediction mechanism model, and constructing a virtual sample according to the working condition parameter combination corresponding to each edge grid and the prior CHF value; taking a virtual sample matching a real CHF distribution as a target virtual sample, and constructing a CHF prediction model according to the target virtual sample and the CHF test sample set.

[0005] In a second aspect, the present application provides an electronic device, comprising: at least one processor; and at least one memory having instructions stored thereon, which, when executed by the at least one processor alone or jointly, cause the electronic device to perform the method of the first aspect.

[0006] In a third aspect, the present application provides a computer readable medium storing computer program codes, which, when executed by a processor, implement the method according to the first aspect.

[0007] The present application provides a CHF prediction model construction method, comprising: identifying edge condition data from a CHF test sample set, and screening an edge test sample set from the CHF test sample set; calculating a prior CHF value corresponding to each edge grid in an edge grid region by using a CHF prediction mechanism model, and constructing a virtual sample according to a working condition parameter combination and the prior CHF value corresponding to each edge grid, taking a virtual sample matching a real CHF distribution as a target virtual sample, and constructing a CHF prediction model according to the target virtual sample and the CHF test sample set. On the one hand, the virtual sample is obtained by expanding the edge condition data under the physical constraint condition of the edge grid region, so that the expanded virtual sample is more in line with the physical law of the edge data and will not be blindly expanded to other non-edge condition regions. On the other hand, the CHF prediction model is constructed based on the expanded virtual sample, which can improve the extrapolation ability and prediction reliability of the model. BRIEF DESCRIPTION OF DRAWINGS

[0008] The accompanying drawings are included to provide a further understanding of the present application, and are incorporated in and constitute a part of this application, illustrate embodiments of the present application, and together with the description serve to explain the principles of the present application. In the drawings: Figure 1 is a flowchart of a CHF prediction model construction method provided by an embodiment of the present application; Figure 2 is a flowchart of a CHF prediction model construction method provided by an embodiment of the present application; Figure 3 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some examples or embodiments of the present application, and for those skilled in the art, the present application can also be applied to other similar scenarios without creative labor. Unless it is obvious from the language environment or otherwise stated, the same reference numbers in the drawings represent the same structure or operation.

[0010] As used in this application, the terms "comprise", "comprising", "include", "including", "contain", "containing", "have", "having", "may" and "may have" shall be construed as non- limiting terms, and thus the interpretation "comprises at least" should be taken into account. As used in this application, the terms "one", "a", "an", and / or "the" shall not be construed as specific terms, and can include plural referents unless the context clearly dictates otherwise. In general, the terms "include", "including", "contain", and / or "containing" shall not be construed to mean an exhaustive listing of the steps and elements included in the process or device.

[0011] Also, the present application uses specific terminology in describing the embodiments of the present application. As used in this application, the terms "one embodiment", "an embodiment", and / or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Therefore, it is emphasized and should be appreciated that a described "embodiment" or "one embodiment" or "an alternative embodiment" is not necessarily an alternative (i.e., mutually exclusive) to some other "embodiment" previously described, but can be included in one or more embodiments.

[0012] The relative arrangement of parts and steps, numerical expressions, and numerical values set forth in the examples are not intended to limit the scope of the application unless otherwise specifically stated. It is to be understood that the drawings are not necessarily to scale as the dimensions of the parts shown are for the clarity of presentation and are shown exaggerated in relation to each other. The techniques, methods, and devices known to those of ordinary skill in the relevant art can not be discussed in detail but should be considered as if the discussion were fully set forth herein. In the examples shown and discussed herein, any specific values should be interpreted as merely illustrative and not as a limitation upon the scope of the exemplary embodiments. Thus, other examples of the exemplary embodiments can have different values. It is noted that like numbers and letters on the figures identify like parts throughout the disclosure, and thus, once an item is defined in one figure, it is not necessary to discuss it further in connection with other figures.

[0013] Further, while the terms used in the present specification are selected from generally known and used terms, some of the terms mentioned in the description of the present application can be created by the applicant in his or her own judgment, and the detailed meaning of these terms can be found in the description of relevant parts of the present specification. It is required that the present application is understood, not by the actual terms used in the description, but by the meaning of each term under the circumstances.

[0014] Flow diagrams have been used herein to illustrate the operation of apparatus or devices in accordance with embodiments of the present application. It should be understood that the operations need not necessarily be performed in the precise order shown. Rather, various steps can be handled in reverse order, or at the same time, or other operations can be added or removed from the processes.

[0015] As described above, it is a core requirement to build a high-precision CHF prediction model to realize accurate prediction of CHF values under different working conditions and to ensure the safe and stable operation of thermal equipment. At present, the known CHF prediction model construction methods all have the problem of poor extrapolation ability and unreliable prediction due to the scarcity of edge condition data.

[0016] To expand the edge condition data used for CHF prediction model construction and overcome the problem of poor extrapolation ability and unreliable prediction of known CHF prediction model construction methods due to the scarcity of edge condition data, see Figure 1 An embodiment of the present application proposes a CHF prediction model construction method based on which a CHF prediction model with wide coverage and high prediction accuracy can be obtained, including the following processes: S10: Obtain a CHF test sample set containing a plurality of CHF test samples, and each CHF test sample contains a working condition parameter combination and a CHF test value corresponding to the working condition parameter combination.

[0017] In some embodiments, the above-mentioned working condition parameter combination can be composed of a plurality of working condition parameters, such as pressure P, mass flow rate G, and steam quality x, etc. For example, the test working condition covers a multi-working condition environment with a pressure of 0.5-15 MPa, a mass flow rate of 200-1600 kg / (m²·s), and a local steam quality of 0-1. For each working condition, the pressure P at the outlet of the rod bundle CHF test piece G , the mass flow rate at the CHF occurrence position x , and the local steam quality at the CHF occurrence position , etc. are synchronously obtained, and the CHF test value corresponding to the set of working condition parameters is recorded. The records are arranged as a structured data matrix of N rows and M columns, and each row represents a CHF test sample: D raw . Taking pressure P , mass flow rate G , and local steam quality x as the plurality of working condition parameters and M=4 as an example, the CHF test sample set D raw may be as follows: .

[0018] S11: Identify edge condition data from the CHF test sample set to filter out an edge test sample set from the CHF test sample set.

[0019] In some embodiments, step S11 includes: Step 1: Perform at least one of the following edge condition data identification processes on the CHF test sample set: sparse grid identification, physical mechanism-based edge condition identification, and isolated sample identification.

[0020] As a feasible approach, the above combination of operating parameters involves K operating parameters (such as pressure). P Mass flow rate G and local vapor content x The processing results of sparse mesh identification include candidate sparse sets corresponding to the CHF test sample set. The above-described process for sparse grid identification of the CHF test sample set includes: uniformly dividing the K-dimensional parameter space composed of K operating parameters into multiple grids; determining the grid j to which each CHF test sample in the CHF test sample set belongs; and further, counting the number of sample points contained in each grid j. n j , will satisfy n j Grids j with fewer points than the threshold are defined as sparse grids, and the data points in these sparse grids form a candidate sparse set corresponding to the CHF test sample set. The aforementioned grids are divided based on the partitioning interval and the number of partitions for each of the K operating parameters. The partitioning interval can be flexibly set and adjusted according to the number of partitions and accuracy requirements.

[0021] For example, suppose the K operating parameters are pressure P Mass flow rate G and vapor content x Dividing the grid can reduce pressure. P Mass flow rate G Steam content x The resulting three-dimensional parameter space is uniformly divided into multiple grids. Specifically, the three-dimensional parameter space ( P , G , x The pressure is uniformly divided into a regular grid according to the division interval and the number of division intervals corresponding to each working condition parameter: P Dimensions: Divided into 10 intervals with 1.5 MPa as the dividing point; mass flow rate ( G Dimensions: Divided into 7 intervals with a spacing of 200 kg / (m²·s); vapor content ( x Dimensions: Divided into 5 intervals with a 0.2 increment. Total number of grids: 10 × 7 × 5 = 350 grids.

[0022] Furthermore, perform grid density statistics: for each grid j Count the number of sample points it contains. n j :

[0023] in, I (·) is an indicator function, when CHF test samples i All operating parameters fall within the grid j If the value is within the specified range, take 1; otherwise, take 0. and They are grids j Minimum and maximum values ​​covered under the pressure dimension. and They are grids j Minimum and maximum values ​​covered in the mass flow rate dimension. and They are grids j Minimum and maximum values ​​covered under the vapor content dimension.

[0024] Furthermore, perform sparse mesh identification: set a density threshold. n th =5 (i.e., the above point quantity threshold = 5), and all those that satisfy... n j A grid with a value less than 5 is labeled as a sparse grid, containing data points... (Corresponding CHF test sample) i Construct candidate sparse sets: .

[0025] This edge condition data identification method identifies sparse regions of data from the perspective of the entire parameter space of the CHF test sample set, rather than focusing on a single sample, thus providing a global distribution perspective for the entire edge condition data identification process. Furthermore, the grid method has low computational complexity, making it suitable for large-scale datasets and enabling efficient screening of edge condition data. In addition, the grid interval can be flexibly set according to the data volume and accuracy requirements, giving this identification method adjustable multi-resolution capabilities.

[0026] As a feasible approach, the processing results of the aforementioned edge condition identification include a physical edge test sample set. The CHF test sample set is subjected to edge condition identification based on physical mechanism, including: identifying CHF test samples in the CHF test sample set that meet the physical mechanism conditions as physical edge test samples, forming a physical edge test sample set containing at least one physical edge test sample, wherein the physical mechanism conditions are associated with one or more of K operating condition parameters.

[0027] For example, suppose the K operating parameters are pressure P Mass flow rate G and vapor content xBased on the knowledge of thermal hydraulic physics, the physical mechanism conditions are pre-set to include the conditions of the edge operating zone to which each operating parameter belongs, as follows: High pressure near-critical region: pressure P > 0.8 × P critical P critical This is the critical pressure value.

[0028] Extremely low mass flow rate region: The region with mass flow rate G < 400 kg / (m²·s) is characterized by unstable flow and complex CHF mechanism.

[0029] Extremely high / extremely low vapor content region: inlet subcooling (corresponding to) x Approaching -0.2) or extremely high dryness at export ( x The >0.8) region corresponds to deviations from the bubbling boiling and drying mechanisms, respectively.

[0030] Any CHF test sample i in the CHF test sample set that falls into any of the three regions—the high-pressure near-critical region, the high-pressure near-critical region, and the extremely low mass flow rate region—is marked as a physical edge test sample. This process is repeated for each CHF test sample in the CHF test sample set to identify all physical edge test samples and construct a physical edge test sample set. .

[0031] As a feasible approach, the processing results of edge condition identification include isolated sample sets. The above-mentioned isolated sample identification of the CHF test sample set includes: The random forest algorithm is used to calculate the anomaly score of each CHF test sample in the CHF test sample set. CHF test samples with anomaly scores greater than or equal to the anomaly score threshold are identified as isolated samples, forming an isolated sample set containing at least one of these isolated samples. .

[0032] Traditional outlier detection methods (such as the 3σ criterion) assume that the data follows a normal distribution and only consider univariate variables, which cannot effectively handle complex anomaly patterns in multivariate coupling relationships. The Isolation Forest algorithm, by constructing a randomized split tree, can efficiently identify samples located in sparse positions in the multidimensional parameter space. These samples usually correspond to abnormal data caused by measurement errors, equipment failures, or real edge case data under extreme operating conditions. Identifying them as edge case data (i.e., the aforementioned isolated samples) avoids the interference of outliers on the subsequent construction of the CHF prediction model, while retaining important extreme case information, providing targets for subsequent targeted data augmentation.

[0033] Step 2: Merge the processing results of each edge condition data identification process to obtain the edge test sample set.

[0034] For example, taking edge condition data identification processing as an example, there are three processing methods: sparse mesh identification, physical mechanism-based edge condition identification, and isolated sample identification. The processing result of each method is a candidate sparse set. Physical edge test sample set and isolated sample sets Then the final edge trial sample set = ∪ ∪ This method, which integrates three criteria—local anomaly identification (isolated sample identification), global distribution identification (sparse grid identification), and physical mechanism identification (edge ​​condition identification based on physical mechanism)—ensures the comprehensiveness and robustness of edge condition identification, overcoming the critical deficiency of traditional methods that rely solely on statistical outlier testing and miss physically important edge regions. Furthermore, the execution order of the three processing methods is not critical; they can be performed in parallel or sequentially, and this application does not impose any limitations on this.

[0035] It should be noted that the edge condition data determined by sparse grid identification and isolated sample identification ( The edge condition data can be understood as statistically significant edge condition data. However, the edge condition data identified by these two methods is not entirely equivalent to the physical "edge." For example, even if some data exists in certain operating areas, it may be difficult to predict due to their unique physical conditions. Therefore, in some embodiments, when one or both of sparse grid identification and isolated sample identification in step 1 are combined with edge condition identification based on physical mechanisms, the final identified edge test sample set can take into account both statistical significance and actual physical significance, resulting in more comprehensive edge data.

[0036] Furthermore, as mentioned above, isolated sample identification focuses on anomaly analysis of individual CHF test samples, while sparse grid identification focuses on the global distribution analysis of the entire CHF test sample set. Therefore, in some embodiments, when sparse grid identification and isolated sample identification are combined in step 1, the limitation of relying solely on "single sample anomaly scores" is overcome. By introducing the "global grid density" criterion, it is ensured that even samples in the isolated forest that are not marked as anomalies but are located in the overall sparse "unpopular area" can be identified, achieving comprehensive capture of edge cases.

[0037] S12: Determine the edge grid region containing multiple edge grids based on the combination of operating parameters corresponding to each edge test sample in the edge test sample set.

[0038] For example, assume that the combination of operating parameters corresponding to each edge test sample is pressure. P Mass flow rate G and local vapor content xBy combining different values ​​of the three operating parameters, and based on the operating parameter combinations corresponding to each edge test sample in the edge test sample set, pressure can be classified. P The numerical range of mass flow rate G The numerical range and local vapor content x The numerical regions of these three dimensions are used to establish a three-dimensional parameter space. The three-dimensional parameter space is then divided into uniform grids (the grids here can be understood as edge grids). This will give us the edge grid region in S12 that contains multiple edge grids.

[0039] S13: Calculate the prior CHF value for each edge grid using a CHF prediction mechanism model within the edge grid region, and construct virtual samples based on the corresponding combination of operating parameters and prior CHF values ​​for each edge grid. The CHF prediction mechanism model can be, for example, the Katto-Ohno or Bowring mechanism models, all of which are physically driven computational tools and are mature CHF mechanism models (not purely data fitting models) in the nuclear power field. Their core is based on mathematical equations derived from the "vapor-liquid two-phase flow heat transfer mechanism" (such as vapor film rupture and bubble release confinement), and has been verified by engineering practice to possess high reliability.

[0040] In some implementations, the prior CHF value for each edge grid is calculated using a CHF prediction mechanism model in the edge grid region, including: based on the parameters of each edge grid center point, such as: G = 1500 kg / (m²) Substituting P=10MPa and x=-0.2 into the core equation of the CHF prediction mechanism model, the prior CHF value corresponding to each grid is calculated. Furthermore, the paired data of "edge grid center point parameter + corresponding prior CHF value" are used as virtual samples. In this way, physically reliable virtual samples can be generated in the edge grid region based on the CHF prediction mechanism model, realizing edge data amplification. As can be seen from the above, the edge data amplification methods proposed in this application are all limited to the physical constraint of the edge grid region, making the amplified virtual samples more consistent with the physical laws of edge data and preventing blind amplification to other non-edge working conditions.

[0041] S14: Use virtual samples that match the real CHF distribution as target virtual samples, and construct a CHF prediction model based on the target virtual samples and the CHF experimental sample set.

[0042] In some embodiments, S14 involves constructing a CHF prediction model based on the target virtual sample and the CHF test sample set, including: Step 1: Obtain the initial CHF prediction model. The initial CHF prediction model is a polynomial function that uses a set of coefficients to weight the feature matrix of the input samples. The output parameters of the initial CHF prediction model are the CHF predicted values. The set of coefficients includes multiple coefficients.

[0043] In some embodiments, each element in the feature matrix is ​​associated with different operating parameters, including pressure P, mass flow rate G, and local vapor content x. The polynomial function corresponding to the initial CHF prediction model is shown below: CHF predicted value, For the characteristic matrix, For the intercept term of the model, For the coefficient set, the feature matrix can be a row vector: = including higher-order terms (such as...) ) and interactive items (such as ), corresponding It can be a column vector, represented as: .

[0044] One of them It includes n coefficients ( , , ..., The number of n is the same as the total number of elements in the feature matrix.

[0045] Step 2: Obtain the target loss function that includes the ridge regression penalty term. With minimizing the target loss function as the optimization objective, combine the target virtual sample and the CHF test sample set to update the values ​​of each coefficient in the coefficient set to optimize the initial CHF prediction model and obtain the CHF prediction model.

[0046] For example, the formula corresponding to the target loss function is as follows:

[0047] in, The goodness-of-fit term is a mean squared error function, representing the average squared error between the model's predicted CHF values ​​and the measured CHF values. It drives the model to reproduce the input-output mapping relationship implied in the data as accurately as possible. Norm penalty term: λ is the coefficient of the norm penalty term. The role of the norm penalty term is to control the complexity of the model and prevent coefficient settling. When the value of a certain coefficient becomes too large, it effectively improves the stability of the model and its resistance to overfitting.

[0048] Furthermore, all target virtual samples and each sample in the CHF experimental sample set can be substituted into the polynomial function corresponding to the initial CHF prediction model, with minimizing the target loss function as the optimization objective, and the coefficient set can be iteratively updated continuously. The values ​​of each coefficient and the intercept term The final output is the optimal set of coefficients. and intercept term The initial CHF prediction model is optimized to obtain the CHF prediction model. For example, the CHF prediction model can be a polynomial function as follows: Coefficient set ( , , ..., The optimal parameters are determined through iterative optimization of the model. Ridge regression is used to fit the polynomial features containing higher-order terms and interaction terms. The norm penalty term introduced effectively suppresses the parameter estimation variance caused by feature multicollinearity, thereby improving the model's numerical stability and robustness. This makes the model's predictions smoother and more reasonable when facing new working conditions not fully covered by training data, avoiding the extrapolation oscillation problem common in complex polynomial models. Furthermore, compared with traditional "black box" machine learning models, it has the following advantages: transforming the "black box" machine learning model into an explicit mathematical formula with clear physical meaning not only facilitates prediction but also directly reveals the coupling mechanism between multiple parameters, greatly enhancing the interpretability and engineering acceptance of the results.

[0049] In some embodiments, see Figure 2 Another embodiment of this application proposes a construction method in which... Figure 1 In addition, the following steps are also included: S20: Construct a generative adversarial network (GAN), which includes a generator and a discriminator. The generator uses the operating parameters involved in the combination of operating parameters in the virtual samples as constraints to generate simulated CHF values ​​that match the constraints. The discriminator takes the operating parameters and the simulated CHF values ​​as input and outputs a authenticity score for the simulated CHF values. For example, the GAN could be a cWGAN.

[0050] S21: Train the generative adversarial network based on the CHF experimental sample set, and optimize iteratively through adversarial interaction between the generator and the discriminator until the training converges, thus obtaining the trained generative adversarial network.

[0051] S22: Input the virtual sample into the trained generative adversarial network, output the authenticity score of the virtual sample, and determine the virtual sample with an authenticity score greater than or equal to a preset threshold as matching the real CHF distribution.

[0052] The core principle of steps S20-S22 above is: based on the distribution matching verification of "conditional constraints + adversarial game", the boundary of the generation and discrimination scenarios is limited by the working condition parameters (such as (P, G, x)), so that the generative adversarial network learns the "working condition-numerical correspondence law of real CHF samples". Finally, the trained discriminator is used to check whether the virtual sample conforms to the law. The essence is to verify whether the probability distribution of the virtual sample is consistent with the distribution of the real sample, rather than whether a single value is the same, so as to ensure that the virtual sample can be integrated into the real sample set (i.e. the CHF experimental sample set mentioned above) for subsequent training of the CHF prediction model without introducing distribution bias.

[0053] For example, the operating parameters (P, G, x) are used as constraints for the generator. The purpose of these constraints is to prevent the generative adversarial network from generating arbitrary values. For instance, it should not generate high CHF values ​​that only occur under "high pressure, high flow rate, high gas content" conditions when operating under "low pressure (P=1MPa), low flow rate (G=500), low gas content (x=-0.2)" conditions, thus ensuring that the learned CHF distribution is the true CHF distribution under specific operating conditions. The training process of the generative adversarial network includes: Step 1: Input each CHF test sample in the CHF test sample set into the generative adversarial network for training. Specifically, input the "operating parameters (P, G, x) + corresponding CHF test value" of the CHF test sample into the discriminator in pairs, telling the discriminator "this is the real CHF value". At the same time, let the generator generate fake CHF values ​​based on the same "operating parameters (P, G, x)" and input them into the discriminator, telling the discriminator "this is the fake CHF value".

[0054] Step 2: The generator continuously adjusts its parameters and learns to generate CHF values ​​that the discriminator may mistake for true. The discriminator continuously adjusts its parameters to try to accurately distinguish between "real CHF values" and "generator's fake CHF values".

[0055] Step 3: When the game reaches equilibrium (the generator can no longer fool the discriminator, and the discriminator can no longer improve the recognition accuracy), training stops. At this point, the discriminator has learned the core patterns of real CHF samples: such as "P=15MPa, G=3000kg / (m²)". 2 Under the conditions of ·s) and x=0.1”, the actual CHF value is usually between 4000-5000 kW / m², and is most likely concentrated around 4500 kW / m², which is the true distribution.

[0056] Furthermore, after the generative adversarial network (GAN) is trained, it is used to verify the virtual samples generated by the CHF prediction mechanism model. The process is as follows: the virtual sample's "operating condition parameters (P, G, x) + prior CHF value" are input as pairs into the trained discriminator. The discriminator outputs a authenticity score based on the previously learned CHF true distribution pattern. The authenticity score is between 0 and 1; the closer to 1, the more likely it is to be a real sample, and the closer to 0, the more likely it is to be a fake sample. If the authenticity score of the virtual sample is ≥ a preset threshold (e.g., 0.8), it means that the numerical distribution of the virtual sample under the corresponding operating condition is consistent with the distribution of the real experimental sample, and it is determined that it matches the real CHF distribution. Conversely, if the authenticity score is lower than the preset threshold, it means that the virtual sample violates the operating condition-numerical correspondence pattern of the real CHF and is discarded or corrected.

[0057] Understandably, the target virtual samples (i.e., the virtual samples that match the real CHF distribution) are used to construct subsequent CHF prediction models, enabling them to learn the CHF distribution patterns of edge case data. This fundamentally alleviates the problem of poor extrapolation ability and unreliable predictions caused by the scarcity of edge case data in traditional CHF prediction model construction. Therefore, the larger the amount of target virtual samples, the higher the CHF prediction accuracy of the subsequent CHF prediction model will be. Based on this, after executing step S22 to select target virtual samples from all virtual samples, the number of target virtual samples can be counted. If this number is greater than or equal to a preset threshold, the subsequent step is triggered: constructing a CHF prediction model based on all target virtual samples and the CHF experimental sample set. If the quantity is less than a preset threshold, the edge grid region in step S13 is expanded according to the expansion step size of each operating parameter involved in the edge grid region, resulting in an expanded grid region. Within the expanded grid region, the prior CHF value corresponding to each edge grid is calculated using the CHF prediction mechanism model. New virtual samples are constructed based on the combination of operating parameters and the prior CHF value corresponding to each edge grid. Step S22 is then performed on the new virtual samples, and this process continues until the total number of selected virtual samples is greater than or equal to the preset threshold. For example, assuming the operating parameters involved in the edge grid region include pressure P, mass flow rate G, and local vapor content x, the expansion step size set for each operating parameter can be: 0.5 MPa, 20 kg / (m³), etc. 2 For each amplification step, the pressure P, mass flow rate G, and local vapor content x are accumulated to the corresponding amplification step size.

[0058] This application also provides a chip, including a circuit system configured to perform the life assessment method for nuclear power materials described above. The chip includes a Field Programmable Gate Array (FPGA) chip, a Complex Programmable Logic Device (CPLD) chip, and an Application Specific Integrated Circuit (ASIC) chip, etc.

[0059] Figure 3 This is a simplified block diagram of an electronic device 300 suitable for implementing embodiments of this application. For example, the CHF prediction model construction method described above can be implemented by the electronic device 300. As shown, the electronic device 300 includes one or more processors 310, one or more memories 320 coupled to the processors 310, and one or more communication modules 340 coupled to the processors 310.

[0060] Communication module 340 is used for bidirectional communication. Communication module 340 has at least one antenna to facilitate communication. The communication interface can represent any interface necessary for communication with other network elements.

[0061] Processor 310 can be any type suitable for a local technology network, and as a non-limiting example, can include one or more of the following: general-purpose computer, special-purpose computer, microprocessor, digital signal processor (DSP), and processor based on a multi-core processor architecture. Electronic device 300 can have multiple processors, such as application-specific integrated circuit (ASIC) chips, which are timely driven to a clock that synchronizes with the main processor.

[0062] Memory 320 may include one or more non-volatile memories and one or more volatile memories. Examples of non-volatile memories include, but are not limited to, read-only memory (ROM), electrically programmable read-only memory (EPROM), flash memory, hard disk, optical disc (CD), digital video disc (DVD), and other magnetic and / or optical storage. Examples of volatile memories include, but are not limited to, random access memory (RAM) and other volatile memories that do not persist during power-off periods.

[0063] Computer program 330 includes computer-executable instructions that are executed by the associated processor 310. Program 330 may be stored in ROM 324. Processor 310 may perform any appropriate actions and processes by loading program 330 into RAM 322.

[0064] The embodiments of this application can be implemented by program 330, enabling electronic device 300 to execute the reference. Figure 1 or Figure 2 Any process disclosed in the discussion. Embodiments of this application may also be implemented by hardware or by a combination of software and hardware.

[0065] In some embodiments, program 330 may be tangibly contained in a computer-readable medium, which may be contained in an electronic device 300 (e.g., memory 320) or other storage device accessible to the electronic device 300. The electronic device 300 may load program 330 from the computer-readable medium into RAM 322 for execution. The computer-readable medium may include any type of tangible non-volatile memory, such as ROM, EPROM, flash memory, hard disk, CD, DVD, etc. Program 330 is stored on the computer-readable medium.

[0066] Generally, the various embodiments of this application can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. Some aspects may be implemented in hardware, while others may be implemented in firmware or software, which may be executed by a controller, microprocessor, or other electronic device. Although various aspects of the embodiments of this application are shown and described as block diagrams, flowcharts, or other graphical representations, it should be understood that, as non-limiting examples, the blocks, devices, systems, techniques, or methods described herein may be implemented in hardware, software, firmware, dedicated circuitry or logic, general-purpose hardware or controllers or other electronic devices, or some combination thereof.

[0067] This application also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which execute in a device on a target real or virtual processor to perform the aforementioned references. Figure 1 or Figure 2 The method described herein. Typically, a program module includes routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of a program module can be combined or separated among program modules as needed. The machine-executable instructions used in the program module can execute on a local or distributed device. In a distributed device, the program module can reside on both local and remote storage media.

[0068] The program code used to perform the methods of this application can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program code is executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented.

[0069] In the context of this application, computer program code or related data may be carried by any suitable carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, etc.

[0070] Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable media can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or apparatuses, or any suitable combination thereof. More specific examples of computer-readable storage media include electrical connections having one or more wires, portable computer floppy disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable optical disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0071] Furthermore, although the operations are described in a specific order, this should not be construed as requiring that these operations be performed in the specific order or sequence shown, or that all of the operations shown be performed to obtain the desired result. In some cases, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the foregoing discussion, these details should not be construed as limiting the scope of this application, but rather as descriptions of features specific to particular embodiments. Certain features described in the context of a single embodiment may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0072] Although this application has been described in language specific to structural features and / or methodological behavior, it should be understood that the application as defined in the appended claims is not necessarily limited to the specific features or behaviors described above. Rather, the specific features and actions described above are disclosed as exemplary forms for implementing the claims.

Claims

1. A method for constructing a CHF prediction model, characterized in that, include: Obtain a CHF test sample set containing multiple CHF test samples. A CHF test sample contains a combination of operating parameters and the CHF test value corresponding to the combination of operating parameters. Edge condition data identification is performed on the CHF test sample set to filter out edge test sample sets from the CHF test sample set; The edge grid region containing multiple edge grids is determined based on the combination of operating parameters corresponding to each edge test sample in the edge test sample set. In the edge grid region, the prior CHF value corresponding to each edge grid is calculated using the CHF prediction mechanism model, and a virtual sample is constructed according to the combination of operating parameters corresponding to each edge grid and the prior CHF value; A CHF prediction model is constructed based on virtual samples that match the real CHF distribution, using the target virtual samples and the CHF experimental sample set.

2. The construction method as described in claim 1, characterized in that, The method further includes: A generative adversarial network is constructed, comprising a generator and a discriminator; the generator generates a simulated CHF value that matches the constraints based on the operating parameters involved in the combination of operating parameters; the discriminator takes the operating parameters and the simulated CHF value as input and outputs a authenticity score for the simulated CHF value. The generative adversarial network is trained based on the CHF test sample set. The adversarial optimization between the generator and the discriminator is carried out until the training converges, and the trained generative adversarial network is obtained. The virtual sample is input into the trained generative adversarial network, which outputs the authenticity score of the virtual sample. Virtual samples with an authenticity score greater than or equal to a preset threshold are identified as matching the real CHF distribution.

3. The construction method as described in claim 1, characterized in that, The step of identifying edge condition data from the CHF test sample set to filter out edge test sample sets from the CHF test sample set includes: The CHF test sample set is subjected to at least one of the following edge condition data identification processes: sparse mesh identification, physical mechanism-based edge condition identification, and isolated sample identification; The processing results of each edge condition data identification process are combined to obtain the edge test sample set.

4. The construction method as described in claim 3, characterized in that, The combination of operating parameters involves K operating parameters, and the processing result of the sparse mesh identification includes a candidate sparse set corresponding to the CHF test sample set. The sparse mesh identification of the CHF test sample set includes: The K-dimensional parameter space formed by the K operating parameters is uniformly divided into multiple grids; Determine the grid j to which each CHF test sample in the CHF test sample set belongs; Count the number of sample points contained in each grid j. n j , will satisfy n j Grids j with fewer than the number of points are identified as sparse grids, and the data points in the sparse grids constitute the candidate sparse set corresponding to the CHF test sample set.

5. The construction method as described in claim 4, characterized in that, The multiple grids are divided based on the division interval and the number of division intervals corresponding to each of the K operating parameters.

6. The construction method as described in claim 3, characterized in that, The combination of operating parameters involves K operating parameters, and the processing result of the edge operating condition identification includes a physical edge test sample set. The edge operating condition identification based on physical mechanisms on the CHF test sample set includes: CHF test samples that meet the physical mechanism conditions in the CHF test sample set are identified as physical edge test samples, forming a physical edge test sample set containing at least one of the physical edge test samples. The physical mechanism conditions are associated with one or more of the K operating condition parameters.

7. The construction method as described in claim 3, characterized in that, The processing result of the edge condition identification includes an isolated sample set. The isolated sample identification of the CHF test sample set includes: The anomaly score of each CHF test sample in the CHF test sample set is calculated using the random forest algorithm. CHF test samples with anomaly scores greater than or equal to the anomaly score threshold are identified as isolated samples, forming an isolated sample set containing at least one of the isolated samples.

8. The construction method according to any one of claims 1-7, characterized in that, The step of constructing a CHF prediction model based on the target virtual sample and the CHF test sample set includes: Obtain the initial CHF prediction model, which is a polynomial function that weights the feature matrix of the input sample by a set of coefficients. The output parameter of the initial CHF prediction model is the CHF prediction value. The set of coefficients includes multiple coefficients, and each element in the feature matrix is ​​associated with different operating condition parameters. Obtain the target loss function that includes the ridge regression penalty term; With minimizing the target loss function as the optimization objective, the values ​​of each coefficient in the coefficient set are updated by combining the target virtual sample and the CHF test sample set to optimize the initial CHF prediction model, thus obtaining the CHF prediction model.

9. A computer-readable medium storing computer program code, characterized in that, The computer program code implements the construction method as described in any one of claims 1-8 when executed by a processor.

10. An electronic device, characterized in that, include: At least one processor; as well as At least one memory storing instructions that, when executed individually or jointly by the at least one processor, cause the electronic device to perform the construction method as described in any one of claims 1-8.

Images