Naive bayes-based pump-driven two-phase flow loop fault diagnosis method, device, medium and equipment

Abstract

The application provides a pump-driven two-phase flow loop fault diagnosis method and device based on Naive Bayes, a medium and equipment. The method comprises the following steps: acquiring real-time monitoring data of a pump-driven two-phase flow loop; converting a target characteristic parameter of the real-time monitoring data into a discrete state value based on a preset discretization rule; inputting the discrete state value into a pre-trained fault diagnosis model to generate a probability curve corresponding to each of a plurality of fault types defined by the fault diagnosis model; determining a diagnosis result of the pump-driven two-phase flow loop based on a preset multi-threshold decision rule and the probability curve; wherein the multi-threshold decision rule is used to determine a fault occurrence, and the multi-threshold decision rule comprises an absolute probability condition, a relative probability value condition and a relative threshold length condition that need to be met by the probability curve of each fault type. The application constrains the model through the multi-threshold decision rule, improves the accuracy of the fault diagnosis result, and the obtained result has interpretability.

Description

Technical Field

[0001] This application relates to the field of fault diagnosis technology for long-life pump-driven two-phase fluid circuits, and more specifically, to a method, apparatus, medium, and equipment for fault diagnosis of pump-driven two-phase flow circuits based on Naive Bayes. Background Technology

[0002] Long-life pump-driven two-phase fluid circuits are widely used in thermal management systems with extremely high reliability requirements, such as nuclear power and aerospace. They have complex structures and involve multi-physics coupling, and are subject to various failure mechanisms and modes, such as feedwater loss and coolant flow loss. The fault characteristics are high-dimensional and highly correlated. Faults may lead to system shutdown or even safety accidents, so the timeliness, accuracy and interpretability of fault diagnosis are required.

[0003] In related technologies, although there are various technical solutions in the current field of fault diagnosis, none of them can meet the diagnostic needs of long-life pump-driven two-phase fluid circuits: First, the original circuit parameters reach 96 items, and existing methods mostly require complete datasets for modeling. In practice, it is difficult to obtain all parameters due to limitations in sensor cost and measurement conditions, resulting in low model practicality. Second, although mainstream neural network methods can fit complex fault characteristics, they are "black boxes" with invisible internal decisions, making it difficult to trace the causes of key fault parameters and failing to meet the needs of critical systems. Third, there is a lack of effective expert knowledge embedding mechanisms, making it difficult to combine expert experience with data-driven models and failing to solve problems such as sensor errors and adaptation to special operating conditions.

[0004] Therefore, this application provides a fault diagnosis scheme for pump-driven two-phase flow loops based on Naive Bayes to solve one of the above-mentioned technical problems. Summary of the Invention

[0005] The purpose of this application is to provide a method, apparatus, medium, and equipment for fault diagnosis of pump-driven two-phase flow loops based on Naive Bayes, which can solve at least one of the aforementioned technical problems. The specific solution is as follows:

[0006] According to a specific embodiment of this application, in a first aspect, this application provides a method for fault diagnosis of pump-driven two-phase flow loops based on Naive Bayes, the method comprising:

[0007] Acquire real-time monitoring data of the pump-driven two-phase flow loop;

[0008] Based on a preset discretization rule, the target feature parameters of the real-time monitoring data are converted into discrete state values.

[0009] The discrete state values ​​are input into a pre-trained fault diagnosis model to generate probability curves for each of the various fault types defined by the fault diagnosis model.

[0010] Based on the preset multi-threshold judgment rules and the probability curves, the diagnostic results of the pump-driven two-phase flow loop are determined; wherein, the multi-threshold judgment rules are used to determine the occurrence of faults, and the multi-threshold judgment rules include the absolute probability conditions, relative probability value conditions and relative threshold length conditions that the probability curves of each fault type must satisfy.

[0011] In one possible embodiment, converting the target feature parameters of the real-time monitoring data into discrete state values ​​based on a preset discretization rule includes:

[0012] Determine the target feature parameters: Use an attribute reduction algorithm based on the class difference matrix to select the target feature parameters from all system parameters of the pump-driven two-phase flow loop;

[0013] Based on preset discretization rules, the target feature parameters in the real-time monitoring data are converted into discrete state values.

[0014] In one possible embodiment, the fault diagnosis model includes a main diagnosis model and multiple sub-diagnosis models, wherein the main diagnosis model is used to diagnose multiple fault types; the sub-diagnosis models are used to diagnose some fault types among the fault types of the main diagnosis model, and the set of fault types of each sub-diagnosis model is a proper subset of the set of all fault types of the main diagnosis model.

[0015] The step of inputting the discrete state values ​​into a pre-trained fault diagnosis model to generate probability curves corresponding to various fault types defined by the fault diagnosis model includes:

[0016] The discrete state values ​​are input into the main diagnostic model, and based on the conditional probability table of the main diagnostic model, a probability curve corresponding to the fault type defined by the main diagnostic model is generated.

[0017] The discrete state values ​​are input into each of the sub-diagnostic models, and probability curves corresponding to the fault types defined by each sub-diagnostic model are generated based on the conditional probability tables of each sub-diagnostic model.

[0018] In one possible embodiment, determining the diagnostic result of the pump-driven two-phase flow loop based on preset multi-threshold judgment rules and the probability curve includes:

[0019] Based on the multi-threshold judgment rules and probability curves corresponding to the main diagnostic model, the main diagnostic results of the pump-driven two-phase flow loop are obtained.

[0020] Based on the multi-threshold judgment rules corresponding to each of the sub-diagnostic models, the sub-diagnostic results corresponding to each of the sub-diagnostic models are obtained respectively;

[0021] Based on the main diagnostic results and the sub-diagnostic results of all the sub-diagnostic models, the redundant diagnostic results of the pump-driven two-phase flow loop are obtained.

[0022] Based on the main diagnostic results and the redundant diagnostic results of the pump-driven two-phase flow circuit, the diagnostic results of the pump-driven two-phase flow circuit are determined.

[0023] In one possible embodiment, determining the diagnostic result of the pump-driven two-phase flow circuit based on the primary diagnostic result and the redundant diagnostic result includes:

[0024] The redundant diagnostic results are determined based on the majority voting mechanism and the diagnostic results corresponding to each of the sub-diagnostic models.

[0025] When the primary diagnostic result and the redundant diagnostic result are consistent, the consistent diagnostic result is determined to be the diagnostic result of the pump-driven two-phase flow circuit.

[0026] When the primary diagnostic result and the redundant diagnostic result are inconsistent, the diagnostic result of the pump-driven two-phase flow circuit shall be determined in one of the following ways:

[0027] Triggers the sending of an uncertain diagnostic result alert to a preset terminal;

[0028] The redundant diagnostic result is determined to be the diagnostic result of the pump-driven two-phase flow loop, and the event of inconsistent diagnostic results is reported as a reminder to the preset terminal;

[0029] The preset arbitration model is activated, and the diagnostic results of the pump-driven two-phase flow loop are determined based on the voting distribution of the main diagnostic model and the sub-diagnostic model.

[0030] In one possible embodiment, the fault diagnosis model includes a main diagnostic model and N sub-diagnostic models. The main diagnostic model is used to diagnose N types of faults. Each sub-diagnostic model diagnoses one fewer type of fault than the main diagnostic model, and any two sub-diagnostic models have different types of faults that are missing.

[0031] In one possible embodiment, the method further includes pre-training the fault diagnosis model, wherein the training method of the fault diagnosis model includes:

[0032] Obtain sample data of pump-driven two-phase flow circuits under normal operating conditions and various fault conditions;

[0033] Discretize the continuous sample data and define the discretization rules for the conditional attributes;

[0034] An algorithm based on the class difference matrix is ​​used to reduce the conditional attribute set and determine the target feature parameters for fault diagnosis.

[0035] Based on the reduced target feature parameters and sample data, the prior probability and conditional probability are calculated according to the Naive Bayes theorem to generate the initial conditional probability table of the fault diagnosis model.

[0036] The initial conditional probability table is corrected based on a preset expert knowledge base to obtain the final conditional probability table;

[0037] The sample data is input into the fault diagnosis model to be trained, and based on the final conditional probability table, a fault probability curve corresponding to each fault type is generated.

[0038] Based on the fault probability curve, the multi-threshold judgment rule is determined.

[0039] According to a specific embodiment of this application, in a second aspect, this application also provides a fault diagnosis device for a pump-driven two-phase flow loop based on Naive Bayes, the device comprising:

[0040] The input unit is used to acquire real-time monitoring data of the pump-driven two-phase flow loop;

[0041] Discrete unit, used to convert the target feature parameters of the real-time monitoring data into discrete state values ​​based on a preset discretization rule;

[0042] The diagnostic unit is used to input the discrete state values ​​into a pre-trained fault diagnosis model and generate probability curves corresponding to various fault types defined by the fault diagnosis model.

[0043] The output unit is used to determine the diagnostic result of the pump-driven two-phase flow loop based on the preset multi-threshold judgment rules and the probability curves; wherein, the multi-threshold judgment rules are used to determine the occurrence of faults, and the multi-threshold judgment rules include the absolute probability conditions, relative probability value conditions and relative threshold length conditions that the probability curves of each fault type must satisfy.

[0044] According to a specific embodiment of this application, in a third aspect, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any of the preceding claims.

[0045] According to a specific embodiment of this application, in a fourth aspect, this application also provides an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to perform the method as described in any of the preceding claims.

[0046] Compared with the prior art, the above-described solutions of this application have at least the following beneficial effects:

[0047] By using an attribute reduction algorithm based on class difference matrices, the original system parameters of the loop are filtered into a target subset. This reduces the number of parameters that need to be collected for real-time monitoring, reduces the scale of sensor deployment, and avoids the hardware costs associated with redundant parameters.

[0048] By generating probability curves from discrete state values ​​input into the model, the diagnostic results not only output the fault type but also inform the operator of key parameter changes and corresponding fault probability bases. The logic is traceable and the results are verifiable, unlike the "black box model" of neural networks, making it suitable for scenarios with high fault traceability requirements, such as nuclear power and aerospace.

[0049] By using absolute probability conditions to exclude low-probability suspected faults, relative probability value conditions to distinguish dominant fault types, and relative threshold length conditions to filter instantaneous false peaks, interference is avoided through triple screening.

[0050] Adapting to complex loop conditions and expanding applicable scenarios: Targeting the fluctuations in loop conditions and the strong correlation of fault characteristics, this method transforms the Naive Bayes method, which is suitable for independent sample classification, into a tool that can process continuous time-series monitoring data by mapping the degree of parameter deviation through discretization rules. This solves the problem that traditional methods are difficult to deal with time-series faults in complex systems.

[0051] Furthermore, existing fault diagnosis methods often involve first detecting anomalies and then inputting the detected faults into the fault diagnosis model. However, this application directly combines anomaly detection and fault diagnosis, meaning it can determine whether real-time data is normal. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings:

[0053] Figure 1 A flowchart illustrating the fault diagnosis method for pump-driven two-phase flow loops based on Naive Bayes provided in this application. Figure 1 ;

[0054] Figure 2 A flowchart illustrating the fault diagnosis method for pump-driven two-phase flow loops based on Naive Bayes provided in this application. Figure 2 ;

[0055] Figure 3a The probability curve of the pump-driven two-phase flow loop fault diagnosis method based on Naive Bayes provided in this application;

[0056] Figure 3bThe graph shows the change over time of the relative probability values ​​of various faults in a certain data of the pump-driven two-phase flow loop fault diagnosis method based on Naive Bayes provided in this application.

[0057] Figure 4 A schematic diagram of the fault diagnosis device for pump-driven two-phase flow loop based on Naive Bayes provided in this application;

[0058] Figure 5 This is a schematic diagram of the electronic device structure shown in an embodiment of this application. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0060] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0061] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0062] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.

[0063] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that an article or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such an article or device. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device that includes said element.

[0064] The optional embodiments of this application are described in detail below with reference to the accompanying drawings.

[0065] Figure 1 A flowchart illustrating the fault diagnosis method for pump-driven two-phase flow loops based on Naive Bayes provided in this application. Figure 1 ,like Figure 1 As shown, the method includes:

[0066] S101. Obtain real-time monitoring data of the pump-driven two-phase flow loop;

[0067] For example, the executing entity of this embodiment may be an electronic device, a terminal device, a server, or a device or apparatus capable of executing the scheme of this embodiment, and there are no limitations on this.

[0068] S102. Based on the preset discretization rules, the target feature parameters of the real-time monitoring data are converted into discrete state values.

[0069] Among them, the discretization rule takes the degree of parameter deviation from normal operating conditions as the core, transforms continuous values ​​into intuitive discrete states, reduces the interference of small fluctuations in real-time data, and at the same time retains key fault information, providing a data foundation for the subsequent generation of fault probability curves.

[0070] S103. Input the discrete state values ​​into the pre-trained fault diagnosis model to generate probability curves corresponding to the various fault types defined by the fault diagnosis model.

[0071] The fault diagnosis model is built around the Naive Bayes theorem and optimized using expert knowledge. During the pre-training phase, the model was trained using sample data from normal and fault conditions, and key parameters (such as WECS, VOID, and 8 other parameters) reduced based on the class difference matrix attributes. The initial conditional probability table was also corrected using expert knowledge to ensure that the probability calculation logic closely matches the actual fault characteristics of the circuit. When discrete state values ​​(derived from target feature parameters through preset rules, such as a WECS parameter deviation of 0.6 corresponding to a discrete value of 2) are input into the pre-trained fault diagnosis model, the model calls the pre-trained conditional probability table and, based on the Naive Bayes posterior probability calculation logic, calculates the probability of occurrence for each fault type defined by the model corresponding to the current discrete state value. Simultaneously, considering the temporal sequence of real-time monitoring data, the fault probability values ​​at different times are concatenated along the time dimension to generate a unique probability curve for each fault type. This curve visually presents the changing trend of the corresponding fault probability at different times.

[0072] S104. Based on the preset multi-threshold judgment rules and probability curves, determine the diagnostic results of the pump-driven two-phase flow circuit; wherein, the multi-threshold judgment rules are used to determine the occurrence of faults, and the multi-threshold judgment rules include the absolute probability conditions, relative probability value conditions and relative threshold length conditions that the probability curves of each fault type must satisfy.

[0073] The absolute probability condition is based on the probability boundary values ​​of each fault type determined during the pre-training phase of the fault diagnosis model. Its function is to filter out low-probability suspected fault signals by using the upper and lower probability limits. During pre-training, the model combines normal and faulty operating condition sample data and uses the quantile method to calculate the lower and upper limits of the absolute values ​​of the probability curves for each fault type (such as FLB water supply loss, LLB auxiliary machine drain pipe breakage, etc.). In the example, the lower limit of the absolute value for FLB faults is 9.87e-05 and the upper limit is 2.39e-02, while the lower limit for LLB faults is 9.25e-05 and the upper limit is 9.27e-03. During the judgment, only fault curves whose probabilities fall within the corresponding upper and lower limit intervals are retained, directly excluding interference signals with excessively low probabilities and no actual fault significance, thus avoiding misjudgment of faults due to small probability fluctuations.

[0074] The relative probability value condition, based on the comparison of multiple fault probabilities, is used to select the fault type with a relatively dominant probability from candidate faults that meet the absolute probability condition. In fault diagnosis of pump-driven two-phase flow loops, multiple fault probabilities may simultaneously fall within the absolute probability range. In such cases, it is necessary to compare the probability curves of each fault type and, at the same time or within the same time window, select the type with a significantly higher probability value than other candidate faults as the core judgment object. This avoids diagnostic confusion caused by the parallel existence of multiple fault probabilities and ensures focus on the high-probability signal corresponding to the true fault. The windowing method successfully applies Naive Bayes (steady-state) to real-time data fault diagnosis (dynamic).

[0075] The relative threshold length condition requires that the fault probability signal that meets the absolute probability and relative probability value conditions must remain stable within the set threshold length to exclude fleeting high probability peaks caused by instantaneous parameter fluctuations, ensuring that the diagnostic results only apply to stable real faults.

[0076] In this embodiment, an attribute reduction algorithm based on a class difference matrix is ​​used to filter the original system parameters of the loop into a target subset. This reduces the number of parameters that need to be collected for real-time monitoring, reduces the scale of sensor deployment, and avoids the hardware costs associated with redundant parameters.

[0077] By generating probability curves from discrete state values ​​input into the model, the diagnostic results not only output the fault type but also inform the operator of key parameter changes and corresponding fault probability bases. The logic is traceable and the results are verifiable, unlike the "black box model" of neural networks, making it suitable for scenarios with high fault traceability requirements, such as nuclear power and aerospace.

[0078] By using absolute probability conditions to exclude low-probability suspected faults, relative probability value conditions to distinguish dominant fault types, and relative threshold length conditions to filter instantaneous false peaks, interference is avoided through triple screening.

[0079] Adapting to complex loop conditions and expanding applicable scenarios: Targeting the fluctuations in loop conditions and the strong correlation of fault characteristics, this method transforms the Naive Bayes method, which is suitable for independent sample classification, into a tool that can process continuous time-series monitoring data by mapping the degree of parameter deviation through discretization rules. This solves the problem that traditional methods are difficult to deal with time-series faults in complex systems.

[0080] Figure 2 A flowchart illustrating the fault diagnosis method for pump-driven two-phase flow loops based on Naive Bayes provided in this application. Figure 2 ,like Figure 2 As shown, the method includes:

[0081] S201. Obtain real-time monitoring data of the pump-driven two-phase flow loop;

[0082] S202. Using an attribute reduction algorithm based on a class difference matrix, target characteristic parameters are selected from all system parameters of the pump-driven two-phase flow loop; based on a preset discretization rule, the target characteristic parameters in the real-time monitoring data are converted into discrete state values.

[0083] The attribute reduction algorithm for the class difference matrix removes redundant parameters and retains the most critical feature parameters for fault diagnosis from the high-dimensional original system parameters of the pump-driven two-phase flow loop. Specifically, it constructs a sample decision table (recording the correspondence between parameters and fault types) and defines the class difference matrix to statistically analyze the distinguishing ability of parameters between different fault categories. For example, the more times a parameter appears as "1" in the class difference matrix, the stronger its classification ability. Finally, this algorithm selects eight key parameters (such as WECS, VOID, SCMA, LWRB, RBLK, WTRA, RM2, and RM4) from the 96 original system parameters of the loop as target feature parameters, significantly reducing data acquisition and model calculation costs.

[0084] All system parameters of the pump-driven two-phase flow loop refer to all thermodynamic, mechanical, and other parameters that can be monitored during the operation of the long-life pump-driven two-phase flow loop. These are the original data sources for fault diagnosis. However, due to the large number of parameters and the difficulty in measuring some parameters, it is necessary to use attribute reduction algorithms to screen core parameters and avoid redundant parameters.

[0085] The target characteristic parameters are a subset of key parameters obtained after screening. They are the core indicators that can reflect the characteristics of loop faults to the greatest extent (such as WECS reflecting changes in medium state and LWRB reflecting abnormal flow).

[0086] Various fault types can include: FLB (Feedwater Line Break), LLB (Letdown Line Break in auxiliary buildings), LOCA (Loss of Coolant Accident), LR (Load Rejection), MD (Moderator Dilution), RW (Rod Withdrawal), SGATR (Steam Generator A Tube Rupture), SGBTR (Steam Generator B Tube Rupture), and SLBOC (Steam Line Break Outside Containment). These are the core objects of model diagnostics, and each fault corresponds to a specific probability curve.

[0087] S203. Input the discrete state values ​​into the main diagnostic model, and generate the probability curves corresponding to the fault types defined by the main diagnostic model based on the conditional probability table of the main diagnostic model; input the discrete state values ​​into each sub-diagnostic model respectively, and generate the probability curves corresponding to the fault types defined by the sub-diagnostic model based on the conditional probability table of each sub-diagnostic model respectively.

[0088] The fault diagnosis model includes a main diagnosis model and multiple sub-diagnosis models. The main diagnosis model is used to diagnose multiple fault types. The sub-diagnosis models are used to diagnose some fault types among the fault types of the main diagnosis model. The set of fault types of each sub-diagnosis model is a proper subset of the set of all fault types of the main diagnosis model.

[0089] Understandably, the main diagnostic model covers all fault types in the loop, while multiple sub-diagnostic models cover subsets of faults in the main model, forming a redundant system of comprehensive diagnosis by the main model and supplementary verification by the sub-models. After inputting discrete state values ​​into the main and sub-models respectively, both generate independent fault probability curves based on their respective conditional probability tables. By comparing the probability curves of the main and sub-models (e.g., if the main model determines "LLB fault," most sub-models also output "high probability of LLB fault"), the diagnostic results can be cross-validated, avoiding misjudgments caused by parameter fluctuations or local data biases in a single model. As a proper subset of the fault types in the main model, the sub-models do not need to cover all faults; they can optimize their conditional probability tables for the specific fault types they diagnose. This allows the sub-models to more accurately capture the characteristic associations of specific faults, thereby improving the diagnostic accuracy for a single fault type. The full fault coverage of the main diagnostic model ensures that no potential fault is missed, meeting the basic requirement of comprehensive monitoring of multiple loop fault types. The sub-models optimize their diagnostic logic for the unique characteristics of each fault, avoiding the problem of broad coverage but low accuracy in a single model. The combination of the two methods achieves both comprehensive fault diagnosis and accurate diagnosis of individual faults.

[0090] For example, the fault diagnosis model includes a main diagnostic model and N sub-diagnostic models. The main diagnostic model is used to diagnose N types of faults. Each sub-diagnostic model has one less type of fault than the main diagnostic model, and the fault types that are missing in any two sub-diagnostic models are different.

[0091] Understandably, the main model diagnoses N types of faults, while each of the N sub-models is missing one different fault. By arbitrarily selecting N-1 faults from the N faults to build fault diagnosis models, N fault diagnosis models are obtained. Cross-validation using probability curves generated by the main and sub-models reduces misjudgments by a single model and improves the reliability of the results. Since each sub-model only diagnoses N-1 faults, it doesn't need to consider the one fault that was removed, reducing interference from that fault's parameters and more accurately capturing the remaining fault characteristics, thus improving the accuracy of specific fault diagnosis. Each sub-model has a different missing fault type; even if a sub-model deviates due to a missing fault, other sub-models and the main model can still diagnose normally, avoiding global failure. The main model fully covers all N faults to prevent omissions, while the sub-models focus on specific faults to improve accuracy, meeting the need for comprehensive monitoring and accurate judgment of multiple faults in a loop.

[0092] S204. Based on the multi-threshold judgment rules and probability curves corresponding to the main diagnostic model, the main diagnostic results of the pump-driven two-phase flow loop are obtained.

[0093] First, the probability curves corresponding to the N fault types that can be diagnosed by the main diagnostic model are called. Then, the multi-threshold judgment rules exclusive to the main diagnostic model (including the absolute probability conditions, relative probability value conditions and relative threshold length conditions that the probability curves of each fault type must meet, such as the absolute probability lower limit of 9.87e-05 for FLB faults) are applied to screen and judge the probability curves, eliminate low probability interference, lock in the fault types that are dominant and stable, and finally obtain the main diagnostic result of the pump-driven two-phase flow loop (i.e. the fault type determined by the main model).

[0094] S205. Based on the multi-threshold judgment rules corresponding to each sub-diagnostic model, obtain the sub-diagnostic results corresponding to each sub-diagnostic model; based on the main diagnostic results and the sub-diagnostic results of all sub-diagnostic models, obtain the redundant diagnostic results of the pump-driven two-phase flow loop.

[0095] Specifically, for N sub-diagnostic models (each diagnosing one different fault type), the probability curves generated by each sub-model corresponding to the N-1 fault types it focuses on are called. Then, the multi-threshold judgment rules adapted to the diagnostic range of each sub-model (the rule logic is consistent with the main model, but adapted to the fault type subset of the sub-model) are applied one by one to judge the probability curves, and finally the sub-diagnostic result corresponding to each sub-model is obtained (i.e., the fault type determined by each sub-model). Combining all the obtained sub-diagnostic results, a majority voting mechanism is used (the fault type that appears most frequently in the sub-diagnostic results is counted), and the fault type that is consistent with the majority is determined as the redundant diagnostic result of the pump-driven two-phase flow loop. The reliability of the diagnosis is improved through cross-validation of the sub-model results.

[0096] S206. Based on the main diagnostic results and redundant diagnostic results of the pump-driven two-phase flow circuit, determine the diagnostic results of the pump-driven two-phase flow circuit.

[0097] For example, based on the main diagnostic results and redundant diagnostic results of the pump-driven two-phase flow loop, the diagnostic results of the pump-driven two-phase flow loop are determined, including:

[0098] Based on the majority voting mechanism and the diagnostic results corresponding to each sub-diagnostic model, redundant diagnostic results are determined.

[0099] When the primary diagnostic result and the redundant diagnostic result are consistent, the consistent diagnostic result is determined to be the diagnostic result of the pump-driven two-phase flow loop.

[0100] When the primary diagnostic result and the redundant diagnostic result are inconsistent, the diagnostic result of the pump-driven two-phase flow loop shall be determined by one of the following methods:

[0101] Triggers the sending of an uncertain diagnostic result alert to a preset terminal;

[0102] The redundant diagnostic result is determined to be the diagnostic result of the pump-driven two-phase flow loop, and the event of inconsistent diagnostic results is reported as an alert to the preset terminal;

[0103] The preset arbitration model is activated, and the diagnostic results of the pump-driven two-phase flow loop are determined based on the voting distribution of the main diagnostic model and the sub-diagnostic model.

[0104] The redundant diagnostic results are generated through a "majority voting mechanism": the diagnostic results of all sub-diagnostic models are statistically analyzed, and the fault type that occurs most frequently is identified as the redundant diagnostic result. Then, the final diagnostic result is determined in two ways: if the primary diagnostic result and the redundant diagnostic result are consistent (e.g., both are determined to be "LLB auxiliary machine drain pipe rupture"), then the consistent result is directly used as the final diagnostic result for the pump-driven two-phase flow loop; if they are inconsistent, the result can be determined in one of the following ways: first, send a "diagnostic result uncertain" reminder to the preset terminal; second, use the redundant diagnostic result as the final result, and simultaneously report the "primary and redundant results inconsistent" event to the preset terminal; third, activate the preset arbitration model, and combine the voting distribution of the primary and sub-model diagnostic results (e.g., the number of times each model supports different faults) to finally determine the loop diagnostic result.

[0105] In this embodiment, a target feature parameter is selected from all system parameters using an attribute reduction algorithm based on the class difference matrix. This reduces the number of parameters to be monitored, lowers the scale of sensor deployment and the difficulty of data acquisition, avoids the model implementation difficulties caused by redundant parameters, and improves the practicality of the method. The discretization rule adapts to the requirements of the Naive Bayes model, ensuring effective data input. Relying on the characteristics of the Naive Bayes model, the diagnostic basis is presented by generating probability curves for each fault type, rather than relying on an untraceable "black box" model. This facilitates the understanding of the fault determination logic and meets the requirements for the reliability of diagnostic results in pump-driven two-phase flow loops (such as nuclear power and aerospace scenarios). The main diagnostic model covers all fault types, and N sub-models each diagnose one less different fault (forming a proper subset of the main model), forming a redundant system of comprehensive diagnosis by the main model and focused verification by the sub-models. By generating probability curves independently by the main and sub-models and cross-validating them, the risk of misjudgment by a single model is reduced, and the sub-models focusing on certain faults can further improve the diagnostic accuracy of specific faults. Multi-threshold judgment rules (absolute probability control of boundaries, relative probability selection of core, threshold length filtering) effectively filter interference signals and avoid false alarms and missed alarms; when the primary and redundant results are inconsistent, the system can send reminders, take the redundant results as the standard, or start the arbitration model to handle the situation, adapting to the diagnostic needs under complex working conditions and ensuring stable and reliable results.

[0106] In some embodiments, the above-described Naive Bayes-based fault diagnosis method for pump-driven two-phase flow loops further includes a pre-trained fault diagnosis model, wherein the training method for the fault diagnosis model includes:

[0107] Obtain sample data of pump-driven two-phase flow circuits under normal operating conditions and various fault conditions;

[0108] Discretize continuous sample data and define discretization rules for conditional attributes;

[0109] An algorithm based on the class difference matrix is ​​used to reduce the conditional attribute set and determine the target feature parameters for fault diagnosis.

[0110] Based on the reduced target feature parameters and sample data, the prior probability and conditional probability are calculated according to the Naive Bayes theorem to generate the initial conditional probability table of the fault diagnosis model.

[0111] The initial conditional probability table is corrected based on a pre-set expert knowledge base to obtain the final conditional probability table.

[0112] The sample data is input into the fault diagnosis model to be trained, and based on the final conditional probability table, a fault probability curve corresponding to each fault type is generated.

[0113] Based on the fault probability curve, a multi-threshold judgment rule is determined.

[0114] For example, initial conditions are obtained by acquiring data on normal and fault conditions using a simulator, and relevant expert knowledge is obtained by consulting with experts. Specifically:

[0115] Table 1 shows the fault descriptions of the selected long-life pump-driven two-phase fluid circuits.

[0116] Table 1 Fault Description

[0117]

[0118] The expert knowledge obtained is shown in Table 2.

[0119] Table 2 Self-built Expert Knowledge Table

[0120]

[0121]

[0122] A fault diagnosis model for a long-life pump-driven two-phase fluid loop based on Naive Bayes is obtained, as follows:

[0123] The data length from 2 / 3 to 3 / 4 was selected as the data condition when the system reaches steady state. Based on this, the data was discretized. The natural language processing results are shown in Table 3.

[0124] Table 3 Natural Language Processing Results

[0125] node meaning scope 2 Significant increase (0.5,+∞) 1 rise [0.3,0.5) 0.5 Slight rise [0.1,0.3) 0 Basically unchanged [-0.1,0.1) -0.5 Slight decline [-0.3,-0.1) -1 decline [-0.5,-0.3) -2 Significant decline (-∞,-0.5)

[0126] When building the model, the training and test sets must first be divided. 40% of the dataset for each fault type is selected as the training set, and the remaining dataset representing the fault severity is used as the test set. First, the degree of variation of all parameters in the fault data is calculated, and a sample decision table is established for all parameters to perform attribute reduction. In this example, the reduced parameters are: ['WECS','VOID','SCMA','LWRB','RBLK','WTRA','RM2','RM4']. The sample decision table for the feature parameters is established using the attribute-reduced parameters under sliding windowing. When windowing, the window length is 10 units, and the window moves 2 units at a time. Note: The shorter the distance moved each time, the better the fault diagnosis result. The results of the sample decision table for the feature parameters are shown in Table 4.

[0127] Table 4 Sample Decision Table Regarding Feature Parameters

[0128]

[0129]

[0130] The following section describes multiple threshold control, combining the sample decision table for feature parameters with the conditional probability table, resulting in the following probability curve: Figure 3b As shown. Figure 3a The figure shows how the relative probability values ​​of various faults occurring in a certain data point change over time. Figure 3b The figure shows the absolute probability values ​​of the data occurrence over time. Position ② in Figure 3 represents the ideal fault identification location. Therefore, a relative threshold should be set to exclude the influence of other faults. However, position ① still affects the calculation results and can easily lead to misdiagnosis. Therefore, the concept of a relative threshold length is introduced, as shown in position ⑤ in Figure 3. Position ⑤ is also truncated using a sliding window method. When a certain proportion of the data within the relative threshold range has an absolute probability value between the upper limit ③ and the lower limit ④ of the absolute threshold, it indicates that the fault represented by the probability curve has occurred in that data point.

[0131] In this embodiment, sample data of the pump-driven two-phase flow circuit under normal operating conditions and various fault conditions are first acquired; then, the continuous sample data is discretized, and discretization rules for conditional attributes are defined; next, the conditional attribute set is reduced using an algorithm based on the class difference matrix to determine the target feature parameters; then, based on the Naive Bayes theorem, the prior probability and conditional probability are calculated by combining the reduced target feature parameters and sample data to generate an initial conditional probability table for the model; the initial table is corrected using a preset expert knowledge base to obtain the final conditional probability table; the sample data is input into the model to be trained, and fault probability curves corresponding to each fault type are generated based on the final table; finally, multi-threshold judgment rules are determined based on the fault probability curves. Thus, comprehensive training data ensures the basic reliability of the model: covering normal and various fault condition samples, ensuring the model can learn fault characteristics across all scenarios and avoiding diagnostic blind spots caused by incomplete data; adapting to model requirements and reducing computational complexity: discretization processing adapts to the Naive Bayes model's requirement for discrete data, attribute reduction reduces redundant parameters, reducing both the model's computational burden and the cost of sensor deployment for subsequent engineering applications; integrating expert experience improves model fit: using an expert knowledge base to correct the conditional probability table compensates for the shortcomings of purely data-driven models that may deviate from engineering reality, making the model's probability calculations more consistent with the real fault patterns of the loop; generating exclusive judgment rules lays the foundation for diagnostic accuracy: determining multi-threshold rules based on the fault probability curves generated during training can specifically filter out instantaneous interference in loop fault diagnosis, reducing the risk of false alarms and missed alarms in subsequent diagnoses.

[0132] This application also provides apparatus embodiments that follow the above embodiments to implement the method steps of the above embodiments. The interpretation of the same names is the same as that of the above embodiments, and they have the same technical effects as those of the above embodiments. They will not be described again here.

[0133] like Figure 4 As shown, this application provides a fault diagnosis device for pump-driven two-phase flow loops based on Naive Bayes, the device comprising:

[0134] Input unit 401 is used to acquire real-time monitoring data of the pump-driven two-phase flow loop;

[0135] Discrete unit 402 is used to convert the target feature parameters of real-time monitoring data into discrete state values ​​based on preset discretization rules;

[0136] The diagnostic unit 402 is used to input discrete state values ​​into a pre-trained fault diagnosis model to generate probability curves corresponding to various fault types defined by the fault diagnosis model.

[0137] The output unit 404 is used to determine the diagnostic results of the pump-driven two-phase flow loop based on preset multi-threshold judgment rules and probability curves. The multi-threshold judgment rules are used to determine the occurrence of faults. The multi-threshold judgment rules include the absolute probability conditions, relative probability value conditions and relative threshold length conditions that the probability curves of each fault type must satisfy.

[0138] In some embodiments, the discrete unit 402 is specifically used to: employ an attribute reduction algorithm based on a class difference matrix to select target feature parameters from all system parameters of the pump-driven two-phase flow loop; and convert the target feature parameters in the real-time monitoring data into discrete state values ​​based on a preset discretization rule.

[0139] In some embodiments, the fault diagnosis model includes a main diagnosis model and multiple sub-diagnosis models. The main diagnosis model is used to diagnose multiple fault types. The sub-diagnosis models are used to diagnose some fault types among the fault types of the main diagnosis model, and the set of fault types of each sub-diagnosis model is a proper subset of the set of all fault types of the main diagnosis model. The diagnosis unit 402 is specifically used to input discrete state values ​​into the main diagnosis model and generate probability curves corresponding to the fault types defined by the main diagnosis model based on the conditional probability table of the main diagnosis model; and to input discrete state values ​​into each sub-diagnosis model respectively and generate probability curves corresponding to the fault types defined by each sub-diagnosis model based on the conditional probability table of each sub-diagnosis model respectively.

[0140] In some embodiments, the output unit 404 is specifically used to obtain the main diagnostic result of the pump-driven two-phase flow loop based on the multi-threshold judgment rules and probability curves corresponding to the main diagnostic model.

[0141] Based on the multi-threshold judgment rules corresponding to each sub-diagnostic model, the sub-diagnostic results corresponding to each sub-diagnostic model are obtained respectively;

[0142] Based on the main diagnostic results and the sub-diagnostic results of all sub-diagnostic models, the redundant diagnostic results of the pump-driven two-phase flow loop are obtained.

[0143] Based on the main diagnostic results and redundant diagnostic results of the pump-driven two-phase flow loop, the diagnostic results of the pump-driven two-phase flow loop are determined.

[0144] In some embodiments, the output unit 404 is specifically used to determine redundant diagnostic results based on the majority voting mechanism and the diagnostic results corresponding to each sub-diagnostic model;

[0145] When the primary diagnostic result and the redundant diagnostic result are consistent, the consistent diagnostic result is determined to be the diagnostic result of the pump-driven two-phase flow loop.

[0146] When the primary diagnostic result and the redundant diagnostic result are inconsistent, the diagnostic result of the pump-driven two-phase flow loop shall be determined by one of the following methods:

[0147] Triggers the sending of an uncertain diagnostic result alert to a preset terminal;

[0148] The redundant diagnostic result is determined to be the diagnostic result of the pump-driven two-phase flow loop, and the event of inconsistent diagnostic results is reported as an alert to the preset terminal;

[0149] The preset arbitration model is activated, and the diagnostic results of the pump-driven two-phase flow loop are determined based on the voting distribution of the main diagnostic model and the sub-diagnostic model.

[0150] In some embodiments, the fault diagnosis model includes a main diagnostic model and N sub-diagnostic models. The main diagnostic model is used to diagnose N types of faults. Each sub-diagnostic model has one fewer type of fault than the main diagnostic model, and the fault types missing from any two sub-diagnostic models are different.

[0151] like Figure 5 As shown, this embodiment provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by a processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method steps of the above embodiment.

[0152] This application provides a non-volatile computer storage medium storing computer-executable instructions that can execute the method steps of the above embodiments.

[0153] The following is for reference. Figure 5 The diagram illustrates a structural schematic of an electronic device suitable for implementing the embodiments of this application. The terminal devices in the embodiments of this application may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0154] like Figure 5As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage device 508 into a random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device. The processing unit 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0155] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0156] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the processing device 501, it performs the functions defined in the methods of the embodiments of this application.

[0157] It should be noted that the computer-readable medium described above in this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0158] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0159] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0160] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0161] The units described in the embodiments of this application can be implemented in software or hardware. The names of the units are not, in some cases, limiting the scope of the unit itself.

Claims

1. A fault diagnosis method for pump-driven two-phase flow loops based on Naive Bayes, characterized in that, The method includes: Acquire real-time monitoring data of the pump-driven two-phase flow loop; Based on a preset discretization rule, the target feature parameters of the real-time monitoring data are converted into discrete state values. The discrete state values ​​are input into a pre-trained fault diagnosis model to generate probability curves for each of the various fault types defined by the fault diagnosis model. Based on the preset multi-threshold judgment rules and the probability curves, the diagnostic results of the pump-driven two-phase flow circuit are determined; wherein, the multi-threshold judgment rules are used to determine the occurrence of faults, and the multi-threshold judgment rules include the absolute probability conditions, relative probability value conditions and relative threshold length conditions that the probability curves of each fault type must satisfy. The fault diagnosis model includes a main diagnosis model and multiple sub-diagnosis models. The main diagnosis model is used to diagnose multiple fault types. The sub-diagnosis models are used to diagnose some fault types among the fault types of the main diagnosis model, and the set of fault types of each sub-diagnosis model is a proper subset of the set of all fault types of the main diagnosis model. The step of inputting the discrete state values ​​into a pre-trained fault diagnosis model to generate probability curves corresponding to various fault types defined by the fault diagnosis model includes: The discrete state values ​​are input into the main diagnostic model, and based on the conditional probability table of the main diagnostic model, a probability curve corresponding to the fault type defined by the main diagnostic model is generated. The discrete state values ​​are input into each of the sub-diagnostic models, and probability curves corresponding to the fault types defined by each sub-diagnostic model are generated based on the conditional probability tables of each sub-diagnostic model. The diagnostic result of the pump-driven two-phase flow loop, determined based on the preset multi-threshold judgment rule and the probability curve, includes: Based on the multi-threshold judgment rules and probability curves corresponding to the main diagnostic model, the main diagnostic results of the pump-driven two-phase flow loop are obtained. Based on the multi-threshold judgment rules corresponding to each of the sub-diagnostic models, the sub-diagnostic results corresponding to each of the sub-diagnostic models are obtained respectively; Based on the main diagnostic results and the sub-diagnostic results of all the sub-diagnostic models, the redundant diagnostic results of the pump-driven two-phase flow loop are obtained. Based on the main diagnostic results and the redundant diagnostic results of the pump-driven two-phase flow circuit, the diagnostic results of the pump-driven two-phase flow circuit are determined. The determination of the diagnostic results for the pump-driven two-phase flow circuit based on the primary diagnostic results and the redundant diagnostic results includes: The redundant diagnostic results are determined based on the majority voting mechanism and the diagnostic results corresponding to each of the sub-diagnostic models. When the primary diagnostic result and the redundant diagnostic result are consistent, the consistent diagnostic result is determined to be the diagnostic result of the pump-driven two-phase flow circuit. When the primary diagnostic result and the redundant diagnostic result are inconsistent, the diagnostic result of the pump-driven two-phase flow circuit shall be determined in one of the following ways: Triggers the sending of an uncertain diagnostic result alert to a preset terminal; The redundant diagnostic result is determined to be the diagnostic result of the pump-driven two-phase flow loop, and the event of inconsistent diagnostic results is reported as a reminder to the preset terminal; The preset arbitration model is activated, and the diagnostic results of the pump-driven two-phase flow loop are determined based on the voting distribution of the main diagnostic model and the sub-diagnostic model.

2. The method according to claim 1, characterized in that, The process of converting the target feature parameters of the real-time monitoring data into discrete state values ​​based on preset discretization rules includes: The target feature parameters are selected from all system parameters of the pump-driven two-phase flow loop using an attribute reduction algorithm based on the class difference matrix. Based on preset discretization rules, the target feature parameters in the real-time monitoring data are converted into discrete state values.

3. The method according to claim 1, characterized in that, The fault diagnosis model includes a main diagnosis model and N sub-diagnosis models. The main diagnosis model is used to diagnose N types of faults. Each sub-diagnosis model diagnoses one fewer type of fault than the main diagnosis model, and the fault types missing from any two sub-diagnosis models are different.

4. The method according to claim 1, characterized in that, The method further includes pre-training the fault diagnosis model, wherein the training method of the fault diagnosis model includes: Obtain sample data of pump-driven two-phase flow circuits under normal operating conditions and various fault conditions; Discretize the continuous sample data and define the discretization rules for the conditional attributes; An algorithm based on the class difference matrix is ​​used to reduce the conditional attribute set and determine the target feature parameters for fault diagnosis. Based on the reduced target feature parameters and sample data, the prior probability and conditional probability are calculated according to the Naive Bayes theorem to generate the initial conditional probability table of the fault diagnosis model. The initial conditional probability table is corrected based on a preset expert knowledge base to obtain the final conditional probability table; The sample data is input into the fault diagnosis model to be trained, and based on the final conditional probability table, a fault probability curve corresponding to each fault type is generated. Based on the fault probability curve, the multi-threshold judgment rule is determined.

5. A fault diagnosis device for pump-driven two-phase flow loops based on Naive Bayes, characterized in that, The apparatus for performing the method according to any one of claims 1 to 4, the apparatus comprising: The input unit is used to acquire real-time monitoring data of the pump-driven two-phase flow loop; Discrete unit, used to convert the target feature parameters of the real-time monitoring data into discrete state values ​​based on a preset discretization rule; The diagnostic unit is used to input the discrete state values ​​into a pre-trained fault diagnosis model and generate probability curves corresponding to various fault types defined by the fault diagnosis model. The output unit is used to determine the diagnostic result of the pump-driven two-phase flow loop based on the preset multi-threshold judgment rules and the probability curves; wherein, the multi-threshold judgment rules are used to determine the occurrence of faults, and the multi-threshold judgment rules include the absolute probability conditions, relative probability value conditions and relative threshold length conditions that the probability curves of each fault type must satisfy.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 4.

7. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 4.

Images