Complex industrial system fault risk assessment method and device

Abstract

The invention provides a complex industrial system fault risk assessment method and device, and the method comprises the steps: obtaining equipment fault rates corresponding to all equipment in a complex industrial system, and a current real-time system operation mode of the complex industrial system; the equipment fault rate and the real-time system operation mode are input into a risk assessment model, a current fault risk assessment result, output by the risk assessment model, of the complex industrial system is obtained, and the risk assessment model is obtained through construction based on a fault tree analysis method and a Bayesian network. According to the method, the fault rate and the current real-time operation mode of each device of the complex industrial system are obtained and then input into the risk assessment model constructed based on the fault tree analysis method and the Bayesian network, and the model can deeply mine the fault causal relationship, effectively resist interference and quickly and accurately output the current fault risk assessment result; and the evaluation accuracy and efficiency are obviously improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent monitoring, and in particular to a complex industrial system fault risk assessment method and device. BACKGROUND

[0002] In the industrial field, complex industrial systems have become the core support of production, covering power, chemical industry, transportation and many other key industries. These systems integrate a large number of advanced equipment and sophisticated technologies, with complex structures, and interrelated and coordinated operation between components.

[0003] At present, there are various technical means for complex industrial system fault risk assessment. Traditional methods are mostly based on expert experience and historical data statistics, which can provide certain reference, but rely on a large number of manual participation, are highly subjective, and are difficult to adapt to system dynamic changes. Some intelligent algorithms such as neural networks and support vector machines have also been applied to fault diagnosis, but these methods face problems such as high data dimension and difficulty in feature extraction when dealing with complex systems, resulting in low fault diagnosis efficiency and difficulty in meeting the high requirements of actual industrial production for system safe and stable operation.

[0004] Therefore, there is an urgent need for a complex industrial system fault risk assessment method and device to solve the above problems. SUMMARY

[0005] In view of the problems existing in the prior art, the present application provides a complex industrial system fault risk assessment method and device.

[0006] The present application provides a complex industrial system fault risk assessment method, comprising: obtaining the device failure rate corresponding to each device in the complex industrial system, and the real-time system operation mode of the complex industrial system at present; inputting the device failure rate and the real-time system operation mode into a risk assessment model to obtain the fault risk assessment result of the complex industrial system at present output by the risk assessment model, wherein the risk assessment model is constructed based on fault tree analysis method and Bayesian network.

[0007] According to the complex industrial system fault risk assessment method provided by the present application, the risk assessment model is constructed by the following steps: based on the fault tree analysis method and the historical fault event data of the complex industrial system, constructing the system operation fault tree of the complex industrial system corresponding to different historical system operation modes; determining the input-output relationship between each fault event in the system operation fault tree, and constructing the subsystem Bayesian network corresponding to each subsystem of the complex industrial system under different historical system operation modes according to the fault events and the input-output relationship. constructing a complex industrial system Bayesian network based on each of the subsystem Bayesian networks; constructing the risk assessment model according to the complex industrial system Bayesian network and each of the subsystem Bayesian networks.

[0008] According to the complex industrial system fault risk assessment method provided by the application, the system operation fault tree of the complex industrial system corresponding to different historical system operation modes is constructed based on the fault tree analysis method and historical fault event data of the complex industrial system, and the system operation fault tree comprises: determining basic events, intermediate events and top events corresponding to the historical fault event data in different historical system operation modes based on the fault tree analysis method, wherein the basic events are equipment fault events in the historical fault event data, the intermediate events are subsystem fault events in the historical fault event data, and the top events are complex industrial system fault events in the historical fault event data; constructing the system operation fault tree according to the basic events, the intermediate events and the top events.

[0009] According to the complex industrial system fault risk assessment method provided by the application, each subsystem Bayesian network corresponding to each subsystem of the complex industrial system in different historical system operation modes is constructed according to the fault events and the input-output relationship, and the complex industrial system fault risk assessment method comprises the following steps: constructing each node in the Bayesian network based on the basic events, the intermediate events and the top events, wherein the basic events correspond to root nodes in the Bayesian network, the intermediate events correspond to intermediate nodes in the Bayesian network, and the top events correspond to leaf nodes in the Bayesian network; constructing a directed arc in the Bayesian network based on the input-output relationship; constructing a conditional probability table in the Bayesian network based on the logical gate in the system operation fault tree; constructing the subsystem Bayesian network according to the root nodes, the intermediate events, the directed arc and the conditional probability table corresponding to each subsystem of the complex industrial system in different historical system operation modes.

[0010] According to the complex industrial system fault risk assessment method provided by the application, a conditional probability table in the Bayesian network is constructed based on the logical gate in the system operation fault tree, and the complex industrial system fault risk assessment method comprises the following steps: converting the logical gate in the system operation fault tree into the conditional probability table according to the system composition type of the complex industrial system, wherein the system composition type comprises a series system type, a parallel system type and a k / n voting system type.

[0011] According to the present invention, a method for assessing the failure risk of a complex industrial system, the method further includes: When it is determined that the system operation mode of the complex industrial system has changed, the target equipment failure rate is obtained, wherein the target equipment failure rate is the equipment failure rate corresponding to each device under the changed system operation mode. The modified system operation mode and the target equipment failure rate are input into the risk assessment model to obtain the intermediate risk assessment results corresponding to each subsystem in the complex industrial system, output by the Bayesian network of each subsystem, and the failure risk assessment results output by the Bayesian network of the complex industrial system based on the intermediate risk assessment results.

[0012] The present invention also provides a fault risk assessment device for complex industrial systems, comprising: The data acquisition module is used to obtain the equipment failure rate of each device in the complex industrial system, as well as the current real-time system operation mode of the complex industrial system. The risk assessment module is used to input the equipment failure rate and the real-time system operation mode into the risk assessment model to obtain the current failure risk assessment result of the complex industrial system output by the risk assessment model. The risk assessment model is constructed based on fault tree analysis and Bayesian network.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the fault risk assessment method for complex industrial systems as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the fault risk assessment method for complex industrial systems as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the fault risk assessment method for complex industrial systems as described above.

[0016] The present invention provides a method and apparatus for assessing the failure risk of complex industrial systems. By acquiring the failure rate and current real-time operating mode of each device in the complex industrial system, and then inputting the two into a risk assessment model constructed based on fault tree analysis and Bayesian network, the model can deeply explore the causal relationship of failures, effectively resist interference, and quickly and accurately output the current failure risk assessment results, which significantly improves the accuracy and efficiency of the assessment. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the fault risk assessment method for complex industrial systems provided by this invention; Figure 2 A schematic diagram of a Bayesian network for a complex industrial system provided by the present invention; Figure 3 A schematic diagram of the architecture of the risk assessment model provided by this invention; Figure 4 A schematic diagram of the overall business process of the risk assessment model provided by this invention; Figure 5 A schematic diagram of the structure of the fault risk assessment device for complex industrial systems provided by the present invention; Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0020] In the industrial sector, complex industrial systems serve as the core force of production, widely covering key industries such as power, chemical, and transportation. They integrate numerous advanced equipment and precision technologies, have complex structures, and their components are closely interconnected and work collaboratively.

[0021] Currently, there are various technologies for assessing the failure risks of complex industrial systems. Traditional methods mainly rely on expert experience and historical data statistics, which, while having some reference value, involve a lot of human intervention, are highly subjective, and are difficult to adapt to dynamic changes in the system. When some intelligent algorithms, such as neural networks and support vector machines, are used for fault diagnosis, the high dimensionality of data in complex systems and the difficulty of feature extraction lead to low diagnostic efficiency, which cannot meet the high requirements of industrial production for the safe and stable operation of systems.

[0022] Failure risk assessment, as a crucial means of quantifying potential defects in complex industrial systems, is of great significance for ensuring personnel safety, maintaining production continuity, and preventing property damage. The process requires a comprehensive analysis of system failure modes, causes, consequences, and probabilities of occurrence to optimize system design, improve fault diagnosis efficiency, and ensure stable operation. Quantitative risk assessment models, based on historical data and system failure processes, use inference algorithms to identify key factors influencing system failures, providing a probability distribution of system states. This numerically quantifies and intuitively presents system reliability, helping engineers and decision-makers identify and manage system weaknesses, guide system design improvements, and reduce future failure risks. Predictive risk assessment can also effectively plan maintenance strategies, optimize resource allocation, and reduce unexpected downtime. From an economic perspective, it can reduce failure costs and improve the overall economic efficiency of the system.

[0023] This invention integrates fault tree analysis and Bayesian network analysis, combining the comprehensiveness of fault tree analysis with the flexibility of Bayesian networks to construct a fault risk assessment model for complex industrial systems, thereby enabling probabilistic risk assessment of operational failures in complex industrial systems.

[0024] Figure 1 A flowchart illustrating the fault risk assessment method for complex industrial systems provided by this invention is shown below. Figure 1 As shown, the present invention provides a method for assessing the failure risk of a complex industrial system, comprising: Step 101: Obtain the equipment failure rate of each device in the complex industrial system, as well as the current real-time system operation mode of the complex industrial system.

[0025] Complex industrial systems consist of numerous devices, each with varying probabilities of failure due to factors such as design, manufacturing quality, service life, and operating environment. Equipment failure rate is a crucial indicator for measuring the likelihood of equipment failure within a specific timeframe. In this invention, by collecting historical operating data, maintenance records, and failure statistics, and employing appropriate statistical methods and reliability analysis tools, the failure rate of each device under given conditions can be calculated. For example, for a motor operating under high temperature and high load conditions for an extended period, analyzing the number of failures and total operating time over a past period allows for the estimation of the motor's failure rate under the current operating environment. This equipment failure rate data serves as important input parameters for subsequently constructing risk assessment models and analyzing system failure risks.

[0026] The operating modes of complex industrial systems are not static; they dynamically adjust based on factors such as production demands, process requirements, and equipment status. Different operating modes significantly impact the system's failure risk. For example, in chemical production, a system may exist in normal production mode, equipment maintenance mode, and emergency shutdown mode. In normal production mode, equipment operates at rated parameters; while in equipment maintenance mode, some equipment may be shut down, and the overall operating logic and risk characteristics of the system differ significantly from those in normal production mode. Therefore, accurately obtaining the system's current real-time operating mode, including the operating status of each piece of equipment (such as start-up, shutdown, and operating parameters), the system's workflow, and process parameters, is crucial for a comprehensive and accurate assessment of system failure risk.

[0027] Step 102: Input the equipment failure rate and the real-time system operation mode into the risk assessment model to obtain the current failure risk assessment result of the complex industrial system output by the risk assessment model, wherein the risk assessment model is constructed based on fault tree analysis and Bayesian network.

[0028] In this invention, the risk assessment model is constructed based on fault tree analysis and Bayesian networks. First, a static logical model of operational failures in a complex industrial system is built using fault tree analysis. Fault tree analysis is a top-down deductive reasoning method that uses the undesirable failure event (the top event) as the analysis target. By analyzing the direct and indirect causes of this failure event layer by layer, the root cause of the system failure is ultimately found. By constructing a fault tree, the logical relationships and propagation paths of system failures can be clearly displayed, providing a basic framework for subsequent risk assessment. Subsequently, considering the characteristics of the system and its operating mode, a direct method is applied to model the complex industrial system and its subsystems using Bayesian networks.

[0029] Bayesian networks are uncertainty reasoning models based on probability theory and graph theory, capable of handling uncertainty and dependencies between variables. In constructing a Bayesian network, each event in the fault tree is used as a node, and the logical relationships between events are represented by directed edges. By collecting relevant historical data and expert knowledge, a conditional probability table for each node is determined. This constructs a Bayesian network model that reflects the probability and uncertainty of system failures.

[0030] Furthermore, the acquired equipment failure rate and real-time system operation mode are input into the constructed risk assessment model. In this invention, the equipment failure rate data serves as the basic probability information for device nodes in the Bayesian network, while the real-time system operation mode data can be used to adjust the conditional probability table in the model to reflect changes in the system failure probability under different operation modes. Based on the input data, the risk assessment model uses the inference algorithm of the Bayesian network to calculate the probability of system failure in the current state, as well as other relevant risk indicators, such as the severity and scope of the failure. Finally, these calculation results are output as the current failure risk assessment result of the complex industrial system, providing decision-making basis for system operation and maintenance personnel, helping them to take timely measures to reduce system failure risks and ensure the safe and stable operation of the system.

[0031] The fault risk assessment method for complex industrial systems provided by this invention obtains the failure rate and current real-time operating mode of each device in the complex industrial system, and then inputs the two into a risk assessment model constructed based on fault tree analysis and Bayesian network. This model can deeply explore the causal relationship of faults, effectively resist interference, and quickly and accurately output the current fault risk assessment results, significantly improving the accuracy and efficiency of the assessment.

[0032] Based on the above embodiments, the risk assessment model is constructed through the following steps: Based on fault tree analysis and historical fault event data of the complex industrial system, a system operation fault tree corresponding to different historical system operation modes of the complex industrial system is constructed. Determine the input-output relationship between each fault event in the system operation fault tree, and construct a subsystem Bayesian network corresponding to each subsystem of the complex industrial system under different historical system operation modes based on the fault events and the input-output relationship; Based on the Bayesian networks of each of the subsystems, a Bayesian network for a complex industrial system is constructed. The risk assessment model is constructed based on the Bayesian network of the complex industrial system and the Bayesian networks of each of the subsystems.

[0033] Fault tree analysis starts with an undesirable failure event (the top event) in the system and then analyzes the direct and indirect causes of the top event layer by layer downwards. These causes are connected by logic gates (such as AND gates and OR gates) to form a tree structure, i.e., a fault tree. This invention, by constructing a fault tree, can clearly demonstrate the logical relationships and propagation paths of system failures, which helps to deepen the understanding of the mechanisms by which system failures occur.

[0034] Complex industrial systems accumulate a large amount of historical fault event data during long-term operation. This data contains information on various faults that occurred under different operating conditions, such as the time, location, type, and cause of the fault. Utilizing this historical fault event data, it is possible to more accurately identify potential system failures and the causal relationships between them, providing rich examples for constructing system fault trees.

[0035] In this invention, complex industrial systems may operate under different modes at different historical periods. For example, during peak and off-peak production periods, system operating parameters and equipment loads will differ, leading to variations in the characteristics and probabilities of system failures. Therefore, it is necessary to construct separate system operation fault trees for different historical system operation modes. Taking a chemical production system as an example, under normal continuous production mode, the main focus may be on equipment wear and pipeline blockages; while under trial operation mode after equipment maintenance, the focus may be more on equipment startup failures and improper parameter adjustments. This invention, by constructing specialized system operation fault trees for different operation modes, can more accurately analyze the system's failure risks under those modes.

[0036] Furthermore, in the system's fault tree, each fault event is connected by logic gates, which define the input-output relationships between these events. For example, in a fault event connected by an AND gate, the output fault event will only occur if all input fault events occur; while in a fault event connected by an OR gate, the output fault event will occur as long as at least one input fault event occurs. By explicitly defining these input-output relationships, the directed edges and conditional probability relationships between nodes in the Bayesian network are determined.

[0037] In this invention, complex industrial systems typically consist of multiple subsystems, each with its specific function and structure. Based on the fault events in the system's fault tree and their input-output relationships, fault events belonging to the same subsystem can be extracted, and a subsystem Bayesian network corresponding to that subsystem can be constructed. In the subsystem Bayesian network, fault events are treated as nodes, and input-output relationships are represented by directed edges. Simultaneously, by collecting relevant historical data and expert knowledge, a conditional probability table for each node is determined, representing the probability that its child nodes will fail given that a node has failed. For example, for a subsystem consisting of a motor and a transmission device, if the probability of the motor failing is known to be 0.1, and the probability of the transmission device failing given the motor failure is 0.8, then the corresponding conditional probabilities can be set in the subsystem Bayesian network.

[0038] In this invention, the subsystems are not independent of each other; a failure in one subsystem may affect the normal operation of other subsystems. Therefore, after constructing the Bayesian networks of each subsystem, they need to be integrated to build a Bayesian network for the complex industrial system. During integration, the interface relationships and mutual influences between subsystems need to be considered. By adding appropriate directed edges and conditional probability relationships, the Bayesian networks of each subsystem are connected into an organic whole. For example, in a complex industrial system containing a production subsystem, a control subsystem, and a power supply system, if a power supply system failure causes a power outage, the production and control subsystems will also be affected and cease operation. When constructing the Bayesian network for the complex industrial system, this dependency relationship between subsystems needs to be reflected.

[0039] Furthermore, the Bayesian networks of complex industrial systems and their respective subsystems collectively constitute a complete risk assessment framework. Based on these Bayesian networks, Bayesian inference algorithms can be used to assess the failure risks of the system under different operating conditions. Specifically, by inputting real-time operating data and equipment status information, the prior probabilities of each node in the Bayesian network are updated. Then, according to the inference rules of the Bayesian network, risk indicators such as the probability of various system failures and the severity of the failures are calculated. For example, by calculating the probability of a major system failure, the current safety status of the system can be assessed; by analyzing the propagation path and impact range of the failure, corresponding emergency plans and maintenance strategies can be formulated.

[0040] Finally, by encapsulating and solidifying the aforementioned Bayesian network-based risk assessment methods and processes, a complete risk assessment model is constructed. This model can serve as a tool to provide decision support for operators and maintenance personnel of complex industrial systems, helping them to promptly identify potential safety hazards, take effective measures to reduce system failure risks, and ensure the safe and stable operation of the system. Furthermore, the risk assessment model can be continuously optimized and updated based on the actual operating conditions of the system and new failure data, improving the accuracy and reliability of risk assessments.

[0041] Based on the above embodiments, the step of constructing a system operation fault tree corresponding to different historical system operation modes of the complex industrial system based on fault tree analysis and historical fault event data of the complex industrial system includes: Based on the fault tree analysis method, the basic events, intermediate events, and top events corresponding to the historical fault event data under different historical system operation modes are determined, wherein the basic events are equipment fault events in the historical fault event data; the intermediate events are subsystem fault events in the historical fault event data; and the top events are complex industrial system fault events in the historical fault event data. Based on the basic events, intermediate events, and top events, construct the system operation fault tree.

[0042] In this invention, based on fault tree analysis, starting from the final result (top event) of a system failure, the direct causes (intermediate events) and indirect causes (basic events) leading to that result are analyzed step by step from top to bottom. These events are then connected using logic gates (such as AND gates and OR gates) to form a tree structure for analyzing system failures. In the failure analysis of complex industrial systems, this method can clearly identify the logical chain of failures occurring under different operating modes.

[0043] Complex industrial systems operate in different modes at different historical periods. For example, during peak production periods, the system may operate at full load, leading to accelerated equipment wear and tear, and the frequency and types of failures may differ from those during low-load operation. During trial operation after equipment maintenance, potential factors such as improper equipment debugging and incorrect parameter settings may cause failures. Therefore, it is necessary to identify the basic events, intermediate events, and culminating events for different historical system operating modes.

[0044] Basic events are the lowest-level, indivisible events in a fault tree. In this invention, equipment failure events from historical fault event data are used as basic events because equipment is the fundamental building block of complex industrial systems, and equipment failure is often the root cause of system failures. For example, in a chemical production system, equipment failure events such as pump failure, motor damage, and valve jamming can all be considered basic events. By statistically analyzing equipment failures in historical fault event data, it is possible to determine which equipment is more prone to failure under different operating modes, as well as the probability of failure.

[0045] Intermediate events are events composed of basic events combined through logic gates. This invention uses subsystem fault events from historical fault event data as intermediate events. A subsystem is composed of multiple devices combined according to certain functions. A fault in a subsystem is usually caused by the combined effect of faults in multiple devices within it. For example, in a power system, a fault in the generator subsystem may be composed of multiple device fault events such as stator winding faults, rotor faults, and cooling system faults, combined through logical relationships such as AND gates and OR gates. By analyzing historical fault event data, the logical relationships between subsystem faults and device faults under different operating modes can be determined, thereby identifying intermediate events.

[0046] The top event is the ultimate goal of fault tree analysis, namely, the undesirable failure event in the system. In this invention, complex industrial system failure events from historical failure event data are used as top events. For example, the shutdown of an entire chemical production system or a complete power outage in a power plant are both complex industrial system failure events. The occurrence of a top event is often the result of the combined effect of multiple intermediate events. By analyzing historical failure event data, the manifestations and conditions of complex industrial system failures under different operating modes can be clearly identified.

[0047] In this invention, reliability block diagrams of each subsystem describe the reliability relationships and logical structures between components within the subsystem, serving as a crucial basis for constructing fault trees. During qualitative analysis, the logical combination of basic events is determined based on the series and parallel relationships of components in the reliability block diagrams. For example, in a series system, the entire system fails if any component fails; the corresponding logic gate in the fault tree is an OR gate. In a parallel system, the system fails only if all components fail; the corresponding logic gate is an AND gate. By combining reliability block diagrams with qualitative analysis, the logical relationships between basic events, intermediate events, and the top event can be accurately determined.

[0048] Furthermore, based on the logical relationships obtained from the qualitative analysis, a system fault tree is drawn. The fault tree uses the top event as the root node and connects intermediate and basic events through logic gates to form a tree structure. When drawing the fault tree, the name of each event and the type of logic gate need to be clearly labeled for subsequent analysis and understanding. For example, for a complex industrial system containing multiple subsystems, the fault of the complex industrial system is taken as the top event, and the faults of each subsystem are taken as intermediate events and connected by OR gates; then, the equipment faults within each subsystem are taken as basic events, and according to the reliability structure of the subsystems, AND and / or gates are used to connect the basic events with the intermediate events, thereby constructing a complete system operation fault tree.

[0049] The constructed system fault tree provides a static logical model foundation for subsequent quantitative analysis based on Bayesian networks, thus enabling the transition from qualitative to quantitative analysis and providing strong support for probabilistic risk assessment of complex industrial systems. In this invention, to ensure the reliability of probabilistic risk assessment of complex industrial systems, the static logical model needs to be modified according to the system's operating mode when constructing the quantitative analysis Bayesian network model, thereby constructing a Bayesian network model corresponding to the operating mode.

[0050] Based on the above embodiments, the step of constructing a subsystem Bayesian network corresponding to each subsystem of the complex industrial system under different historical system operation modes according to the fault events and the input-output relationships includes: Based on the basic events, intermediate events, and top events, the nodes in the Bayesian network are constructed, wherein the basic events correspond to the root nodes in the Bayesian network; the intermediate events correspond to the intermediate nodes in the Bayesian network; and the top events correspond to the leaf nodes in the Bayesian network. Based on the input-output relationship, a directed arc in the Bayesian network is constructed; Based on the logic gates in the system fault tree, construct a conditional probability table in the Bayesian network; Based on the root node, intermediate event, directed arc, and conditional probability table corresponding to each subsystem of the complex industrial system under different historical system operation modes, construct the Bayesian network of the subsystem.

[0051] In this invention, the basic event is the most fundamental and indivisible event in the fault tree, typically representing a device-level fault in the system. In complex industrial systems, events such as motor failure, valve jamming, and sensor malfunction are all basic events. These events are the source of subsequent faults, with no other events serving as their direct cause.

[0052] In a Bayesian network, a root node is a node without a parent node (i.e., no other node points to it). Basic events are mapped to root nodes because they are the starting points for fault propagation. For example, in a chemical production system, a pump failure, as a basic event, would exist as a root node in a Bayesian network. It is not directly affected by failures in other equipment, but its occurrence may trigger a series of subsequent failures.

[0053] Intermediate events are events composed of basic events combined through logic gates. In complex industrial systems, they often correspond to subsystem-level failures. For example, a failure in a transmission subsystem consisting of multiple devices may be caused by a combination of multiple basic events, such as a faulty drive motor and a broken drive shaft, through logical relationships.

[0054] In a Bayesian network, an intermediate node has a parent node (the node that points to it) and child nodes (the nodes that the child node points to). An intermediate event, acting as an intermediate node, inherits the influence of the basic events and passes it on to the top event. For example, the intermediate node representing a transmission subsystem failure might have a parent node corresponding to the root node, such as a transmission motor failure or a broken drive shaft, while its child nodes might be leaf nodes corresponding to the top event, such as a failure of the entire production system.

[0055] The top event is the ultimate goal of fault tree analysis, representing the undesirable failure event in the system. In complex industrial systems, it signifies a major system-wide failure, such as a plant-wide power outage or production line shutdown. In a Bayesian network, leaf nodes are nodes without children. The top event, as a leaf node, is the endpoint of fault propagation, receiving influences from intermediate and fundamental events. For example, the leaf node corresponding to the top event of a plant-wide power outage is caused by the combined effects of multiple subsystem failures (intermediate events) and equipment failures (fundamental events).

[0056] In a fault tree, there are explicit input-output relationships between events, which reflect the direction of fault propagation. For example, a fault in one device (input) may lead to a fault in its subsystem (output). In Bayesian networks, directed arcs are used to represent causal and dependency relationships between nodes, which can correspond to the input-output relationships in a fault tree.

[0057] Specifically, a directed arc points from a parent node to a child node, clearly defining the path of fault propagation. For example, in a Bayesian network, drawing a directed arc from the root node (parent node) of motor fault to the intermediate node (child node) of transmission subsystem fault indicates that a motor fault may trigger a transmission subsystem fault, clearly demonstrating the fault propagation process.

[0058] In this invention, the logic gates (such as AND gates and OR gates) in the fault tree define the logical combination methods between events. For example, in multiple events connected by an AND gate, the output event will only occur if all input events occur; while in multiple events connected by an OR gate, the output event will occur as long as one input event occurs. In Bayesian networks, conditional probability tables are used to describe the probability of a node occurring under different states of its parent nodes, which corresponds precisely to the logic gates in the fault tree. Taking an OR gate logic as an example, suppose there is an intermediate node (subsystem fault) and two parent nodes (device A fault and device B fault). When constructing the conditional probability table, it is necessary to determine the probability of the intermediate node occurring under different combinations of parent node states. For example, when device A fails (state 1) and device B does not fail (state 0), the probability of a subsystem failure is 0.7; when device A does not fail but device B fails, the probability of a subsystem failure is 0.6; when both device A and device B fail, the probability of a subsystem failure is 0.9; and when neither device A nor device B fails, the probability of a subsystem failure is 0.1. In this way, the logical relationships in the fault tree are transformed into conditional probabilities in a Bayesian network.

[0059] In this invention, the fault characteristics and probabilities of complex industrial systems differ under different historical system operation modes. For example, during peak production periods, the system operates at full load, equipment wear intensifies, and the probability of fault occurrence may increase; while in the trial operation mode after equipment maintenance, there may be problems such as improper equipment debugging, and the fault type may differ from the normal operation mode. Therefore, it is necessary to construct subsystem Bayesian networks based on different historical system operation modes.

[0060] Furthermore, based on the various subsystems of a complex industrial system under different historical system operation modes, the constructed root nodes (equipment failure nodes corresponding to basic events), intermediate nodes (subsystem failure nodes corresponding to intermediate events), directed arcs (representing fault propagation paths), and conditional probability tables (describing the probabilistic relationships between nodes) are integrated. For example, for an electronic power supply system composed of multiple devices, under peak operation mode, the probability of occurrence of each device failure (root node) and the logical relationship between device failures and subsystem failures (represented by directed arcs and conditional probability tables) are determined based on historical data and expert experience. This allows for the construction of a Bayesian network model of the subsystem under peak operation mode, enabling accurate fault analysis and risk assessment for subsystems under different operation modes.

[0061] Based on the above embodiments, the step of constructing a conditional probability table in a Bayesian network based on the logic gates in the system fault tree includes: Based on the system composition type of the complex industrial system, the logic gates in the system operation fault tree are converted into the conditional probability table, wherein the system composition type includes series system type, parallel system type and k / n voting system type.

[0062] In this invention, fault trees use logic gates to represent the logical relationships between events, thus describing the mechanism of system failure. Bayesian networks, on the other hand, use nodes to represent events, connected only by directed arcs, requiring conditional probability tables to represent the logical relationships between nodes. Therefore, to apply fault tree analysis to Bayesian networks, the logic gates in the fault tree need to be transformed into conditional probability tables for the Bayesian network. Furthermore, to simplify research and address data acquisition challenges, this invention assumes that complex industrial systems are binary systems, meaning the system has only two states: fault (represented by "1") and normal (represented by "0"). This facilitates subsequent logic gate transformation and conditional probability table determination.

[0063] Furthermore, corresponding conversion methods are applied to different system composition types, as follows: 1. Converting logic gates in a series system to conditional probability tables: In a series system, all subsystems or components are connected sequentially. If any one of these subsystems or components fails, the entire system will fail. In a fault tree, this relationship is usually represented by an "OR gate". For example, in a system consisting of three components A, B, and C connected in series, the system will fail if A, B, or C fails.

[0064] In Bayesian networks, taking a series system of two components as an example, let component A and component B be two nodes in the series system. When both component A and component B are in state 0 (normal), the probability that the system state is 0 (normal) is 1. Similarly, when component A is in state 0 (normal) and component B is in state 1 (fault), or component A is in state 1 (fault) and component B is in state 0 (normal), or component A is in state 1 (fault) and component B is in state 1 (fault), the probability that the system state is 1 (fault) is 1. Through this logical relationship, a conditional probability table for the nodes of the Bayesian network of the series system can be constructed.

[0065] 2. Converting logic gates into conditional probability tables in parallel systems: In a parallel system, multiple subsystems or components are connected in parallel. The entire system will only fail when all subsystems or components fail. In a fault tree, this relationship is usually represented by an AND gate. For example, in a system consisting of two components C and D connected in parallel, the system will only fail if both C and D fail.

[0066] Taking a parallel system with two components as an example, let components C and D be two nodes in the parallel system. When the state of component C is 0 (normal) and the state of component D is 0 (normal), or the state of component C is 0 (normal) and the state of component D is 1 (fault), or the state of component C is 1 (fault) and the state of component D is 0 (normal), the probability of the system state being 0 (normal) is 1. Only when the state of component C is 1 (fault) and the state of component D is 1 (fault) is the probability of the system state being 1 (fault) is 1. Based on this logical relationship, a conditional probability table for the nodes of the Bayesian network of the parallel system can be constructed.

[0067] 3. Converting logic gates to conditional probability tables in k / n voting systems: A k / n voting system refers to a system composed of n components. The system fails when at least k components malfunction. For example, a 3 / 5 voting system consists of 5 components; the system fails when 3, 4, or 5 of these components fail. In a fault tree, this complex logical relationship can be represented by combinations of multiple logic gates, but it essentially follows the k / n rule.

[0068] Specifically, in a Bayesian network, taking a 2 / 3 voting system as an example, let component E, component F, and component G be three nodes in the system. When at least two components are in state 0 (normal), the probability of the system being in state 0 (normal) is 1, i.e., the system is normal in the cases of (E=0, F=0, G=0), (E=0, F=0, G=1), (E=0, F=1, G=0), and (E=1, F=0, G=0). When two or three components are in state 1 (fault), the probability of the system being in state 1 (fault) is 1, i.e., the system is faulty in the cases of (E=0, F=1, G=1), (E=1, F=0, G=1), (E=1, F=1, G=0), and (E=1, F=1, G=1). Based on this logical relationship, a conditional probability table for the nodes of the Bayesian network of the k / n voting system is constructed.

[0069] In this invention, as can be seen from the system reliability block diagram, a complex industrial system is a complex system composed of three typical systems: series systems, parallel systems, and k / n systems. For example, in one embodiment, the complex industrial system is a typical series system, which is composed of subsystems connected in series. Figure 2 This is a schematic diagram of a Bayesian network for a complex industrial system provided by the present invention. A Bayesian network established using the direct method can be referenced. Figure 2 As shown, the top event of the fault tree, which is the leaf node of the Bayesian network, represents a complex industrial system (denoted as T). The root node of the Bayesian network represents subsystems (ST1 to ST8, respectively). This invention employs the traversal minimum cut set method. The minimum cut sets that make up the system are in a series relationship, meaning that if one cut set fails, the entire system fails. Based on these logical relationships, a conditional probability table for the Bayesian network can be established.

[0070] Based on the different composition types of complex industrial systems, this invention converts the logic gates in the system operation fault tree into a conditional probability table of a Bayesian network, thereby utilizing the Bayesian network to perform more in-depth analysis and prediction of system faults.

[0071] Based on the above embodiments, the method further includes: When it is determined that the system operation mode of the complex industrial system has changed, the target equipment failure rate is obtained, wherein the target equipment failure rate is the equipment failure rate corresponding to each device under the changed system operation mode. The modified system operation mode and the target equipment failure rate are input into the risk assessment model to obtain the intermediate risk assessment results corresponding to each subsystem in the complex industrial system, output by the Bayesian network of each subsystem, and the failure risk assessment results output by the Bayesian network of the complex industrial system based on the intermediate risk assessment results.

[0072] In this invention, complex industrial systems employ different operating modes under varying operational scenarios. For instance, during peak production periods, the system may operate at full load, with equipment running at high intensity for extended periods. During the trial operation phase after equipment maintenance, the system may operate at low load and intermittently. Additionally, energy-saving and emergency operating modes may also exist. When the system's operating mode changes, factors such as the equipment's working environment, load, and operating time all change, directly impacting the equipment's failure probability and the overall system's failure risk. Therefore, accurately identifying changes in the system's operating mode is a crucial prerequisite for subsequent risk assessment.

[0073] The target equipment failure rate refers to the probability of failure of various devices in a complex industrial system under the new operating mode after a change in system operation mode. Due to the change in system operation mode, the working pressure and operating conditions of each device may differ, resulting in varying failure rates. For example, in full-load operation mode, motors may experience an increased failure rate due to prolonged high-load operation; while in energy-saving operation mode, the operating frequency of some auxiliary equipment decreases, and their failure rate may decrease accordingly.

[0074] In this invention, the failure rate of the target equipment can be obtained through multiple methods. On the one hand, historical data can be referenced to analyze the historical failure patterns of various devices under similar operating modes, and the frequency of failures can be statistically analyzed to provide an estimate of the failure rate. On the other hand, expert experience can be used to provide an estimate of the failure rate of each device under new operating modes. In addition, information provided by the equipment manufacturer can be used to understand the performance parameters and failure probability information of the equipment under different operating conditions.

[0075] In this invention, the risk assessment model typically consists of Bayesian networks for each subsystem and a Bayesian network for the complex industrial system. The subsystem Bayesian network is established for each subsystem within the complex industrial system (such as the drive subsystem, control subsystem, and power supply system), and it describes the fault logic relationships between the various devices within the subsystem and the fault risk of the subsystem. The complex industrial system Bayesian network, based on the subsystem Bayesian network, further integrates the interrelationships between the various subsystems to assess the fault risk of the entire complex industrial system.

[0076] Furthermore, the modified system operating mode and the target equipment failure rate are input into the risk assessment model. The system operating mode information helps the model understand the current working environment and operating conditions of the equipment, thereby more accurately assessing the failure risk. The target equipment failure rate provides the model with basic data on the failure probability of each device under the new operating mode. Based on this input information, and combining the conditional probability table defined in the Bayesian network with the logical relationships between nodes, the model performs complex probability calculations and inferences.

[0077] In this invention, the Bayesian networks of each subsystem reason and calculate the fault logic relationships within the subsystem based on the input target equipment failure rate and system operating mode information, outputting intermediate risk assessment results for each subsystem in the complex industrial system. These results reflect the probability or risk level of failure of each subsystem under the new system operating mode. For example, the Bayesian network of the transmission subsystem might output a failure probability of 0.2 for the transmission subsystem under full-load operating mode, and the Bayesian network of the control subsystem might output a failure probability of 0.15 for the control subsystem under the new operating mode, etc.

[0078] Then, the Bayesian network for complex industrial systems receives the intermediate risk assessment results output by the Bayesian networks of each subsystem. Combining this with the interrelationships between subsystems, it performs further reasoning and calculations, ultimately outputting the failure risk assessment result for the entire complex industrial system. This result comprehensively considers the failure risks of each subsystem and their interactions, providing a more comprehensive and accurate reflection of the overall failure risk level of the complex industrial system under the new operating mode. For example, the Bayesian network for complex industrial systems might output a failure probability of 0.3 for the entire system under the new operating mode, providing important information for enterprise risk management and decision-making.

[0079] In one embodiment, the method for assessing the failure risk of complex industrial systems provided by the present invention is described in general. Figure 3 A schematic diagram of the risk assessment model provided by this invention can be referred to. Figure 3 As shown, the risk assessment model aims to comprehensively and accurately evaluate the risk level of the system under different operating conditions, providing a scientific basis for decisions on the safe operation and maintenance of the system. The risk assessment model consists of three core parts: the input module, the subsystem risk reasoning module, and the main system risk reasoning module. These modules cooperate with each other to complete the entire process from data input to risk assessment result output.

[0080] Specifically, the input module provides foundational data for subsequent failure risk reasoning and includes two key input files: State.json and Failure-rate.json. The State.json file contains information about the system's current operational status, specifically covering system state and task duration. System state describes whether the system is operating normally, experiencing partial failure, or being completely paralyzed; task duration records the duration of the currently executing task. The State.json file also allows specifying the Bayesian network model to be visualized. In complex industrial systems, multiple subsystems and their interconnected complex network structures exist. By explicitly specifying this in the file, operators can focus on the Bayesian network of a specific subsystem or the entire system to more intuitively observe the risk reasoning process and results.

[0081] The `failure-rate.json` file records detailed information about the equipment names and their corresponding failure rates. Equipment names accurately identify each component in the system, while the failure rate reflects the likelihood of a device failing under specific operating conditions. Failure rates vary among different devices due to differences in their structure, function, usage frequency, and maintenance status. Failure rate information is a key basis for risk assessment. When calculating risk, the system considers the failure rates of different devices, combined with the importance and relevance of each device within the system, to calculate the contribution of each device to the overall risk. For example, the failure of core equipment can severely impact the operation of the entire system, and its failure rate directly affects the assessment result of the overall system risk level.

[0082] Furthermore, based on the current system's actual operating mode, a comprehensive and detailed assessment of the equipment's failure probability is conducted. This requires considering various factors, including the equipment's historical operating data, technical parameters provided by the manufacturer, the actual working environment, and recent maintenance records. Through comprehensive analysis of these factors, a relatively accurate equipment failure probability assessment result can be generated and compiled into two input files: State.json and Failure-rate.json, providing data support for the subsequent operation of the risk assessment model.

[0083] In this invention, the subsystem risk reasoning module (i.e., the subsystem Bayesian network) is a crucial component of the risk assessment model. It is responsible for performing independent risk reasoning and analysis on each subsystem within a complex industrial system. Each subsystem Bayesian network (implemented via Subsystem.py) possesses the capability to perform independent risk reasoning. Upon receiving data from the State.json and Failure-rate.json files from the input module, the subsystem Bayesian network performs risk calculations based on this data. Specifically, it uses the variable elimination algorithm to perform probabilistic reasoning based on information such as system status, task time, and equipment failure rate recorded in the files, combined with its internally defined Bayesian network model and conditional probability table.

[0084] After a series of calculations and inferences, each subsystem's Bayesian network generates intermediate risk assessment results. These results quantitatively reflect the risk status of each subsystem under its current operating conditions, such as the probability of subsystem failure and the potential extent of loss. The intermediate risk assessment results are a crucial output of the subsystem's Bayesian network, serving as the foundational data for subsequent risk inference in the main system.

[0085] In this invention, the risk analysis function of each subsystem Bayesian network is implemented through an independent Python script (such as subsystem1.py, subsystem2.py, etc.). This modular design makes the development and maintenance of each subsystem Bayesian network more flexible and convenient. The Python scripts for different subsystem Bayesian networks can be customized with corresponding risk inference algorithms and model parameters according to the specific characteristics and needs of the subsystem, thereby more accurately assessing the risk level of the subsystem. After completing the risk analysis, the Python scripts of each subsystem Bayesian network return the intermediate risk assessment results to the main system (i.e., the complex industrial system Bayesian network) for further comprehensive processing.

[0086] The main system risk reasoning module (i.e., the Bayesian network for complex industrial systems) is the core of the entire risk assessment model. It is responsible for coordinating the work of various subsystems and performing a comprehensive risk assessment of the entire complex industrial system. Specifically, the main system script Mainsystem.py, as the core program of the main system risk reasoning module, first receives data from the State.json and Failure-rate.json files from the input module; then, it sequentially calls each subsystem in a preset order to perform local risk reasoning. During the call process, input data is passed to each subsystem, ensuring smooth data interaction and collaborative work between subsystems.

[0087] After receiving the intermediate risk assessment results returned by the Bayesian networks of each subsystem, the main system's risk inference module summarizes and analyzes these results. It comprehensively considers the risk status of each subsystem and their interrelationships and impacts. Based on the constructed Bayesian network model of the entire system, it employs more complex probabilistic inference algorithms to obtain a risk assessment result (Risk-Probability.csv) at the level of a complex industrial system. This result comprehensively and accurately reflects the overall risk level of the entire system under its current operating state, providing crucial information for enterprise risk management and decision-making.

[0088] In this invention, according to the operator's instructions, the main system risk reasoning module also has the function of visualizing the assessment results. It can display the system-level risk assessment results in intuitive graphics and charts (such as Baycsian-model-X.png). For example, it can display the risk contribution ratio of different subsystems through bar charts, and display the probability range of system failure through probability distribution charts. This visual display method enables the operator to understand the risk status of the system more clearly.

[0089] In this invention, the operator assesses the probability of equipment failure based on the current system operating mode, generating State.json and Failure-rate.json input files. Then, the risk assessment model is run. The model first reads the parameter information from the input files. Based on the variable elimination algorithm, the model sequentially calls each subsystem (implemented through Subsystem.py) to perform local risk inference. Each subsystem calculates based on the input data, generates intermediate risk assessment results, and returns them to the main system. Furthermore, the main system script Mainsystem.py summarizes the risk assessment results of the subsystems and combines them with the Bayesian network model of the entire system to perform a probabilistic risk assessment at the level of a complex industrial system.

[0090] Figure 4 This is a schematic diagram of the overall business process of the risk assessment model provided by the present invention, which can be referred to. Figure 4 As shown, in this invention, it is first necessary to clarify the current operating mode of the system. Different operating modes (such as full-load operation, low-load operation, maintenance mode, etc.) will affect the failure probability of the equipment. At the same time, the failure probability of key equipment needs to be assessed, which can be determined through various methods such as historical data statistics, expert experience, and information provided by equipment manufacturers.

[0091] Next, based on the system's structure and functions, Bayesian network models of each subsystem need to be constructed, and the parameters of the nodes and edges in the network need to be initialized. Using Bayesian network inference algorithms, the failure probability of each subsystem is calculated based on the input device failure probabilities and the structure of the Bayesian network. The inference process considers the mutual influence and dependencies between devices within the subsystems.

[0092] After obtaining the failure probabilities of the subsystems, these results are used as input to instantiate the Bayesian network of the main system. The Bayesian network of the main system describes the interrelationships between the subsystems and the logical structure of the overall system. Using the Bayesian network inference algorithm again, based on the structure of the main system's Bayesian network and the failure probabilities of the subsystems, the failure probability of the entire main system is calculated. This step comprehensively considers the failure scenarios of all subsystems and their impact on the main system.

[0093] Furthermore, the calculated main system failure probability and other relevant risk assessment results are presented visually, such as through charts and graphs. Visualization helps operators and managers understand the system's risk status more intuitively, enabling them to determine whether to change the system's operating mode based on the visualization results and actual needs.

[0094] If a change in operating mode is required, the probability of equipment failure needs to be reassessed, and the process must return to the "Determine the current system operating mode and assess the probability of critical equipment failure" step to conduct a new risk assessment. If no change in operating mode is required, the process ends.

[0095] This invention organizes fault information of key equipment in complex industrial systems, forming a corresponding fault information table. Based on this information, and according to the system reliability diagram, a qualitative fault tree analysis is performed on the complex industrial system and its subsystems. A static logical model is constructed, with the failure of the complex industrial system as the top event, subsystems as intermediate events, and key equipment as basic events, laying the logical foundation for the subsequent risk assessment model development. Next, using the static logical model as the core, and based on Bayesian network theory and the characteristics of system operation modes, the transformation from fault tree events to Bayesian network nodes, and from fault tree logic gates to directed arcs and conditional probability tables in the Bayesian network, is completed, building the corresponding Bayesian network model. Finally, according to the technical architecture, a risk assessment model is developed using Python to realize probabilistic risk assessment and result display under different operating states of the complex industrial system.

[0096] The following describes the fault risk assessment device for complex industrial systems provided by the present invention. The fault risk assessment device for complex industrial systems described below can be referred to in correspondence with the fault risk assessment method for complex industrial systems described above.

[0097] Figure 5 A schematic diagram of the structure of the fault risk assessment device for complex industrial systems provided by the present invention is shown below. Figure 5 As shown, the present invention provides a fault risk assessment device for complex industrial systems, including a data acquisition module 501 and a risk assessment module 502. The data acquisition module 501 is used to acquire the equipment failure rate corresponding to each device in the complex industrial system, as well as the current real-time system operation mode of the complex industrial system. The risk assessment module 502 is used to input the equipment failure rate and the real-time system operation mode into a risk assessment model to obtain the current fault risk assessment result of the complex industrial system output by the risk assessment model. The risk assessment model is constructed based on fault tree analysis and Bayesian network.

[0098] The fault risk assessment device for complex industrial systems provided by this invention obtains the failure rate and current real-time operating mode of each device in a complex industrial system, and then inputs the two into a risk assessment model constructed based on fault tree analysis and Bayesian network. This model can deeply explore the causal relationship of faults, effectively resist interference, and quickly and accurately output the current fault risk assessment results, significantly improving the accuracy and efficiency of the assessment.

[0099] The system provided in this embodiment of the invention is used to execute the above-described method embodiments. For specific processes and details, please refer to the above embodiments, which will not be repeated here.

[0100] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 6 As shown, the electronic device may include: a processor 601, a communications interface 602, a memory 603, and a communication bus 604, wherein the processor 601, communications interface 602, and memory 603 communicate with each other via the communication bus 604. The processor 601 can call logical instructions in the memory 603 to execute a fault risk assessment method for complex industrial systems. This method includes: obtaining the equipment failure rate corresponding to each device in the complex industrial system, and the current real-time system operation mode of the complex industrial system; inputting the equipment failure rate and the real-time system operation mode into a risk assessment model to obtain the current fault risk assessment result of the complex industrial system output by the risk assessment model, wherein the risk assessment model is constructed based on fault tree analysis and Bayesian networks.

[0101] Furthermore, the logical instructions in the aforementioned memory 603 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0102] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute the complex industrial system fault risk assessment method provided by the above methods, the method comprising: obtaining the equipment failure rate corresponding to each device in the complex industrial system, and the current real-time system operation mode of the complex industrial system; inputting the equipment failure rate and the real-time system operation mode into a risk assessment model, and obtaining the current fault risk assessment result of the complex industrial system output by the risk assessment model, wherein the risk assessment model is constructed based on fault tree analysis and Bayesian network.

[0103] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program is implemented to perform the fault risk assessment method for complex industrial systems provided in the above embodiments. The method includes: obtaining the equipment failure rate corresponding to each device in the complex industrial system and the current real-time system operation mode of the complex industrial system; inputting the equipment failure rate and the real-time system operation mode into a risk assessment model to obtain the current fault risk assessment result of the complex industrial system output by the risk assessment model, wherein the risk assessment model is constructed based on fault tree analysis and Bayesian networks.

[0104] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0105] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable 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, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0106] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for failure risk assessment of a complex industrial system, characterized by, The method comprises: obtaining device failure rates corresponding to respective devices in a complex industrial system and a current real-time system operation mode of the complex industrial system; inputting the device failure rates and the real-time system operation mode into a risk assessment model to obtain a current failure risk assessment result of the complex industrial system output by the risk assessment model, wherein the risk assessment model is constructed based on a fault tree analysis method and a Bayesian network.

2. The method of claim 1, wherein, The risk assessment model is constructed by the following steps: based on the fault tree analysis method and historical failure event data of the complex industrial system, constructing a system operation fault tree of the complex industrial system corresponding to different historical system operation modes; determining input-output relationships between respective failure events in the system operation fault tree, and constructing a subsystem Bayesian network corresponding to respective subsystems of the complex industrial system under different historical system operation modes according to the failure events and the input-output relationships; based on respective subsystem Bayesian networks, constructing a complex industrial system Bayesian network; constructing the risk assessment model according to the complex industrial system Bayesian network and respective subsystem Bayesian networks.

3. The method of claim 2, wherein, The construction of the system operation fault tree of the complex industrial system corresponding to different historical system operation modes based on the fault tree analysis method and the historical failure event data of the complex industrial system comprises: based on the fault tree analysis method, determining basic events, intermediate events and top events corresponding to the historical failure event data under different historical system operation modes, wherein the basic events are device failure events in the historical failure event data; the intermediate events are subsystem failure events in the historical failure event data; and the top events are complex industrial system failure events in the historical failure event data; constructing the system operation fault tree according to the basic events, the intermediate events and the top events.

4. The method of claim 3, wherein, The construction of the subsystem Bayesian network corresponding to respective subsystems of the complex industrial system under different historical system operation modes according to the failure events and the input-output relationships comprises: based on the basic events, intermediate events and top events, constructing respective nodes in the Bayesian network, wherein the basic events correspond to root nodes in the Bayesian network; the intermediate events correspond to intermediate nodes in the Bayesian network; and the top events correspond to leaf nodes in the Bayesian network; based on the input-output relationships, constructing directed arcs in the Bayesian network; based on the logical gates in the system operation fault tree, constructing conditional probability tables in the Bayesian network; constructing the subsystem Bayesian network according to the root nodes, the intermediate events, the directed arcs and the conditional probability tables corresponding to respective subsystems of the complex industrial system under different historical system operation modes.

5. The method of claim 4, wherein, The construction of the conditional probability tables in the Bayesian network based on the logical gates in the system operation fault tree comprises: The logic gate in the system operation fault tree is converted into the conditional probability table according to a system composition type of the complex industrial system, wherein the system composition type comprises a series system type, a parallel system type, and a k / n voting system type.

6. The method of claim 2 to 5, wherein, The method further comprises: When determining that a system operation mode of the complex industrial system is changed, obtaining a target device failure rate, wherein the target device failure rate is a device failure rate corresponding to each device under the changed system operation mode; inputting the changed system operation mode and the target device failure rate into the risk assessment model to obtain an intermediate risk assessment result corresponding to each subsystem in the complex industrial system output by each of the subsystem Bayesian networks, and the failure risk assessment result output by the complex industrial system Bayesian network based on each of the intermediate risk assessment results.

7. A complex industrial system failure risk assessment apparatus, characterized by, comprises: a data acquisition module configured to obtain a device failure rate corresponding to each device in a complex industrial system and a real-time system operation mode of the complex industrial system; a risk assessment module configured to input the device failure rate and the real-time system operation mode into a risk assessment model to obtain a current failure risk assessment result of the complex industrial system output by the risk assessment model, wherein the risk assessment model is constructed based on a fault tree analysis method and a Bayesian network.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, The processor executes the computer program to implement the complex industrial system failure risk assessment method according to any one of claims 1 to 6. 9.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the complex industrial system failure risk assessment method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the complex industrial system failure risk assessment method according to any one of claims 1 to 6.

Images