Operation and maintenance data processing method and device based on a core system

Abstract

The present specification provides a nuclear system-based operation and maintenance data processing method and device. On the one hand, by introducing and using a first intelligent agent and a second intelligent agent to assist in completing the operation and maintenance data processing of the nuclear system, the data processing process can be effectively simplified and the processing difficulty can be reduced. On the other hand, by determining and according to the current working condition scene, different situations are distinguished, a matching target exploration model is used for multi-mode differentiation, and targeted exploration processing is performed, which can effectively reduce the overall data processing amount while ensuring the data processing accuracy, avoid meaningless state expansion and calculation redundancy, and improve the overall processing efficiency. Thus, it can be well adapted to the complex nuclear system operation and maintenance scene, and a nuclear system operation and maintenance report with high reference value and small error can be efficiently and intelligently generated at a small data processing cost, so as to effectively guide the safe operation and maintenance management of the nuclear system and ensure the safe operation of the nuclear system.

Description

Technical Field

[0001] This manual belongs to the field of nuclear system safety operation and maintenance technology, and in particular relates to methods and devices for operation and maintenance data processing based on nuclear systems. Background Technology

[0002] With the development and promotion of nuclear technology, more and more nuclear systems (such as nuclear power generation equipment) are being applied to industrial production.

[0003] However, nuclear systems are inherently complex in structure and contain numerous components. Furthermore, operational analysis requires consideration of a large number of different operating states and the complex transition mechanisms between these states. Consequently, using existing methods to analyze and manage nuclear systems based on their operating states often results in cumbersome and complex processes, high processing difficulty, a high technical threshold for technical personnel, low efficiency, and a tendency to produce errors.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This specification provides a method and apparatus for processing operation and maintenance data based on a nuclear system. It can be well adapted to complex nuclear system operation and maintenance scenarios and can efficiently and intelligently generate operation and maintenance reports with high reference value and low error in nuclear systems with relatively low data processing costs.

[0006] This manual provides a method for processing operation and maintenance data based on a kernel system, including: Obtain the description text of the target core system; The first intelligent agent extracts the component parameters of the target core system based on the description text; and determines the success criteria for the target core system. Based on the component parameters of the target core system, a transfer rule for the target core system is generated; Determine and, based on the current working conditions, identify a matching target exploration model from the preset exploration models; wherein the preset exploration models include at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; Using the target exploration model, the target core system's component parameters, success criteria, and transition rules are used to perform corresponding exploration processing, resulting in the state transition diagram of the target core system. The second intelligent agent generates an operation and maintenance report for the target core system based on the state transition diagram.

[0007] In one embodiment, the component parameters include at least one of the following: number of components, component identifier, component type, connection path, backup startup failure rate, failure rate, repair success rate, common cause failure rate, and system structure. The system structure includes at least one of the following: series structure, parallel structure, k-n structure, one-use-one-standby structure, and one-use-multiple-standby structure.

[0008] In one embodiment, the transfer rule includes at least one of the following: fault rule, repair rule, backup activation rule, and common cause fault rule.

[0009] In one embodiment, determining and identifying a matching target exploration model from a preset exploration model based on the current working conditions includes: Determine the number of components in the target core system and / or the operational and maintenance requirements parameters for the target core system; Based on the number of components and / or operation and maintenance requirements parameters, determine the current operating scenario of the target core system; Based on the current working conditions, a matching target exploration model is determined from the preset exploration models.

[0010] In one embodiment, when the target exploration model includes a first exploration model, the step of using the target exploration model to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain the state transition diagram of the target core system includes: The current round of exploration will be processed as follows: Obtain the previous state of the target core system; Using the target exploration model, the state label of the previous round's state is determined based on the success criteria and component parameters of the target kernel system; When the status label of the previous round indicates that the status is normal, determine the status of the current round based on the status of the previous round and the transition rules; and calculate the transition rate of the current round from the status of the previous round to the status of the current round. Update the state set based on the state of the previous round and the corresponding state label; and update the transition set based on the transition rate of the current round.

[0011] In one embodiment, when the target exploration model includes a second exploration model, the step of using the target exploration model to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain the state transition diagram of the target core system includes: The current round of exploration will be processed as follows: Obtain the previous state of the target core system; Check whether the link length of the state link between the previous state and the initial state of the target core system is less than the preset budget threshold; When the link length of the state link between the previous state and the initial state of the target core system is less than the preset budget threshold, the state label of the previous state is determined by the target exploration model based on the success criteria and component parameters of the target core system. When the state label of the previous state indicates that the state is normal, determine the successor state of the previous state based on the previous state and the transition rules; and calculate the path priority parameter from the previous state to the successor state. Based on the path priority parameter, determine the state of the current round from the successor states; and calculate the transition rate from the state of the previous round to the state of the current round. Update the state set based on the state of the previous round and the corresponding state label; and update the transition set based on the transition rate of the current round and the path priority parameter of the current round's state.

[0012] In one embodiment, after detecting whether the link length of the state link between the previous state and the initial state of the target core system is less than a preset budget threshold, the method further includes: When the link length of the state link between the previous state and the initial state of the target core system is greater than or equal to the preset budget threshold, state evolution features are extracted based on the state link between the previous state and the initial state of the target core system. Based on the state evolution characteristics, detect whether the adaptive extension condition with respect to the preset budget threshold is met; When it is determined that the adaptive extension condition is met, a matching extension adjustment value is determined based on the state evolution characteristics; Based on the aforementioned extension adjustment value, the preset budget threshold for the state link between the previous state and the initial state of the target core system is dynamically updated.

[0013] In one embodiment, after detecting whether an adaptive extension condition with respect to a preset budget threshold is met, the method further includes: When it is determined that the adaptive extension condition is not met, the exploration process based on the state of the previous round ends; Update the state set based on the state of the previous round and the corresponding state label; Check if the exploration termination conditions are met; When the exploration termination condition is met, the state that is not included in the state set is determined as the absorption state; The absorption states are saved to an absorption state set; and the transition set is updated based on the absorption state set.

[0014] In one embodiment, generating an operation and maintenance report for the target core system using a second intelligent agent based on the state transition diagram includes: The state transition graph is processed by a second intelligent agent to extract key parameters; wherein, the key parameters include: path priority parameters, activity column parameters, state labels, and absorption state information; Based on the aforementioned key parameters, risk transfer paths and risk components are determined; Based on the risk transfer path and risk components, the risk pattern type is determined; Based on the risk mode type, generate an operation and maintenance report for the target core system.

[0015] This specification also provides a kernel-based operation and maintenance data processing device, including: The acquisition module is used to acquire the description text of the target core system; The processing module is used to extract component parameters of the target core system based on the description text using a first intelligent agent; and to determine success criteria for the target core system. The first generation module is used to generate transfer rules for the target core system based on the component parameters of the target core system. The determination module is used to determine and, based on the current working conditions, identify a matching target exploration model from the preset exploration models; wherein the preset exploration models include at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; The exploration module is used to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system using the target exploration model, and to obtain the state transition diagram of the target core system. The second generation module is used to generate an operation and maintenance report of the target core system based on the state transition diagram using the second intelligent agent.

[0016] Based on the operation and maintenance data processing method and apparatus for core systems provided in this specification, before specific implementation, a first intelligent agent and a second intelligent agent that meet the requirements and are adapted to the operation and maintenance of the core system can be trained using a large language model. In specific implementation, firstly, the description text of the target core system is obtained; the first intelligent agent extracts the component parameters of the target core system based on the description text; and automatically determines the success criteria for the target core system; then, based on the component parameters of the target core system, transition rules for the target core system are generated; a matching target exploration model is determined from the preset exploration models according to the current working scenario; wherein the preset exploration models include at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; then, the target exploration model performs corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain the state transition diagram of the target core system; and the second intelligent agent generates an operation and maintenance report for the target core system based on the state transition diagram. On the one hand, by introducing and using first and second intelligent agents to assist in the processing of nuclear system operation and maintenance data, the data processing process can be effectively simplified, the user's processing burden reduced, and the processing difficulty lowered. On the other hand, by identifying and differentiating different situations based on the current operating scenario, and using matching target exploration models for multi-mode, differentiated, and targeted exploration processing, the overall data processing volume can be effectively reduced while ensuring data processing accuracy, avoiding meaningless state expansion and computational redundancy, and improving overall processing efficiency. Therefore, it can be well adapted to complex nuclear system operation and maintenance scenarios, efficiently and intelligently generating operation and maintenance reports with high reference value and low error at a relatively low data processing cost. These reports can then be used to effectively guide users in the safe operation and maintenance management of the nuclear system, ensuring its operational safety and lowering the user's operational threshold. Attached Figure Description

[0017] To more clearly illustrate the embodiments of this specification, the accompanying drawings used in the embodiments will be briefly introduced below. The drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an embodiment of a kernel-based operation and maintenance data processing method provided in this specification. Figure 2 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 3This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 4 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 5 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 6 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 7 This is a schematic diagram of the structural composition of a server provided in one embodiment of this specification; Figure 8 This is a schematic diagram of the structural composition of a kernel-based operation and maintenance data processing device provided in one embodiment of this specification; Figure 9 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 10 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 11 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 12 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 13 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 14 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 15 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 16 This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Figure 17This is a schematic diagram illustrating one embodiment of the operation and maintenance data processing method based on a kernel system provided in this specification, applied in a scenario example. Detailed Implementation

[0019] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0020] It should be noted that the information and data related to users involved in the embodiments of this specification are all information and data authorized by the user or fully authorized by the relevant parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with relevant laws, regulations, and standards, and necessary confidentiality measures have been taken. They do not violate public order and good morals, and corresponding operation entry points are provided for users or relevant parties to choose to authorize or refuse.

[0021] It should also be noted that in the embodiments of this specification, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, they do not mean that the applicant has used or necessarily used the solution.

[0022] See Figure 1 As shown in the embodiments of this specification, a method for processing operation and maintenance data based on a kernel system is provided. Specifically, this method may include the following: S101: Obtain the description text of the target core system; S102: The first intelligent agent extracts the component parameters of the target core system based on the description text; and determines the success criteria for the target core system. S103: Generate transfer rules for the target core system based on the component parameters of the target core system; S104: Determine and, based on the current working conditions, identify a matching target exploration model from the preset exploration models; wherein the preset exploration models include at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; S105: Using the target exploration model, the target core system's component parameters, success criteria, and transition rules are used to perform corresponding exploration processing to obtain the target core system's state transition diagram. S106: The second intelligent agent generates an operation and maintenance report of the target core system based on the state transition diagram.

[0023] The aforementioned target nuclear system can be understood as any system involving nuclear power equipment that requires safe operation and maintenance management. Specifically, for example, the target nuclear system could be a nuclear reactor coolant system or a nuclear power equipment cooling water system, etc. It should be noted that the target nuclear systems listed above are merely illustrative. In actual implementation, depending on the specific circumstances and processing requirements, the aforementioned target nuclear system may also include other types of equipment systems. This specification does not limit this.

[0024] The aforementioned descriptive text can specifically be text data input by a user (e.g., a technician) to describe aspects of the target core system, such as system architecture, component functions, and connectivity.

[0025] The aforementioned first intelligent agent can be understood as an algorithmic model that is pre-trained based on a large language model and can automatically extract component parameters of the kernel system from the descriptive text; and automatically generate success criteria of the kernel system based on the component parameters of the kernel system.

[0026] Specifically, the aforementioned component parameters can be understood as parameter data effective for the operation and maintenance management of the kernel system, and may include at least one of the following: component quantity, component identifier, component type, connection path, and standby startup failure rate (for example, it can be denoted as...). Failure rate (or failure rate, which can be denoted as ) ), Repair success rate (or repair rate, for example, it can be denoted as ), Common cause failure rate (for example, it can be denoted as...) ), system structure, etc.

[0027] The aforementioned common-cause faults differ from conventional individual faults. Specifically, they refer to faults in the core system where multiple specified component groups fail simultaneously due to a common fault source caused by special internal relationships (which can be denoted as CCF). The reason for separating common-cause faults from conventional individual faults is that the two types of faults differ significantly when performing state transition analysis based on a controlled transition system. By considering them separately, a more accurate state transition diagram can be generated.

[0028] Furthermore, the above system structure may specifically include at least one of the following: series structure, parallel structure, k-n structure, one-use-one-standby structure, one-use-multiple-standby structure, etc.

[0029] It should be noted that, considering the complexity of nuclear system structure and the high requirements for nuclear system operation safety, most nuclear systems adopt k-out-of-n structure, one-in-one-backup structure, or one-in-multiple-backup structure in whole or in part.

[0030] Specifically, for example, based on a k-choose-n structure, the core system can be configured with n devices having corresponding functions. As long as at least k of these devices are operating normally, the overall operation of the core system can be ensured. Based on a one-use-one-standby structure, under normal circumstances, the primary device is running while the standby device is in standby mode. When the primary device malfunctions, the standby device is activated, and when it successfully starts, the overall operation of the core system can be maintained. Based on a one-use-multiple-standby structure, under normal circumstances, the primary device is running while multiple standby devices are in standby mode, and each standby device has a corresponding startup priority. When the primary device malfunctions, the standby devices are activated sequentially according to their startup priorities. When one standby device successfully starts, the overall operation of the core system can be maintained.

[0031] The aforementioned success criteria can be understood as judgment rules corresponding to the nuclear system, used to determine whether the nuclear system as a whole is operating normally. Specifically, the aforementioned success criteria can integrate sub-criteria related to the system architecture of the nuclear system.

[0032] In practice, the descriptive text about the target core system input by the user can be obtained first; the first intelligent agent can process the descriptive text to extract text elements related to the operation and maintenance of the core system; and then the preset core system parameter template and the text elements can be used together to extract the required component parameters.

[0033] Furthermore, the first intelligent agent can determine the overall system structure and local system structure of the target system based on the component parameters; then, based on the preset configuration rules, it can determine the logical expressions corresponding to the overall system structure and local system structure respectively; and by combining the component parameters and logical expressions, it can configure the corresponding success criteria.

[0034] The aforementioned preset nuclear system parameter template can be a parameter template generated in advance by performing correlation analysis and cluster learning on a large amount of historical operation and maintenance data of nuclear systems.

[0035] In practice, a corresponding component parameter configuration interface can be built based on a preset kernel system parameter template; then, the structured component parameters input by the user can be directly obtained by showing the component parameter configuration interface to the user.

[0036] In practice, based on the preset configuration rules, for a series structure, an AND connection expression can be determined and used as the corresponding logical expression. Specifically, a series structure can be converted into an AND connection expression indicating that all devices are "normal" simultaneously, for example, ((state['A']=='normal') and (state['B']=='normal')).

[0037] According to the preset configuration rules, for parallel structures, an OR connection expression can be determined and used as the corresponding logical expression. Specifically, a parallel structure can be converted into an OR connection expression that accepts "any normal state", such as ((state['A']=='normal') or (state['B']=='normal')).

[0038] According to the preset configuration rules, for the k-to-n structure, a combined enumeration expression can be determined and used as the corresponding logical expression. Specifically, the k-to-n structure can list all combinations of "randomly selecting k devices to function normally simultaneously," with expressions connected by AND within each combination and expressions connected by OR between combinations. Note that the sum() expression is prohibited.

[0039] According to the preset configuration rules, for a one-use-one-standby structure and a one-use-multiple-standby structure, a sequential combination enumeration expression can be determined and used as the corresponding logical expression. Specifically, based on the combination enumeration expression, the various combinations can be concatenated in sequence according to the principle of "primary first, then standby" and the startup priority of the standby devices. The aforementioned one-use-one-standby structure and one-use-multiple-standby structure can also be collectively referred to as a primary-standby structure.

[0040] In practice, after determining the success criteria for the target core system using the first intelligent agent, the first intelligent agent can further convert the success criteria into a corresponding single-line Python Boolean expression (e.g., System_working( Specifically, the Boolean expressions mentioned above can be used to analyze specific states (e.g., When the return value is "Yes" or "True", it indicates that the status is normal; when the return value is "No" or "False", it indicates that the status is faulty.

[0041] The aforementioned transfer rules can be understood as rule data configured based on the expression of a controlled transfer system to describe the transfer mechanism between different states of the nuclear system.

[0042] Specifically, for example, the above transfer rule can be expressed in the following form: .

[0043] Among them, Event (abbreviated as E) is used to identify the event type, such as fault, repair, standby activation, common cause failure, etc. Guard (abbreviated as G) is used to indicate the current state of the event (e.g., How is it triggered? An Action (which can be denoted as A) is used to indicate updating the current state to the next state (or successor state, for example, ...). Rate (transition rate) is used to represent the probability of a successful transition from the current state to the successor state.

[0044] In practice, a knowledge graph about the relationships between components in the target core system can be constructed first based on the component parameters of the target core system. Then, based on the knowledge graph, the types of state transitions that exist in the state transition process of the target core system can be determined. Based on the knowledge graph, the types of state transitions, and the success criteria, the transition rules for the target core system can be constructed.

[0045] The aforementioned transfer rules may include multiple transfer rules, each of which corresponds to at least one state transfer type. Specifically, the aforementioned transfer rules may include at least one of the following: fault rules, repair rules, backup activation rules, and common-cause fault rules.

[0046] For details, please refer to Figure 2 As shown, the above failure rules can be used to represent component i with a failure rate. A failure leads to a state transition. The aforementioned repair rules can be used in component i to increase the success rate of repair. A successful repair results in a state transition. Standby activation rules are used to indicate when a standby device is activated to mitigate startup failures in the event of a primary device failure. Successful or failed startup results in a state transition. Common Cause Failure (CCF) rules can be used to represent common sources of failure, expressed as a common cause failure rate. This can cause multiple specified component groups to fail simultaneously, resulting in a state transition.

[0047] The first exploration model mentioned above can be understood as an algorithm model based on a complete exploration mode, which can traverse all states of the kernel system and generate a complete state transition diagram.

[0048] The aforementioned second exploration model can be specifically understood as an algorithm model that, based on a local exploration mode, selectively and purposefully retrieves some key states of the kernel system and generates a local state transition diagram.

[0049] The aforementioned operation and maintenance report may specifically include one or more of the following operation and maintenance reference data: operational security risks of the target core system, high-risk and vulnerable components in the target core system, information on high-risk states and state transitions of the target core system, and operation and maintenance strategies and recommendations for the target core system.

[0050] The aforementioned second intelligent agent can be understood as an algorithmic model that is pre-trained based on a large language model and can automatically analyze and process the state transition diagram of the nuclear system. By analyzing and based on the weak components and risk states in the nuclear system, it can automatically generate operation and maintenance reports to guide the operation and maintenance of the nuclear system.

[0051] The state transition diagram described above includes at least a set of states and a set of transitions. Specifically, it can be represented as: .in, A state set can contain at least several states indicating that the target kernel system is operating normally, with corresponding state labels. This represents a transition set, specifically containing transition relationships based on a state in the state set, from one state to the next state (or successor state). The state labels can include labels indicating a normal state and labels indicating a faulty state. The transition relationships must at least include the state identifier of the starting state, the state identifier of the successor state, the transition rate from the starting state to the successor state, and the transition rule upon which the transition from the starting state to the successor state is based. Specifically, for example, any transition relationship can be represented in the following form: Based on the above transition relationships, it can be known that the transition path is the state. Transition to state The transfer rate is rate(e); where e is the transfer rule (which can also be denoted as r).

[0052] Furthermore, the aforementioned transition relationship may also include a transition path priority parameter. This transition path priority parameter represents the probability that, when a starting state has multiple transitionable successor states, the user will transition to different successor states through their respective transition paths.

[0053] Specifically, for example, the state transition diagram mentioned above may include a Markov state transition diagram.

[0054] Furthermore, the state transition diagram described above may also include a set of absorbing states (which can be denoted as...). Specifically, the aforementioned absorption state set includes at least states with state labels indicating state failures. Furthermore, the aforementioned absorption state set may also include truncated states.

[0055] In specific implementation, the current operating scenario of the target core system can be determined first by acquiring and using the component parameters of the target core system and / or the operation and maintenance requirements parameters of the target core system input by the user. Then, based on the current operating scenario, a matching exploration model is determined from the first exploration model and the second exploration model as the target exploration model. Next, the initial state of the target core system is determined based on the component parameters of the target core system. In this initial state, all components in the target core system are operating normally. Then, starting from the initial state of the target core system, the target exploration model is used to perform multiple rounds of exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain the state transition diagram of the target core system. The state transition diagram includes a complete state transition diagram or a partial state transition diagram.

[0056] Furthermore, a second intelligent agent can be used to extract key parameters by processing the state transition diagram. These key parameters include path priority parameters, activity column parameters, state labels, and absorption state information. Based on these key parameters, a maintenance report for the target core system can be generated by performing specific fault analysis and prediction on the target core system.

[0057] Based on the above embodiments, firstly, by introducing and using a first intelligent agent to automatically extract and determine component parameters and success criteria of the target core system based on the user-input description text, the user's operational threshold can be effectively reduced, the success criterion generation process can be simplified, and the subjective human error introduced by relying on manual configuration of success criteria by the user can be reduced. Secondly, by identifying and differentiating different situations according to the current working scenario, selecting and using a matching target exploration model for multi-mode differentiation and targeted exploration processing, a high-quality state transition diagram can be efficiently generated while ensuring data processing accuracy. This effectively reduces the overall data processing volume, avoids meaningless state expansion and computational redundancy, and improves overall processing efficiency. Furthermore, by introducing and using a second intelligent agent to automatically perform fault analysis and prediction based on the generated state transition diagram, a high-value operation and maintenance report for the target core system can be automatically generated. This reduces the user's operational threshold, simplifies the state transition diagram analysis and processing process, and efficiently and accurately generates a compliant operation and maintenance report.

[0058] In some embodiments, the component parameters may specifically include at least one of the following: number of components, component identifier, component type, connection path, backup startup failure rate, failure rate, repair success rate, common cause failure rate, system structure, etc. The system structure may specifically include at least one of the following: series structure, parallel structure, k-n structure, one-use-one-standby structure, one-use-multiple-standby structure, etc.

[0059] It should be noted that the component parameters and system structures listed above are only illustrative. In actual implementation, other component parameters and other system structures may be included depending on the specific circumstances and processing requirements. This specification does not limit this.

[0060] In some embodiments, the transfer rules may specifically include at least one of the following: fault rules, repair rules, backup activation rules, common cause fault rules, etc.

[0061] In some embodiments, see Figure 3 As shown above, based on the current working conditions, a matching target exploration model is determined from the preset exploration models. In specific implementation, this may include the following: S3-1: Determine the number of components in the target core system and / or the operation and maintenance requirements parameters for the target core system; S3-2: Determine the current operating scenario of the target core system based on the number of components and / or operation and maintenance requirements parameters; S3-3: Based on the current working conditions, determine a matching target exploration model from the preset exploration models.

[0062] Specifically, the aforementioned operation and maintenance requirements parameters may include the operation and maintenance security level and / or the operation and maintenance management precision. These parameters can be customized by the user based on specific circumstances and processing needs.

[0063] In practice, the size of the state space of the target system can be estimated based on the number of components in the target system; the size of the state space can be checked to see if it is less than a preset size threshold, and the security level of operation and maintenance can be checked to see if it is higher than a preset security level threshold, and / or the accuracy of operation and maintenance management can be checked to see if it is higher than a preset accuracy threshold.

[0064] When the state space size is smaller than a preset size threshold, the operation and maintenance security level is higher than a preset security level threshold, and the operation and maintenance management accuracy is higher than a preset accuracy threshold, the current operating scenario can be determined as the first operating scenario. Therefore, the first exploration model can be determined as the matching target exploration model. In this way, the first exploration model can be used to explore the entire state space, generating a complete state transition diagram for accurate operation and maintenance analysis of the target core system.

[0065] Otherwise, the current operating scenario is identified as the second operating scenario, and the second exploration model can then be determined as the matching target exploration model. In this way, the second exploration model can be used to explore some important state spaces and generate local state transition diagrams, so as to simultaneously balance accuracy and efficiency in the operation and maintenance analysis of the target core system.

[0066] In this way, by simultaneously considering the objective dimensions of the target core system itself and the subjective needs of the user in combination with specific circumstances, the current working conditions can be finely distinguished, thereby accurately determining the appropriate target exploration model.

[0067] In some embodiments, when the target exploration model includes a first exploration model, the target exploration model is used to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain the state transition diagram of the target core system. (See reference...) Figure 4 As shown, in specific implementation, the exploration process for the current round based on the full mode can be performed as follows: S4-1: Obtain the state of the target kernel system in the previous round; S4-2: Using the target exploration model, determine the state label of the previous round's state based on the success criteria and component parameters of the target core system; S4-3: When the status label of the previous round indicates that the status is normal, determine the status of the current round according to the status of the previous round and the transition rules; and calculate the transition rate of the current round from the status of the previous round to the status of the current round. S4-4: Update the state set based on the state of the previous round and the corresponding state label; and update the transition set based on the transition rate of the current round.

[0068] In practice, based on the success criteria of the target core system and in conjunction with component parameters, it can be determined whether there was a fault in the previous state of the target core system, and the corresponding state label of the previous state can be generated.

[0069] Specifically, if there is no fault and the target core system can operate normally, a status label indicating that the target core system was in normal condition in the previous round is generated; conversely, if there is a fault and the target core system cannot operate normally, a status label indicating that the target core system was in fault condition in the previous round is generated.

[0070] Specifically, when the state label of the previous round indicates a state failure, the exploration process based on the previous round's state can be terminated; the previous round's state is saved in the absorption state set; simultaneously, the detection state set (which can be denoted as...) is... If a state from the previous round exists in the state set, then remove that state from the state set.

[0071] Furthermore, the candidate state set can be searched again to determine if there are any unexplored remaining states. If so, the remaining states are acquired and used to initiate the next round of exploration. If no remaining states are found, i.e., the candidate state set is empty, the exploration process ends, and the current state set and transition set are combined to obtain the state transition diagram of the target kernel system.

[0072] When the state label of the previous state is normal, the successor state of the previous state can be determined based on the previous state and transition rules. The system then checks if the number of successor states is one. If it is, this successor state is used as the current state. The transition rate function is then called to calculate the transition rate from the previous state to the current state based on component parameters and transition rules. The previous state and its corresponding state label are then stored in a state set to update the state set. Simultaneously, using the current transition rate, a transition relationship from the previous state to the current state is constructed and stored in a transition set to update the transition set. Furthermore, states from the previous state can be detected and removed from the subsequent state set.

[0073] Furthermore, following the above method, based on the state of the current round, we can proceed with the exploration process for the next round regarding the state of the current round.

[0074] Specifically, when the number of successor states of the previous round is greater than 1, a successor state can be randomly selected from multiple successor states as the current round's state; and the remaining successor states are stored in the candidate state set. After traversing all state links based on the successor state through the above exploration process, the remaining unexplored successor states are extracted from the candidate state set, and the above process is repeated to explore the remaining successor states until the candidate state set is empty. At this point, the exploration process ends, and the current state set and transition set are combined to obtain the state transition diagram of the target core system.

[0075] Based on the above embodiments, by using the first exploration model as the target exploration model, it is possible to better match working scenarios with a small state space but a high level of security and / or high requirements for operation and maintenance management accuracy. With a relatively reasonable amount of data processing, the exploration and traversal of each state transition process of the target core system can be completed comprehensively and completely, thereby obtaining a more comprehensive and accurate complete state transition diagram.

[0076] Specifically, the first exploration model mentioned above can be an exploration model based on an improved breadth-first search (BFS).

[0077] In some embodiments, when the target exploration model includes a second exploration model, the target exploration model is used to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain the state transition diagram of the target core system. (See reference...) Figure 5 As shown, in specific implementation, the current round of exploration based on local mode can be processed in the following way: S5-1: Obtain the previous state of the target kernel system; S5-2: Detect whether the link length of the state link between the previous state and the initial state of the target core system is less than the preset budget threshold. S5-3: When the link length of the state link between the previous state and the initial state of the target core system is less than the preset budget threshold, the state label of the previous state is determined by the target exploration model based on the success criteria and component parameters of the target core system. S5-4: When the state label of the previous round indicates that the state is normal, determine the successor state of the previous round based on the previous round state and transition rules; and calculate the path priority parameter from the previous round state to the successor state. S5-5: Determine the state of the current round from the successor states based on the path priority parameter; and calculate the transition rate from the state of the previous round to the state of the current round. S5-6: Update the state set based on the state of the previous round and the corresponding state label; and update the transition set based on the transition rate of the current round and the path priority parameter of the current round's state.

[0078] The aforementioned preset budget threshold (which can be denoted as S) can be a cutoff value determined based on the number of components in the target core system and / or the operation and maintenance requirements parameters of the target core system, or it can be a value customized by the user according to specific circumstances and needs.

[0079] The link length of the state link between the previous state and the initial state of the target core system can specifically refer to the number of states between the initial state and the previous state, or the number of transition paths.

[0080] In addition, a state upper limit value for the state transition diagram can be determined based on a preset budget threshold; and during the exploration process, the total number of states contained in the generated state transition diagram is constrained to be less than or equal to this state upper limit value.

[0081] In practice, it can be first checked whether the link length of the state link between the previous state and the initial state of the target core system is less than the preset budget threshold.

[0082] When the link length between the previous state and the initial state of the target core system is greater than or equal to a preset budget threshold, the previous state can be considered a truncated state. Under normal circumstances, no further exploration processing is performed based on the previous state. In this case, the previous state can be saved to the absorption state set. Then, check if the subsequent state set is empty. If it is empty, end the exploration process, combine the current state set and transition set to obtain the state transition diagram of the target core system. If it is not empty, randomly extract a state from the subsequent state set and perform further exploration processing based on that state.

[0083] When the link length of the state link between the previous state and the initial state of the target core system is less than the preset budget threshold, it can be determined that the previous state is not a truncated state, and further exploration processing can be carried out based on the previous state.

[0084] Specifically, the target exploration model can be used to determine the state label of the previous round of state based on the success criteria and component parameters of the target core system.

[0085] When the state label of the previous round indicates a state failure, the exploration process based on the previous round's state can be terminated; the previous round's state is saved in the absorption state set; simultaneously, it is checked whether the previous round's state exists in the state set, and if it does, it is removed from the state set. The subsequent state set is then searched to determine if there are any remaining unexplored states.

[0086] When the state label of the previous state indicates that the state is normal, the successor state of the previous state can be determined based on the previous state and the transition rules, and the path priority parameter from the previous state to the successor state can be calculated.

[0087] Specifically, when the number of successor states of the previous state is 1, the path priority parameter d from the previous state to the successor state can be set to 1. In this case, the successor state can be directly determined as the state of the current round.

[0088] When the number of successor states of the previous state is greater than 1, it means that the previous state can be transitioned to multiple different successor states (i.e., successor states) through multiple different transition paths. In this case, we can calculate the transition rate from the previous state to each successor state; then, based on the transition rate from the previous state to each successor state, we can calculate the path priority parameters (which can be understood as the path priority parameters of the transition paths) for each successor state.

[0089] Specifically, for example, the path priority parameter can be calculated using the following formula:

[0090] in, The state of the previous round To the successor state numbered 0 Path priority parameter, The state of the previous round To the successor state numbered j The transition rate is given by j, where j is the number of the successor state from which the state in the previous round can directly transition, and p is the number of successor states from the state in the previous round. The state of the previous round To the successor state numbered 0 The transfer rate.

[0091] After determining the current state from the successor states, the path priority parameter of the previous state can be obtained. Then, by multiplying the path priority parameter of the previous state with the path priority parameter from the previous state to the current state, the path priority parameter of the current state can be determined.

[0092] Based on the path priority parameters determined by the above method, the larger the value, the higher the security risk of the state to which the transfer path is transferred, the greater the importance of the transfer and the greater the tendency to occur, and the more likely it is to be explored and processed.

[0093] Based on the path priority parameter, the successor state with the largest path priority parameter is determined from multiple successor states and used as the state of the current round; and the other successor states, along with their corresponding path priority parameters, are stored in the set of selected successor states.

[0094] Next, calculate the transition rate from the previous state to the current state. Then, store the previous state and its corresponding state label into the state set to update the state set. Simultaneously, using the current state's transition rate, construct the transition relationship from the previous state to the current state; store this transition relationship into the transition set to update the transition set. Furthermore, states from the previous state can be detected and removed from the subsequent state set.

[0095] Furthermore, following the above method, based on the state of the current round, we can proceed with the exploration process for the next round regarding the state of the current round.

[0096] It should be added that the path priority parameter of the successor state in the subsequent state set can be detected to see if it is greater than the preset priority parameter threshold. Based on the detection results, the successor states in the subsequent state set whose path priority parameter is less than or equal to the preset priority parameter threshold are removed, and the subsequent states in the removed subsequent state set are further explored and processed.

[0097] Based on the above embodiments, by using the second exploration model as the target exploration model, it is possible to better match working scenarios with a large state space and / or low security level and low operation and maintenance management accuracy requirements. It can selectively and purposefully complete the exploration of some state transition processes of the target core system, thereby generating a local state transition diagram with high reference value intelligently and efficiently with a small data processing cost.

[0098] Specifically, the second exploration model mentioned above can be an exploration model based on an improved priority queue search (Dijkstra).

[0099] In some embodiments, after detecting whether the link length of the state link between the previous state and the initial state of the target core system is less than a preset budget threshold, refer to Figure 6 As shown, in specific implementations, the method may also include the following: S6-1: When the link length of the state link between the previous state and the initial state of the target core system is greater than or equal to the preset budget threshold, extract the state evolution features based on the state link between the previous state and the initial state of the target core system. S6-2: Based on the state evolution characteristics, detect whether the adaptive extension condition with respect to the preset budget threshold is satisfied; S6-3: When it is determined that the adaptive extension condition is met, a matching extension adjustment value is determined according to the state evolution characteristics; S6-4: Based on the aforementioned extension adjustment value, dynamically update the preset budget threshold for the state link between the previous state and the initial state of the target core system.

[0100] In practice, the states contained in the state link between the previous state and the initial state of the target core system can be mapped to corresponding feature codes according to the corresponding coding mapping rules; then, the above feature codes are concatenated in order to obtain the corresponding feature code sequence, which serves as the state evolution feature.

[0101] Then, based on the state evolution characteristics, the preset successor state risk prediction model is used to predict whether there is a critical risk state in the successor state of the state link whose security risk is higher than the preset risk threshold; when it is determined that there is a critical risk threshold, the adaptive extension condition is satisfied.

[0102] If the adaptive extension conditions are met, the matching extension adjustment value can be determined and utilized based on the state evolution characteristics. Further exploration and processing can be carried out based on this state link to further explore and analyze the key risk states that have a significant impact on the operation of the target core system, so as to perform more accurate operation and maintenance data processing on the target core system.

[0103] Among them, the aforementioned preset successor state risk prediction model can be a prediction model based on a hybrid network structure of Transformer structure and bidirectional GR structure.

[0104] The Transformer structure described above can be understood as a deep learning model architecture that primarily replaces the traditional Recurrent Neural Network (RNN) with a self-attention mechanism. Its core consists of encoder and decoder components. By introducing the Transformer structure into the model, it gains a better ability to capture global dependencies. This allows the model to analyze the temporal sequence of state features of the currently analyzed states in a state chain—that is, the state evolution features—to determine the global feature information of the unanalyzed subsequent states in that state chain based on the overall state evolution trend.

[0105] The aforementioned bidirectional GRU (Bidirectional Gated Recurrent Unit) structure can be understood as an extension of the gated recurrent unit, capable of simultaneously acquiring and processing both positive information (feature encoding of states following a single state) and negative information (feature encoding of states preceding a single state) of state evolution characteristics. By introducing this bidirectional GRU structure into the model, on the one hand, the model can determine the long-term change trend of subsequent states in a state chain from a holistic perspective, based on global feature information and combining historical and future interactions; on the other hand, the model can capture the local dynamic feature information of a state chain based on a single state, based on the feature changes of several adjacent states.

[0106] In this way, by integrating the Transformer structure and the bidirectional GRU structure into the model, we can simultaneously acquire and, based on the global feature information of the state link, as well as the hidden, relatively complex long-range change trends and short-term local dynamic feature information, predict the state characteristics of the subsequent state of the simulated state link and predict the related risk situation.

[0107] In practical implementation, based on the state evolution characteristics and the prediction results output by the preset subsequent state risk prediction model based on the state evolution characteristics, the number of interval states that can transition to the key risk state of concern can be determined as a matching delay adjustment value. Then, based on the delay adjustment value, a preset budget threshold for the state link between the previous state and the initial state of the target core system can be dynamically updated; and based on this preset budget threshold, further exploratory processing can be performed on this state link to obtain a local transition state diagram with relatively higher reference value and better results.

[0108] In this way, by introducing and utilizing the aforementioned pre-defined successor state risk prediction model, it can be accurately and automatically determined whether the adaptive extension condition is met.

[0109] In some embodiments, after detecting whether an adaptive extension condition with respect to a preset budget threshold is met, the method may further include the following: S1: When it is determined that the adaptive extension condition is not met, the exploration process based on the state of the previous round ends; S2: Update the state set based on the state of the previous round and the corresponding state label; S3: Check if the exploration termination condition is met; S4: When the exploration termination condition is met, determine the state that is not included in the state set and take it as the absorption state; S5: Save the absorption state to the absorption state set; and update the transition set according to the absorption state set.

[0110] In practice, the state of the previous round and its corresponding state label can be saved to a state set to update the state set. Then, the candidate state set is checked to determine if it is empty; if it is empty, the exploration termination condition can be determined.

[0111] If the candidate state set is not empty, that is, there are still unexplored states (referred to as potential states) in the candidate state set; further, query and determine, based on the current exploration and processing records, whether the link length of the state link between the potential state and the initial state of the target core system is greater than or equal to a preset budget threshold; when it is determined that it is greater than the preset budget threshold, it can be determined that the exploration termination condition is met.

[0112] When the exploration termination condition is met, the exploration process can be terminated. Based on the current exploration process record, the fault states indicated by the status labels are identified; at the same time, the remaining states (i.e., truncated states) in the candidate state set are extracted and determined to be not included in the state set; these are identified as absorbing states; and the absorbing states are then saved to the absorbing state set.

[0113] In addition, based on the absorption states contained in the absorption state set, the transfer relationships that originally transferred to the above-mentioned fault state or truncation state in the original transfer set can be uniformly modified to transfer to the absorption state, thereby realizing the update of the transfer set.

[0114] Finally, by combining and using the updated state set and the updated transition set, a state transition graph is constructed.

[0115] In some embodiments, the above-described method of generating an operation and maintenance report for the target core system using a second intelligent agent based on the state transition diagram may specifically include the following: S1: The second agent processes the state transition graph to extract key parameters; wherein, the key parameters include: path priority parameters, active column parameters (e.g., ActiveTrain), state labels, and absorption state information; S2: Determine the risk transfer path and risk components based on the aforementioned key parameters; S3: Determine the risk pattern type based on the risk transfer path and risk components; S4: Generate an operation and maintenance report for the target core system based on the risk mode type.

[0116] The aforementioned risk mode types may specifically include at least one of the following: independent failure, backup startup failure, common cause failure, independent repair, etc.

[0117] Specifically, the above-mentioned independent failure (single component "normal → failure"): if only one component X changes from "normal" to "failure", then the transition is classified as X independent failure.

[0118] The above repair (single component "fault → normal"): If only one component X changes from "fault" to "normal", it is classified as X repair.

[0119] The above-mentioned standby startup failure (primary failure triggers standby status change, which may be accompanied by changes in multiple components): If the primary equipment failure is involved at the same time and the standby equipment is enabled, it is classified as primary-standby switchover / standby enablement; if the primary equipment failure is shown and the standby also becomes faulty, it is further classified as startup failure.

[0120] The above common-cause failure: If multiple components change from normal to failure at the same time, and this group of components is declared as a CCF group (or the corresponding rule event name is common_cause_fail) in the input file, it is classified as "CCF multi-component common-cause failure".

[0121] In practical implementation, the second intelligent agent can calculate the risk score of each component and the risk score of the transfer path based on the aforementioned key parameters and the component parameters of the target core system. Then, based on the risk scores of the components and the transfer paths, risky components and risky transfer paths are selected. Finally, based on the risky components and risky transfer paths, risk pattern type matching is performed using preset risk pattern type instances to determine the corresponding risk pattern type. The preset risk pattern type instances are representative template instances compiled from a large amount of historical test data through risk pattern type clustering.

[0122] When calculating the risk score, the transition paths with potential risks can be screened based on the component parameters; then, the risk score of the above transition paths can be calculated based on the state transition diagram.

[0123] Specifically, when the state transition diagram is generated based on the first exploration model through exploration processing, since the state transition diagram does not contain path priority parameters, the risk score of the transition path can be obtained and calculated based on the transition rate of the transition path. When the state transition diagram is generated based on the second exploration model through exploration processing, since the state transition diagram also contains path priority parameters, the risk score of the transition path can be obtained and calculated based on the transition rate and path priority parameters of the transition path.

[0124] Then, based on the risk score and key parameters of the transfer path, combined with the priority parameters of the transfer path in the relevant state, the components with potential risks are identified by analyzing and screening the components related to the transfer rules of the above transfer path; and the risk score of the component is calculated by statistically analyzing and based on the risk score of the transfer path related to the above component.

[0125] In practice, when generating the operation and maintenance report of the target core system, the system can detect whether there are any operational risks in the target core system as a whole based on the operation and maintenance report. If operational risks are found to exist, the system can further locate the risky components and / or risk transfer paths that cause the operational risks. Then, based on the risky components and / or risk transfer paths, the system can generate operation and maintenance strategies or recommendations to eliminate the above-mentioned operational risks, so that users can maintain and manage the target core system based on the above-mentioned operation and maintenance strategies or recommendations to ensure the safe operation of the target core system.

[0126] Based on the above embodiments, a second intelligent agent can be used to automatically generate operation and maintenance reports for the target core system that have high reference value and good results by processing the state transition diagram.

[0127] In some embodiments, the first and second agents described above can be trained by jointly using the Dify platform and a large language model.

[0128] As can be seen from the above, the operation and maintenance data processing method based on the core system provided in this specification can be trained using a large language model to obtain a first intelligent agent and a second intelligent agent that meet the requirements before specific implementation. In specific implementation, firstly, the description text of the target core system is obtained; the first intelligent agent extracts the component parameters of the target core system based on the description text; and determines the success criteria for the target core system; based on the component parameters of the target core system, a transition rule for the target core system is generated; a matching target exploration model is determined from the preset exploration models based on the current working scenario; wherein the preset exploration model includes at least: a first exploration model based on a complete exploration mode and a second exploration model based on a local exploration mode; the target exploration model performs corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain a state transition diagram of the target core system; the second intelligent agent generates an operation and maintenance report of the target core system based on the state transition diagram. On the one hand, by introducing and using first and second intelligent agents to assist in the processing of nuclear system operation and maintenance data, the data processing process can be effectively simplified and the processing difficulty reduced. On the other hand, by identifying and differentiating different situations based on the current operating scenario, and using matching target exploration models for multi-mode differentiation and targeted exploration processing, the overall data processing volume can be effectively reduced while ensuring data processing accuracy, avoiding meaningless state expansion and computational redundancy, and improving overall processing efficiency. Therefore, it can be well adapted to complex nuclear system operation and maintenance scenarios, and can efficiently and intelligently generate operation and maintenance reports with high reference value and low error at a relatively low data processing cost. These reports can then be used to effectively guide the safe operation and maintenance management of the nuclear system, ensuring its operational safety.

[0129] This specification provides an embodiment of a server, see below. Figure 7 As shown. The server includes a network communication port 701, a processor 702, and a memory 703. These structures are connected by internal cables so that they can perform specific data interaction.

[0130] Specifically, the network communication port 701 can be used to obtain the description text of the target core system.

[0131] The processor 702 can specifically be used to: extract component parameters of the target core system using a first intelligent agent based on the description text; determine success criteria for the target core system; generate transition rules for the target core system based on the component parameters; determine and identify a matching target exploration model from a preset exploration model based on the current operating scenario; wherein the preset exploration model includes at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; perform corresponding exploration processing using the target exploration model based on the component parameters, success criteria, and transition rules of the target core system to obtain a state transition diagram of the target core system; and generate an operation and maintenance report of the target core system using a second intelligent agent based on the state transition diagram.

[0132] The memory 703 can be used to store the corresponding instruction program and related intermediate data.

[0133] Based on the above methods, the relevant structural performance of the server can be effectively utilized to improve the data processing speed of electronic devices and efficiently realize the operation and maintenance data processing of the kernel-based system.

[0134] In this embodiment, the network communication port 701 can be a virtual port bound to different communication protocols, thereby enabling the sending or receiving of different data. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. Furthermore, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM or CDMA; it can also be a Wi-Fi chip; or it can be a Bluetooth chip.

[0135] In this embodiment, the processor 702 can be implemented in any suitable manner. For example, the processor can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers, etc. This specification is not limiting.

[0136] In this embodiment, the memory 703 may include multiple layers. In a digital system, anything that can store binary data can be a memory. In an integrated circuit, a circuit with storage function but no physical form is also called a memory, such as RAM, FIFO, etc. In a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0137] This specification also provides a computer-readable storage medium based on the above-described operation and maintenance data processing method for a core system. The computer-readable storage medium stores computer program instructions that, when executed, implement the following: obtaining a description text of a target core system; using a first intelligent agent to extract component parameters of the target core system based on the description text; determining success criteria for the target core system; generating transition rules for the target core system based on the component parameters; determining and identifying a matching target exploration model from a preset exploration model based on the current operating scenario; wherein the preset exploration model includes at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; using the target exploration model to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain a state transition diagram of the target core system; and using a second intelligent agent to generate an operation and maintenance report for the target core system based on the state transition diagram.

[0138] In this embodiment, the storage medium includes, but is not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Cache, Hard Disk Drive (HDD), or Memory Card. The memory can be used to store computer program instructions. The network communication unit can be an interface configured according to standards specified in the communication protocol for network connection communication.

[0139] In this embodiment, the specific functions and effects implemented by the program instructions stored in the computer-readable storage medium can be explained in comparison with other embodiments, and will not be repeated here.

[0140] This specification also provides a computer program product, comprising at least a computer program, which, when executed by a processor, implements the following method steps: obtaining a description text of a target core system; using a first intelligent agent to extract component parameters of the target core system based on the description text; determining success criteria for the target core system; generating transition rules for the target core system based on the component parameters; determining and identifying a matching target exploration model from a preset exploration model based on the current operating scenario; wherein the preset exploration model includes at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; using the target exploration model to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain a state transition diagram of the target core system; and using a second intelligent agent to generate an operation and maintenance report of the target core system based on the state transition diagram.

[0141] See Figure 8 As shown, at the software level, this specification also provides an operation and maintenance data processing device based on a kernel system, which may specifically include the following structural modules: The acquisition module 801 can be used to acquire the description text of the target core system. Processing module 802 can be specifically used to extract component parameters of the target core system based on the description text using the first intelligent agent; and to determine the success criteria for the target core system; The first generation module 803 can be specifically used to generate transfer rules for the target core system based on the component parameters of the target core system. The determination module 804 can be used to determine and, based on the current working conditions, identify a matching target exploration model from a preset exploration model; wherein the preset exploration model includes at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; The exploration module 805 can be used to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system using the target exploration model, and obtain the state transition diagram of the target core system. The second generation module 806 can be used to generate an operation and maintenance report of the target core system based on the state transition diagram using the second intelligent agent.

[0142] In some embodiments, the component parameters may specifically include at least one of the following: number of components, component identifier, component type, connection path, backup startup failure rate, failure rate, repair success rate, common cause failure rate, system structure, etc. The system structure may specifically include at least one of the following: series structure, parallel structure, k-n structure, one-use-one-standby structure, one-use-multiple-standby structure, etc.

[0143] In some embodiments, the transfer rules may specifically include at least one of the following: fault rules, repair rules, backup activation rules, common cause fault rules, etc.

[0144] In some embodiments, when the determination module 804 is specifically implemented, it can determine and determine the matching target exploration model from the preset exploration model according to the current working condition scenario in the following manner: determine the number of components of the target core system and / or the operation and maintenance requirements parameters of the target core system; determine the current working condition scenario of the target core system according to the number of components and / or the operation and maintenance requirements parameters; and determine the matching target exploration model from the preset exploration model according to the current working condition scenario.

[0145] In some embodiments, when the target exploration model includes a first exploration model, the exploration module 805, when specifically implemented, can perform the current round of exploration processing in the following manner: obtain the state of the target core system in the previous round; use the target exploration model to determine the state label of the previous round's state according to the success criteria and component parameters of the target core system; when the state label of the previous round's state indicates that the state is normal, determine the state of the current round according to the state of the previous round and the transition rules; and calculate the transition rate of the current round from the state of the previous round to the state of the current round; update the state set according to the state of the previous round and the corresponding state label; and update the transition set according to the transition rate of the current round.

[0146] In some embodiments, when the target exploration model includes a second exploration model, the exploration module 805, when specifically implemented, can perform the current round of exploration processing as follows: obtain the state of the target core system in the previous round; detect whether the link length of the state link between the previous round state and the initial state of the target core system is less than a preset budget threshold; when the link length of the state link between the previous round state and the initial state of the target core system is less than the preset budget threshold, use the target exploration model to determine the state label of the previous round state according to the success criteria and component parameters of the target core system; when the state label of the previous round state indicates that the state is normal, determine the successor state of the previous round state according to the previous round state and transition rules; and calculate the path priority parameter from the previous round state to the successor state; determine the state of the current round from the successor state according to the path priority parameter; and calculate the current round transition rate from the previous round state to the current round state; update the state set according to the previous round state and the corresponding state label; and update the transition set according to the current round transition rate and the path priority parameter of the current round state.

[0147] In some embodiments, after detecting whether the link length of the state link between the previous state and the initial state of the target core system is less than a preset budget threshold, the exploration module 805 may further be used to: extract state evolution features based on the state link between the previous state and the initial state of the target core system when the link length of the state link between the previous state and the initial state of the target core system is greater than or equal to the preset budget threshold; detect whether an adaptive extension condition with respect to the preset budget threshold is met based on the state evolution features; determine a matching extension adjustment value based on the state evolution features when the adaptive extension condition is met; and dynamically update the preset budget threshold for the state link between the previous state and the initial state of the target core system based on the extension adjustment value.

[0148] In some embodiments, after detecting whether the adaptive extension condition with respect to a preset budget threshold is met, the exploration module 805 may further be used to: end the exploration process based on the previous round's state when it is determined that the adaptive extension condition is not met; update the state set according to the previous round's state and the corresponding state label; detect whether the exploration end condition is met; when the exploration end condition is met, determine the state not included in the state set as the absorbing state; save the absorbing state to the absorbing state set; and update the transition set according to the absorbing state set.

[0149] In some embodiments, when the second generation module 806 is specifically implemented, it can generate an operation and maintenance report of the target core system using the second intelligent agent based on the state transition diagram in the following manner: The second intelligent agent processes the state transition diagram to extract key parameters; wherein, the key parameters include: path priority parameters, activity column parameters, state labels, and absorption state information; based on the key parameters, risk transfer paths and risk components are determined; based on the risk transfer paths and risk components, risk mode types are determined; and based on the risk mode types, an operation and maintenance report of the target core system is generated.

[0150] It should be noted that the units, devices, or modules described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above devices are described by dividing them into various modules according to their functions. Of course, in implementing this specification, the functions of each module can be implemented in one or more software and / or hardware, or the module that implements the same function can be implemented by a combination of multiple sub-modules or sub-units, etc. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection between the devices or units shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0151] As can be seen from the above, the operation and maintenance data processing device based on the nuclear system provided in the embodiments of this specification can be well adapted to complex nuclear system operation and maintenance scenarios. It can efficiently and intelligently generate operation and maintenance reports with high reference value and low error of the nuclear system with low data processing cost. In order to effectively guide the safety operation and maintenance management of the nuclear system and ensure the safe operation of the nuclear system, the operation and maintenance report can be used in the future.

[0152] In a specific scenario example, the operation and maintenance data processing method based on the kernel system provided in this manual can be applied to achieve intelligent generation of Markov state transition diagrams for the kernel system, as well as related operation and maintenance decisions. For detailed implementation procedures, please refer to the following content.

[0153] In this scenario example, considering complex engineering systems like nuclear systems, Markov chains / continuous-time Markov processes can be used for quantitative assessment of availability and failure risk to guide specific operation and maintenance management. However, such systems typically contain complex architectures with multiple components and redundancies, and exhibit dynamic and complex behavioral logic such as operation-failure-maintenance-recovery-switchover. Modeling often requires using multiple features such as failure rate, repair rate, or trigger probability to describe state transitions, further increasing complexity. In practical engineering implementation, the core steps are to start from the system's initial state, automatically generate the state set Σ and transition set Θ, then construct the transition rate matrix and calculate indicators such as availability curves and failure probabilities. To support online operation and maintenance, the state transition diagram and calculation results need to be further transformed into executable warning, maintenance, and redundancy switching strategies. Therefore, "state space generation—result output—operation and maintenance decision-making" constitutes the key link in practical applications.

[0154] However, due to the unique characteristics of nuclear systems, the following shortcomings exist in practical implementation: 1) Lack of intelligent model generation and update mechanisms: Most existing toolchains rely on manual intervention, such as manually setting fault and repair rules, resulting in a cumbersome and error-prone modeling process. Even with automated tools for generating some state transition diagrams, the results usually require extensive manual verification and correction. Existing model updates also lack efficient mechanisms; once the system or strategy changes, the model update process typically requires manual redefinition of states and transitions, leading to high update costs and a high risk of errors. 2) Inability to handle complex temporal logic and redundancy mechanisms: Existing technologies often cannot effectively handle systems with complex temporal logic, especially structures such as primary / backup redundancy, parallel redundancy, and column-level switching. In these systems, state transitions are not only driven by the faults and repairs of individual components but also constrained by system-level decision functions and switching logic. Existing Markov modeling tools struggle to uniformly express these complex mechanisms and logic within the same framework, resulting in inconsistencies between the generated transition diagrams and actual engineering logic. For example, in one-in-one redundancy and k-out-of-n redundancy systems, state transitions depend on component failures and repair sequences, which are complex and unintuitive to handle in existing toolchains. 3) Lack of operable decision support and automatic parsing capabilities: Existing state transition diagrams are usually exported in the form of text files or transition matrices. Engineers need to manually analyze these outputs to locate weak components, identify high-risk transition paths, and determine switching trigger conditions. This process heavily relies on expert experience, is time-consuming and error-prone, and lacks the ability to automatically parse text outputs and generate strategies, making it difficult to achieve a closed loop of modeling-output-operational recommendations. In addition, existing operation and maintenance decision support systems usually lack traceability and interpretability of the generated Markov state transition diagrams, making it difficult for operation and maintenance personnel to quickly understand and implement relevant recommendations. 4) LLM lacks operability and stability in engineering practice: Although many existing prototypes based on Large Language Models (LLMs) simplify the model generation process to some extent, most lack operability and stability, and their outputs are often unexecutable or unverifiable. In addition, the lack of deployable implementations and standardized interfaces makes it difficult for these tools to be widely used in real-world engineering environments. In comparison, such as Figure 1 As shown, this invention provides a traceable workflow for graph generation and reuse by combining LLM agents with deployable tools, overcoming the limitations of existing LLM-assisted prototypes in engineering applications.

[0155] To address the aforementioned issues, this scenario example provides two paths for users to choose from: a full generation mode (Breadth First Search, BFS, corresponding to the first exploration model) and a budget generation mode (Best-First Search, corresponding to the second exploration model). In full generation mode, the system generates a complete state transition diagram, while in budget generation mode, the system generates a partial state transition diagram based on the user-defined state budget. Furthermore, the generated text file will be parsed by an intelligent operations and maintenance agent, automatically identifying weak points and providing actionable alarms, maintenance suggestions, and redundancy switching strategies, thus achieving a closed loop of modeling—output—operations and maintenance decision-making. The agent-assisted operations and maintenance decision support function can identify high-risk paths and weak components based on the generated state transition diagram, providing targeted decision support.

[0156] See Figure 9 As shown, this scenario example proposes a MarkovOps system for automating the generation of Markov state transition diagrams and providing operational decision support. Specifically, the MarkovOps system achieves intelligent Markov modeling and operational decision support through four key steps. Step 1: The system extracts system parameters (e.g., component parameters) and success criteria from structured or unstructured text input (e.g., description text of the kernel system) with the assistance of an agent (e.g., the first agent). Step 2: The system encodes key transition mechanisms into controlled transition system transition rules. Step 3: The system generates Markov state transition diagrams (e.g., state transition diagrams) with transition rate annotations under a full breadth-first search algorithm or a budget-first search algorithm. Step 4: The system reuses the exported graphical files through a MarkovGuard module (e.g., the second agent) to generate traceable vulnerability diagnostic reports and operational recommendations (e.g., operational reports). This workflow improves the efficiency, consistency, and reusability of Markov modeling, supports reliability and availability assessments of complex systems, and provides intelligent operational decision support.

[0157] Regarding the extraction of parameters and success criteria, the key parameters and success criterion logic of the system can be extracted through the following steps during implementation.

[0158] Regarding component parameters, the user specifies the number of components in the system, and the software automatically generates component identifiers (such as A, B, C, etc.). For each component i, the system records its failure rate. and repair rate The system architecture is specified through user-defined path numbers: identical numbers indicate serial paths, while different numbers indicate parallel paths. Redundancy configuration specifies the primary device, backup device, and their startup failure probabilities. Common-cause faults are defined by a set of components and their common-cause failure rates; redundancy configurations and common-cause fault settings can be added or removed as needed.

[0159] Regarding the success criterion expression, the system working condition is represented by an optional single-line Python Boolean expression (such as `System_working`), which returns "True" or "False", where "True" corresponds to `System_working='Yes'`. Experts can provide this expression directly. Alternatively, the system offers an option assisted by a large language model agent, which transforms the natural language description into an executable logical expression through a success criterion expression proxy. This option supports non-expert users and can handle logical patterns such as "one-to-many backup" and "k-to-n redundancy".

[0160] Regarding GTS transition rule generation, the system mechanism is expressed through a unified controlled transition system. Each transition rule r∈R is encoded in the form of "Event-Guard-Action-Rate": R=<"Event";"Guard";"Action";"Rate>", where Event identifies the event type; Guard specifies the state of the event. When is it triggered; Action: Update the state to the successor state. ; Transfer rate (Rate): Defines the transfer rate for this transfer.

[0161] Specifically, the above transfer rules can be instantiated from four types of mechanisms: Failure rules: Component i is assigned a failure rate... A failure has occurred. Repair rules: Component i is configured with a repair rate... Repair. Standby activation rules: When the primary device fails, the standby device is activated, reducing the probability of startup failure. Optional. Common Cause Failure (CCF) rule: A specified group of components fails simultaneously due to a common failure source, with a failure rate of [missing information]. .

[0162] To further illustrate how each mechanism triggers Markov state changes Figure 2 A diagram illustrating four transition modes is provided (from state). to state These correspond to faults, repairs, backup activation (including startup failures), and common-cause faults, respectively.

[0163] In addition, options are provided to support column-level timing logic, suitable for scenarios such as "one in use, one backup" configurations. Users can choose to restrict component failover to occur only within the same column. When a column fails, the failover no longer occurs within that column, but instead moves to the next column. This feature increases the complexity of state transition diagram generation.

[0164] Regarding the generation of Markov state transition diagrams, in practice, Markov state transition diagrams can be generated from the initial state. ,in, For a set of states, For transfer sets. MarkovOps supports two user-selectable exploration modes: Among them, the full exploration mode (BFS): when the state space is small and can be fully expanded (e.g., the first scenario type), breadth-first search (BFS) is used to enumerate all reachable states. MarkovOps candidate set (e.g., candidate state set) C, visited state set and transfer set Whenever a state is retrieved from the queue At that time, MarkovOps evaluates all transition rules for that state and generates a successor state for each transition rule. And record the transfer rate "rate" → To avoid unnecessary expansion, states marked as System_working = 'No' are treated as non-transitive absorbing states.

[0165] Budget Exploration Mode (Best-First Search, or Local Exploration Mode): When the state space is large (e.g., in a second scenario), MarkovOps uses a budget-first search mode to generate a partial graph under a user-defined state budget S (e.g., a preset budget threshold). MarkovOps maintains a candidate set C and calculates based on priority scores d( (For example, path priority parameters) continuously expand the candidate states with the highest scores until the state set is reached. The budget limit S is reached or there are no remaining candidate states. In each transition state... MarkovOps will generate successor states. The corresponding transition information is recorded. The score of the successor state is updated using the following formula: .

[0166] in, This is the priority factor for each transfer, which is calculated by normalized transfer rate by default: .

[0167] After exploration is complete, MarkovOps creates an absorbing state. And redirect all transitions pointing to the unreserved state to this absorb state. .

[0168] The specific pseudocode for the Markov state transition graph generation algorithm can be found in Table 1, which includes both full BFS and budget-first search modes, allowing users to choose.

[0169] Table 1

[0170] For decision-oriented operation and maintenance support, please refer to the following for specific implementation details. Figure 10 As shown, MarkovGuard modules (e.g., second agents) can be used to convert exported state transition diagrams into decision-oriented operational support evidence. MarkovGuard generates traceable vulnerability diagnostic reports and provides targeted operational recommendations through the exported standardized .txt file.

[0171] Regarding the input, MarkovGuard reads the standardized .txt file exported by the state transition diagram generator of this invention, and parses the state index mapping and transition sets with transition rate annotations. If available, MarkovGuard will also read the path score d( ), the "ActiveTrain" column, the system status flags System_working, and the absorption state. Information such as system report graph size statistics and the number of failure states identified by `System_working} = 'No'`. If a required part is missing, MarkovGuard will list the missing items and only analyze the existing content.

[0172] Regarding the processing procedure, for each transition, MarkovGuard compares the source state with the target state, identifying component mode changes (e.g., normal → fault). Then, it ranks the risks based on the transition graph content. If the path score d( (This is available; MarkovGuard calculates the risk score "risk_proxy": risk_proxy=d() )×rate( → ).

[0173] If path scoring is unavailable, MarkovGuard ranks moves based on move rate and explicitly states this limitation. For budget maps, MarkovGuard summarizes... The transition of the absorption state is used to indicate the cutoff effect under budget S.

[0174] Regarding the output, MarkovGuard supports interactive querying and returns the following: (1) Graph scale overview: including statistics such as the number of states, transitions, and failure states; (2) Weak component ranking: generating a ranking of weak components based on the highest risk transitions; (3) High-risk transition ranking: listing high-risk transitions and providing a ranking of the weak components. The transition of the absorption state is summarized separately. (4) Summary of fault type patterns: Based on state changes, a concise summary of common fault risk pattern types is provided. (5) Actionable operation and maintenance suggestions: Generate maintenance priorities and maintenance / switching suggestions based on risk warnings, with a brief explanation of limitations (e.g., missing fields in the exported results or obvious truncation effects).

[0175] Furthermore, in Markov graphs of nuclear systems (such as reactor coolant systems (RCS) and equipment cooling water systems (CCWS), the "failure mode" is not inferred out of thin air, but is jointly determined by "source state → target state" + transition rule semantics (event types of GTS): MarkovGuard assigns a value to each transition edge ( → First, identify which components have changed their modes (e.g., "normal → fault"), then map these changes to interpretable fault modes and summarize and rank them.

[0176] In specific nuclear systems, this mapping of "fault type determined by state change" (e.g., risk mode type) is typically done in the following four categories: 1) Independent Failure (Single Component "Normal → Failure"): If only one component X changes from "Normal" to "Failure", then this transition is classified as an independent failure of X. For example, the action of A_fails in RCS is to write {'A':'Failure'} into the successor state. Therefore, the change from S1 to S2 can be directly interpreted as "A Failure: Normal → Failure".

[0177] 2) Repair (Single Component "Fault → Normal"): If only one component X changes from "Fault" to "Normal", it is classified as X repair. For example, in RCS, A_repairs: the guard is state['A']=='Fault', the action is {'A':'Normal'}, so the corresponding edge is "A repair".

[0178] 3) Standby Activation / Startup Failure (Primary device failure triggers standby status change, possibly accompanied by changes in multiple components): If both primary and standby devices are involved in a failure, it is classified as primary / standby switchover / standby activation; if the primary device fails and the standby device also fails, it is further classified as startup failure. For example, in RCS, the action of the I_fails + J_startup_fails event sets both I and J to failure. MarkovGuard sees I: normal → failure and J: normal → failure, and interprets it as "Primary device I failure triggers J startup failure / standby failure to be activated". The rules for CCWS are similar: A_fails_B_starts_successfully and A_fails_B_startup_fails correspond to "primary A fails and B starts successfully" and "primary A fails and B starts unsuccessfully", respectively.

[0179] 4) Common Cause Failure: If multiple components simultaneously fail from normal operation, and this group of components is declared as a CCF group (or the corresponding rule event name is common_cause_fail) in the input file, it is classified as "CCF Multi-Component Common Cause Failure". For example, the RCS common_cause fail I_J action sets I and J to failure simultaneously, and the CCWS common_cause_fail A_B and common_cause_fail_E_F follow the same pattern.

[0180] Among them, RCS (Engineering Example 1, more specific examples will be given later): S1→S2 (A normal → fault) and S1→S3 (C normal → fault) appearing in the main path are all "high λ single component failure"; while the transfer of I&J failures at the same time, such as S1→S10, will be classified as "multi-component failure (CCF or startup failure)".

[0181] CCWS (Engineering Example 2, more specific examples will be given later): S1→S2 (A normal → fault but the system is still operable) will be marked as "degradation (master device failure, not yet in a failed state)"; S2→S3 (B normal → fault) is "standby failure leading to system failure"; S1→S3 (A and B fail at the same time) will be classified as "AB common cause failure / startup failure type multi-component failure".

[0182] In this scenario example, during implementation, the MarkovOps system can be used to evaluate the graph generation under two settings: full generation mode and budget generation mode. In full generation mode, breadth-first search (BFS) is used to derive the complete reachable graph. , In budget generation mode, a partial graph is generated under state budget S using a best-first search. And redirect the truncated transition to the absorption state through absorption state closure. .

[0183] In addition to graph generation, two agents were used, implemented on the Dify + GPT-5.2 platform. The first agent (e.g., Agent 1) is the success criterion expression agent, which translates natural language success criteria into a single-line Python Boolean expression for system status labeling. The second agent (e.g., Agent 2) is MarkovGuard, which acts as an operations assistant, reading the exported Markov graph file and generating decision-oriented operations outputs. All experiments were conducted on a PC equipped with an AMD Ryzen 7 7840H CPU, 32 GB of RAM, and 64-bit Windows 11 (version 24H2). The implementation code used Python 3.9 and ran with the SymPy and NetworkX libraries. At each run, the system reported wall clock time (Time(s)) and peak memory usage (Peak mem(MB)).

[0184] Case studies of nuclear power plant engineering, component failure rate parameters It can be obtained in two ways: as the default option. Taken from the "Nuclear Power Equipment Reliability Data Report". For equipment whose failure rate is affected by environmental factors, The failure rate is then calculated using the non-electrical equipment failure rate analysis tool software (ILS-RPM1.0).

[0185] Regarding the generation of Markov state transition diagrams for typical systems, this paper considers typical system architectures in detail for kernel systems to verify the applicability of the proposed method to general systems. Each component has two modes: normal and fault, and it is assumed that the components are homogeneous unless otherwise specified. Specifically, the following five structures are included.

[0186] Nuclear systems, such as reactor coolant systems, are not associated with specific equipment, and the system configurations they comprise are not simply limited to these five types; they are far more complex. For example, a reactor coolant system might have four loops in series, plus two safety valves, one active and one standby. The equipment cooling water system uses a tiered switching mechanism; if one tier fails, the next tier is activated. Within each tier, one pump is active and one is standby.

[0187] 1) Series systems: contain n∈{3,10} components.

[0188] 2) Parallel systems: contain n∈{3,8,12} components.

[0189] 3) k-out-of-n systems: where (k,n)∈{(1,2),(2,3),(3,5)}.

[0190] 4) One-active-one-standby redundancy.

[0191] 5) One-active-multiple-standby redundancy, where m=4 standby units are activated sequentially when a fault occurs.

[0192] To separate the effects of the structure from other mechanisms, three configurations are defined here: 1) Configure C0: Enable independent faults and standardize the baseline failure rate. Repair rate Common cause failures (CCF) are not considered. For standby systems, C0 is also configured with a fixed startup failure probability. .

[0193] 2) Configure C1: Enable repair in the selected typical system, with a repair rate of At the same time, the independent fault remains unchanged.

[0194] 3) Configure C2: Enable common cause faults and repair in the selected typical system by adding common cause groups, the failure rate is Simultaneously enable repair, with a repair rate of [missing information]. For backup systems, a fixed probability of startup failure is used. Table 2 summarizes all system definitions and parameter settings used in this section.

[0195] Table 2

[0196] Regarding the Markov state transition diagram generation results for typical systems, Tables 3 and 4 summarize the explicit Markov state transition diagrams generated for kernel systems. Table 3 reports the complete reachability graph derived in BFS mode. This includes the number of states and the number of transitions labeled with transition rates. Table 4 shows the budget graphs derived using different state budgets S under the budget mode (best-first search). This includes runtime and peak memory usage. And display the same output fields.

[0197] Table 3

[0198] Table 4

[0199] In BFS mode (Table 3), MarkovOps derives complete explicit graphs for most representative architectures (such as cascaded, k-of-n, and spare configurations) with low computational cost, typically completing in sub-second runtimes, and negligible memory consumption. Parallel architectures, however, experience a significant increase in computational cost as n increases: for example, P-8 requires 22.690 seconds (…). The cost of P-12 increased to 403.490 seconds. Mechanism extensions (such as repair (K23-R vs. K23) and common cause failure (P-8-CCF vs. P-8)) increased the number of transfers.

[0200] Table 4 shows that in the Best-First Search budget mode, generation costs are controlled by the state budget S. For P-8, when S is reduced from 200 to 100, runtime (13.776 seconds to 3.672 seconds) and peak memory (2.36 MB to 0.75 MB) are significantly reduced, and the budget graph is also derived. For P-12, the runtime increases from 13.654 seconds at S=20 to 376.896 seconds at S=1000, which is an adjustable trade-off compared to the full generation mode (403.490 seconds in Table 3).

[0201] Overall, based on the results in Tables 3 and 4, the MarkovOp system can: efficiently generate complete explicit graphs for a wide range of typical systems; and provide budget-controlled generation in larger parallel instances, enabling the export of explicit graphs under limited computational resources.

[0202] Furthermore, this scenario example also specifically applies and tests two actual engineering examples of core systems.

[0203] Engineering Case Study 1: Generation of Markov State Transition Diagrams for Reactor Coolant Systems

[0204] The first engineering case study considers the reactor coolant system of a high-flux reactor under development. This high-flux reactor has a design power of 80 MWth and uses light water as both coolant and neutron moderator. The considered reactor coolant system employs a four-loop symmetrical layout (see [reference]). Figure 11As shown): Each loop contains one reactor cooling pump and one heat exchanger, and the four loops together provide normal core cooling. In this embodiment, the four loops share two safety valves, using a one-in-use, one-out-of-service configuration.

[0205] To generate a Markov state transition diagram for the reactor coolant system, we model the system as a set of repairable two-state components and specify the success criterion as an executable Python Boolean expression. We then report the inputs used for generation and the outputs derived. The generated transition rules are summarized by rule category in Table 5.

[0206] Table 5

[0207] Table 5 shows that the reactor coolant system rule set covers the main operational mechanisms required to generate the state transition diagram, including component failure, component repair, standby behavior with startup failure, and common-cause failure. A total of 22 transition rules are used in this case study, providing an executable event-guard-action specification for driving subsequent graph exploration and derivation.

[0208] Based on these rules, the derived Markov state transition diagram outputs are reported in Table 6. Since the number of states in the reactor coolant system state transition diagram is finitely available, the complete BFS generation output is reported.

[0209] Table 6

[0210] Table 6 reports the Markov state transition diagrams generated under full BFS exploration. For the reactor coolant system case, there are a total of each state and A transfer with a transfer rate label. The measured runtime and peak memory usage were relatively small, indicating that in this case, a complete exploration was sufficient to export explicit graphical files.

[0211] Engineering Case Study 2: CCWS Markov State Transition Diagram Generation

[0212] The second engineering case study considers a high-throughput equipment cooling water system. This system is a dedicated system that operates only under emergency conditions to support waste heat removal. The system has two loops, with one in operation and one on standby (see [reference]). Figure 12 (As shown). Each loop includes two pumps (one operating pump / one standby pump), one heat exchanger, and one expansion tank. The pump recovery rate is... Common cause failure rate (Using the beta factor) .

[0213] Success criteria for the equipment's cooling water system were obtained through a success criterion expression agent. Since loop switching logic is difficult to express directly in the tool, LLM assistance was used to generate executable Boolean expressions. It is worth noting that the equipment's cooling water system employs a one-in-one-out loop configuration and incorporates timing logic. By enabling optional features, the system automatically switches to the backup loop when one loop fails.

[0214] The inputs and outputs generated under both the full generation mode (BFS) and the budget generation mode (Best-First Search) are then reported. The generated transfer rules are summarized by rule category in Table 7.

[0215] Table 7

[0216] Table 7 shows that the rule set for the equipment cooling water system covers component failure, component repair, standby startup (with startup failure), and common cause failure. A total of 20 event-guard-action transfer rules were generated, providing executable specifications for subsequent graph exploration and export.

[0217] Based on these transition rules, Markov state transition diagrams were generated under two exploration modes, as shown in Table 8. Table 8 reports the Markov state transition diagrams generated under the full BFS exploration mode. For the equipment cooling water system case, the size of the state transition diagram is kept moderate, with a total of [number missing]. each state and The transfer with transfer rate annotation has low runtime and memory overhead. In subsequent MarkovGuard operational decision analysis, when path-oriented scoring or resource constraint exploration is necessary, a budget-first search mode is applied here.

[0218] In practice, MarkovGuard can perform operational decision analysis based on the exported graph file generated by MarkovOps under a budget-first search setting. The exported file is in plain text .txt format and contains the information required for downstream queries, specifically including: components and parameters; standby relationships; common failure (CCF) groups; and executable system success logic expressions.

[0219] The document also reports on the generation of summaries (patterns, etc.). , (Runtime and peak memory), followed by a list of states (S# mapped to a state dictionary), state path weights d(σ), and transition details ( (Transfer rate). Using and MarkovGuard first loads the exported state-transition structure. Then, the system utilizes state annotations (such as System_working, ActiveTrain (where applicable), and This is used to answer maintenance queries related to budget control and path guidance.

[0220] In practice, MarkovGuard identifies system weaknesses from two complementary perspectives: At the path level, MarkovGuard enumerates candidate paths from the operational state to the system failure state (System_working = 'NO'), ranks these paths, and generates a global list of dominant failure paths. At the component level, it integrates all contributions from the transition from "normal" to "failure," ranks components by path-weighted risk impact, and identifies the critical components most likely to cause system failure.

[0221] Based on ranking paths and components, MarkovGuard generates executable online alerts, maps high-risk states and transition trigger conditions and recommended actions, and summarizes them into priority-ranked operational recommendations. When sequential logic is involved, MarkovGuard also provides loop-level switching guidance (when sequential logic applies). MarkovGuard reports its decision support results using a fixed seven-part template (see Figure 13). The seven parts include: graph-scale overview; component-level weaknesses; high-risk edges at the transition level; summary of failure modes and main failure paths; online alerts; operational priorities; uncertainty descriptions and modeling follow-up work.

[0222] Finally, MarkovGuard supports interactive follow-up queries. Users can delve deeper into specific paths or components, trace the support status and transitions behind each suggestion, and view the analysis results under different operational scenarios. See Tables 8 and 9 for details.

[0223] Table 8

[0224] Table 9

[0225] Tables 8 and 9 summarize the operation and maintenance decision support results of MarkovGuard for case studies of reactor coolant systems and equipment cooling water systems, reporting the top-ranked dominant failure paths and top-ranked weak components, supporting representative high-risk transfers.

[0226] In summary, Tables 8 and 9 show that MarkovGuard can transform the derived truncated Markov diagrams into actionable vulnerability evidence, providing detailed analysis results at both the path and component levels. (See the reactor coolant system case study for more details.) Figure 14 In the example shown, the results are primarily dominated by the cascaded success chain, where a single component failure directly leads to functional loss from a fully healthy state, with high-failure-rate components (such as A / C / E / G) contributing the most. The I→J standby relationship mainly serves as a mechanism-level risk channel (especially the common failure path of I and J) and as a degraded but operable state, promoting accelerated recovery. In the case of the equipment cooling water system (see...),... Figure 15 As shown, vulnerabilities are concentrated in the one-in-one-backup links (A / B and E / F), loop-level switching, and one-step fatal paths. Tables 8 and 9, along with the dominant paths, consistently reveal the same pattern: a failure of the working unit causes the system to degrade, a subsequent failure of the backup unit leads to a loss of availability, and startup failure or a common failure can also trigger a transition to a critical state (e.g., S1→S2, S2→S3, S1→S3). Based on these ranked paths and evidence edges, MarkovGuard maps high-risk transitions to actionable priorities and generates executable, evidence-based operational guidance.

[0227] In addition, a sensitivity analysis was conducted on the equipment cooling water system case, taking into account the repair rate of changing parameters. (Values ​​are 0.1 / 0.25 / ), while keeping all fault and operating mechanism parameters unchanged. The exported truncation plots are identical under all settings ( and Therefore, this comparative study can isolate and repair how path weights are dynamically reallocated. This, in turn, affects the ranking score used by MarkovGuard.

[0228] In the tests Within the scope, the dominant failure mechanism remains unchanged. The highest-ranking path is consistently S1→S3→S20, which enters the high-risk degradation zone and eventually transitions to a failed state (System_working = 'NO'). See also Figure 16 As shown, in terms of effect size, when Increased from 0.1 to When the total weight of the failure state is reduced by approximately an order of magnitude (when...), the total weight of the failure state is reduced by approximately an order of magnitude (when...). When the value is 0.25, the weight is at an intermediate level.

[0229] Meanwhile, the top-ranked weaknesses remain highly stable: the top five weak component sets remain unchanged (top five overlap = 100%), and the overall ranking maintains strong consistency across all settings. ; These results demonstrate that MarkovGuard possesses structural robustness to the uncertainty of the repair rate. It primarily affects the size of the tail accumulation, without changing the dominant mechanism or the order of operation triggers.

[0230] In practice, MarkovOps is implemented as an industrial-grade methodology framework and executable software, packaged as a standalone Windows executable file (.exe) for immediate use upon system startup. As shown in Figure 17, the tool's interface follows a three-step guided workflow: system parameters, transition rules, and state transition diagram. The tool exports the explicit Markov state transition diagram as a plain text file (including a state / transition list and comments), which users can directly upload to MarkovGuard for operational decision support. Furthermore, it provides in-depth support for states and transitions through evidence-based tracing capabilities.

[0231] This tool supports two complementary methods for specifying success logic: 1. Expert Structured Parameters: Users can directly input system parameters and mechanism settings through the interface. The agent generates an executable Python Boolean success criterion expression for users to review and choose to edit before recompiling.

[0232] 2. Unstructured Description: For users unfamiliar with the system architecture, the agent provides a plain text / JSON-like template and instructions. Users can provide a brief text description, which the agent parses into the same executable success logic file as described above.

[0233] In addition, the agent can be deployed locally, supporting complete offline use in the local environment, ensuring normal operation even without a network connection.

[0234] The above scenario examples validate the operation and maintenance data processing method for kernel-based systems provided in this specification. First, an LLM agent is used to automatically extract system parameters and success criteria from structured input or unstructured engineering text, generating executable Boolean expressions for state labeling and transition rule definition. Second, this invention automatically generates transition rules including mechanisms such as faults, repairs, standby switching, startup failures, and common-cause failures through a unified controlled transition system format, supporting modeling of multi-component redundancy and complex fault propagation. Third, the system provides two generation modes: full breadth-first search (BFS) and best-first search, enabling the generation of partial state spaces even with limited computing resources. Furthermore, the MarkovGuard module utilizes the generated Markov state transition diagram to automatically identify high-risk paths and weak components, generating evidence-based operation and maintenance recommendations such as warnings, maintenance priorities, and redundancy switching strategies, thereby providing decision-oriented operation and maintenance support. Specifically, the system employs several methods: Executable Boolean expressions are automatically generated using natural language descriptions to label system states; the GTS transition rule generation method compiles system mechanisms into a standardized event-guardian-action-transition rate format to automatically generate state transition diagrams, supporting modeling of redundancy mechanisms and fault propagation; a budget-first search mode generates partial state transition diagrams under user-defined state budgets; and MarkovGuard operational decision support automatically identifies high-risk transitions by parsing the generated state transition diagrams, providing targeted maintenance suggestions and fault predictions. Finally, standardized text output and intelligent parsing ensure the generated state transition diagrams are traceable and operable, supporting subsequent verification, decision support, and model reuse. Based on these methods, not only is modeling efficiency and accuracy improved, but the operational management capabilities of complex systems are also significantly enhanced, ensuring system security and reliability.

[0235] While this specification provides the steps of operation for the methods described in the embodiments or flowcharts, more or fewer steps may be included based on conventional or non-inventive means. The order of steps listed in the embodiments is merely one possible order of execution among many steps and does not represent the only possible order. In actual device or client product execution, the methods shown in the embodiments or drawings may be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment, or even a distributed data processing environment). The terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Without further limitations, the presence of other identical or equivalent elements in a process, method, product, or apparatus that includes said elements is not excluded. The terms "first," "second," etc., are used to denote names and do not indicate any particular order.

[0236] Those skilled in the art will also know that, besides implementing the controller using purely computer-readable program code, the same functions can be achieved by logically programming the method steps, making the controller function as logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices within it used to implement various functions can also be considered structures within that hardware component. Alternatively, the devices used to implement various functions can be considered as both software modules implementing the method and structures within a hardware component.

[0237] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer-readable storage media, including storage devices.

[0238] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this specification can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions of this specification can essentially be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments of this specification.

[0239] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. This specification can be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0240] Although this specification has been described by way of examples, those skilled in the art will recognize that many variations and modifications are possible without departing from the spirit of this specification, and it is intended that the appended text include such variations and modifications without departing from the spirit of this specification.

Claims

1. A method for processing operation and maintenance data based on a kernel system, characterized in that, include: Obtain the description text of the target core system; The first intelligent agent extracts the component parameters of the target core system based on the description text; And determine the success criteria for the target nuclear system; Based on the component parameters of the target core system, a transfer rule for the target core system is generated; Determine and, based on the current working conditions, identify a matching target exploration model from the preset exploration models; wherein the preset exploration models include at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; Using the target exploration model, the target core system's component parameters, success criteria, and transition rules are used to perform corresponding exploration processing, resulting in the state transition diagram of the target core system. The second intelligent agent generates an operation and maintenance report for the target core system based on the state transition diagram.

2. The method according to claim 1, characterized in that, The component parameters include at least one of the following: number of components, component identifier, component type, connection path, backup startup failure rate, failure rate, repair success rate, common cause failure rate, and system structure; The system structure includes at least one of the following: series structure, parallel structure, k-n structure, one-use-one-standby structure, and one-use-multiple-standby structure.

3. The method according to claim 2, characterized in that, The transfer rules include at least one of the following: fault rules, repair rules, backup activation rules, and common cause fault rules.

4. The method according to claim 1, characterized in that, The step of determining and identifying a matching target exploration model from the preset exploration models based on the current working conditions includes: Determine the number of components in the target core system and / or the operational and maintenance requirements parameters for the target core system; Based on the number of components and / or operation and maintenance requirements parameters, determine the current operating scenario of the target core system; Based on the current working conditions, a matching target exploration model is determined from the preset exploration models.

5. The method according to claim 4, characterized in that, When the target exploration model includes a first exploration model, the step of using the target exploration model to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain the state transition diagram of the target core system includes: The current round of exploration will be processed as follows: Obtain the previous state of the target core system; Using the target exploration model, the state label of the previous round's state is determined based on the success criteria and component parameters of the target kernel system; When the status label of the previous round indicates that the status is normal, determine the status of the current round based on the status of the previous round and the transition rules; and calculate the transition rate of the current round from the status of the previous round to the status of the current round. Update the state set based on the state of the previous round and the corresponding state label; and update the transition set based on the transition rate of the current round.

6. The method according to claim 4, characterized in that, When the target exploration model includes a second exploration model, the process of using the target exploration model to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system to obtain the state transition diagram of the target core system includes: The current round of exploration will be processed as follows: Obtain the previous state of the target core system; Check whether the link length of the state link between the previous state and the initial state of the target core system is less than the preset budget threshold; When the link length of the state link between the previous state and the initial state of the target core system is less than the preset budget threshold, the state label of the previous state is determined by the target exploration model based on the success criteria and component parameters of the target core system. When the state label of the previous state indicates that the state is normal, determine the successor state of the previous state based on the previous state and the transition rules; and calculate the path priority parameter from the previous state to the successor state. Based on the path priority parameter, determine the state of the current round from the successor states; and calculate the transition rate from the state of the previous round to the state of the current round. Update the state set based on the state of the previous round and the corresponding state label; and update the transition set based on the transition rate of the current round and the path priority parameter of the current round's state.

7. The method according to claim 6, characterized in that, After detecting whether the link length of the state link between the previous state and the initial state of the target core system is less than a preset budget threshold, the method further includes: When the link length of the state link between the previous state and the initial state of the target core system is greater than or equal to the preset budget threshold, state evolution features are extracted based on the state link between the previous state and the initial state of the target core system. Based on the state evolution characteristics, detect whether the adaptive extension condition with respect to the preset budget threshold is met; When it is determined that the adaptive extension condition is met, a matching extension adjustment value is determined based on the state evolution characteristics; Based on the aforementioned extension adjustment value, the preset budget threshold for the state link between the previous state and the initial state of the target core system is dynamically updated.

8. The method according to claim 7, characterized in that, After detecting whether an adaptive extension condition with respect to a preset budget threshold is met, the method further includes: When it is determined that the adaptive extension condition is not met, the exploration process based on the state of the previous round ends; Update the state set based on the state of the previous round and the corresponding state label; Check if the exploration termination conditions are met; When the exploration termination condition is met, the state that is not included in the state set is determined as the absorption state; The absorption states are saved to an absorption state set; and the transition set is updated based on the absorption state set.

9. The method according to claim 1, characterized in that, The step of generating an operation and maintenance report for the target core system using a second intelligent agent based on the state transition diagram includes: The state transition graph is processed by a second intelligent agent to extract key parameters; wherein, the key parameters include: path priority parameters, activity column parameters, state labels, and absorption state information; Based on the aforementioned key parameters, risk transfer paths and risk components are determined; Based on the risk transfer path and risk components, the risk pattern type is determined; Based on the risk mode type, generate an operation and maintenance report for the target core system.

10. A maintenance data processing device based on a nuclear system, characterized in that, include: The acquisition module is used to acquire the description text of the target core system; The processing module is used to extract the component parameters of the target core system based on the description text using the first intelligent agent; And determine the success criteria for the target nuclear system; The first generation module is used to generate transfer rules for the target core system based on the component parameters of the target core system. The determination module is used to determine and, based on the current working conditions, identify a matching target exploration model from the preset exploration models; wherein the preset exploration models include at least: a first exploration model based on a complete exploration mode and a second exploration model based on a partial exploration mode; The exploration module is used to perform corresponding exploration processing based on the component parameters, success criteria, and transition rules of the target core system using the target exploration model, and to obtain the state transition diagram of the target core system. The second generation module is used to generate an operation and maintenance report of the target core system based on the state transition diagram using the second intelligent agent.

Images