Fault prediction method and system based on FTA fault tree

Abstract

The application discloses a fault prediction method and system based on FTA fault tree, and relates to the technical field of data processing.The method comprises the following steps: acquiring historical equipment state parameters, performing parameter correlation degree analysis, and acquiring correlation parameters; acquiring a historical fault event library, extracting fault state parameters as bottom nodes of the FTA fault tree; extracting equipment fault types from the historical fault event library as middle nodes of the FTA fault tree, and extracting global fault results as top nodes of the FTA fault tree; based on the correlation parameters and the historical fault event library, acquiring node fault contribution degree parameters, and performing weight distribution on the bottom nodes and the middle nodes; searching for node correlation characteristics in the historical fault event library, performing logic gate configuration, and completing construction of the FTA fault tree; and inputting current hydropower equipment state parameters into the FTA fault tree, and acquiring a predicted fault type and probability.The application effectively improves the accuracy and real-time performance of fault prediction.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to a fault prediction method and system based on FTA fault trees. Background Technology

[0002] As hydropower equipment develops towards intelligent, integrated, and long-term stable operation, equipment operation status monitoring and fault prediction have become key aspects of ensuring the safe operation and maintenance of hydropower systems. Current hydropower equipment fault analysis primarily employs the traditional Fault Tree Analysis (FTA) method. This method is mainly used for qualitative tracing and logical delineation after a fault occurs, relying heavily on manual experience to construct the fault tree and classify fault levels.

[0003] However, existing methods have obvious shortcomings in practical applications: the weight allocation of fault nodes and the configuration of logic gates rely too much on subjective judgment, and the contribution of faults cannot be accurately quantified; at the same time, the fault tree hierarchy is vague and it is difficult to adapt to the multi-cause and multi-level fault propagation patterns of hydropower equipment, resulting in insufficient accuracy and real-time performance of fault prediction, and failing to meet the actual needs of predicting equipment faults in advance. Summary of the Invention

[0004] This invention provides a fault prediction method and system based on FTA fault trees, aiming to solve the technical problems of insufficient accuracy and real-time performance in fault prediction in the prior art.

[0005] In view of the above problems, the present invention provides a fault prediction method and system based on FTA fault tree.

[0006] In a first aspect, the present invention provides a fault prediction method based on FTA fault trees, comprising: Historical device status parameters are obtained, and parameter correlation analysis is performed on the historical device status parameters to obtain correlation parameters; Obtain the historical fault event database and extract the fault status parameters corresponding to the fault events as the bottom-level nodes of the FTA fault tree; The equipment fault types are extracted from the historical fault event database and used as the middle-level nodes of the FTA fault tree, and the global fault results are extracted and used as the top-level nodes of the FTA fault tree. Based on the correlation parameters and the historical fault event database, the node fault contribution parameters are obtained, and weights are assigned to the bottom-level nodes and the middle-level nodes. Node correlation characteristics are retrieved from the historical fault event database, and the logic gates of the FTA fault tree are configured based on the node correlation characteristics to complete the construction of the FTA fault tree. Input the current status parameters of the hydropower equipment into the FTA fault tree to obtain the predicted fault type and predicted fault probability.

[0007] Secondly, the present invention provides a fault prediction system based on FTA fault trees, comprising: The parameter correlation analysis module is used to obtain historical device status parameters and perform parameter correlation analysis on the historical device status parameters to obtain correlation parameters. The bottom-level node extraction module is used to obtain a historical fault event database and extract the fault status parameters corresponding to the fault events as the bottom-level nodes of the FTA fault tree. The middle and top-level node extraction module is used to extract equipment fault types from the historical fault event database as middle-level nodes of the FTA fault tree, and extract global fault results as top-level nodes of the FTA fault tree. The node weight allocation module is used to obtain the node fault contribution parameter based on the correlation parameter and the historical fault event database, and to allocate weights to the bottom-level nodes and the middle-level nodes. The logic gate configuration module is used to retrieve node correlation characteristics from the historical fault event database, and configure the logic gates of the FTA fault tree based on the node correlation characteristics to complete the construction of the FTA fault tree. The fault prediction output module is used to input the current status parameters of the hydropower equipment into the FTA fault tree to obtain the predicted fault type and the predicted fault probability.

[0008] One or more technical solutions provided in this invention have at least the following technical effects or advantages: This invention provides a fault prediction method and system based on FTA (Fault Tree Analysis). By analyzing the correlation of historical equipment state parameters, it extracts correlation parameters as the basis for constructing bottom-level nodes, solving the problem of strong subjectivity in traditional fault tree node selection. Based on correlation parameters and a historical fault event database, it analyzes the fault contribution of nodes, achieving objective weight allocation for bottom-level and mid-level nodes, avoiding bias caused by experience-based weight assignment. By retrieving node correlation features and configuring logic gates, the fault tree structure better reflects actual fault propagation mechanisms, improving the model's ability to describe complex fault logic. Finally, the current equipment state parameters are input into the constructed FTA fault tree to accurately predict fault types and probabilities. This invention effectively improves the accuracy and real-time performance of fault prediction for hydropower equipment, reduces false alarm and missed alarm rates, effectively reduces the risk of unplanned equipment downtime, and provides reliable data support and decision-making basis for intelligent equipment operation and maintenance. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating the fault prediction method based on FTA fault tree provided in an embodiment of the present invention. Figure 2A schematic diagram of the structure of a fault prediction system based on FTA fault tree provided in an embodiment of the present invention; The components represented by each number in the attached diagram are explained below: Parameter correlation analysis module 11, bottom layer node extraction module 12, middle and top layer node extraction module 13, node weight allocation module 14, logic gate configuration module 15, and fault prediction output module 16. Detailed Implementation

[0010] This invention provides a fault prediction method and system based on FTA fault trees, which is used to address the technical problems of insufficient accuracy and real-time performance in fault prediction in the prior art.

[0011] Example 1, as Figure 1 As shown, this invention provides a fault prediction method based on FTA fault trees, the method comprising: S100: Obtain historical device status parameters, and perform parameter correlation analysis on the historical device status parameters to obtain correlation parameters.

[0012] In this embodiment of the invention, historical equipment status parameters are acquired, and parameter correlation analysis is performed on these parameters to obtain correlated parameters. During the operation of hydropower equipment, its status is influenced by multiple dimensions of parameters, including electrical, mechanical, operating conditions, and environmental factors. These parameters are not independent but exhibit potential linear or nonlinear correlations. Traditional FTA fault tree construction does not systematically review and analyze the historical equipment status parameters; it relies solely on manual experience to select parameters. This can easily lead to the omission of key correlated parameters and the introduction of irrelevant parameters, resulting in inaccurate fault tree node division and unreasonable weight allocation, thus affecting the accuracy of fault prediction. Therefore, it is necessary to first acquire the historical status parameters of the hydropower equipment and then extract correlated parameters through scientific correlation analysis to provide accurate and reliable basic data support for the subsequent construction of the FTA fault tree.

[0013] Step S100 in the method provided in this embodiment of the invention includes: Among them, obtaining historical device status parameters includes: Electrical, mechanical, operating condition, and environmental parameters of hydropower equipment are collected as historical equipment status parameters. The electrical parameters include at least current and voltage; the mechanical parameters include at least rotational speed and pressure; the operating condition parameters include at least switch open / closed status and protection action signals; and the environmental parameters include at least temperature and humidity. Historical equipment status parameters refer to various quantitative or qualitative data recorded by monitoring sensors and data acquisition systems during the past operating cycles of the hydropower equipment, reflecting the equipment's operating status. These parameters serve as the foundation for analyzing equipment failure patterns and uncovering parameter correlations.

[0014] Specifically, through the monitoring system built into the hydroelectric equipment, external sensors, and operation and maintenance records, electrical parameters, mechanical parameters, operating status parameters, and environmental parameters throughout the entire life cycle of the equipment are comprehensively collected to ensure that the parameters cover the key dimensions of equipment operation. After collection, deduplication and completion preprocessing are performed to remove abnormal and invalid data, and finally, a historical equipment status parameter dataset is formed.

[0015] For example, taking a hydropower station turbine generator unit as the research object, historical operating parameters of the unit over the past three years were collected, specifically including: electrical parameters: current: stator current of the unit, ranging from 800-1200A; voltage: outlet voltage of the unit, ranging from 9.5-10.5kV; mechanical parameters: speed: rotor speed of the unit, ranging from 2950-3050r / min; pressure: bearing oil pressure of the unit, ranging from 0.3-0.5MPa; operating condition parameters: switch opening and closing status: real-time records of the unit's main switch being closed and open, with 1 indicating closed and 0 indicating open; protection action signals: action records of overcurrent protection and overvoltage protection, with 1 indicating action and 0 indicating no action; environmental parameters: temperature: ambient temperature of the unit's generator room, ranging from 15-35℃; humidity: relative humidity of the generator room, ranging from 40%-60%. After collection, abnormal data caused by three sets of sensor malfunctions were removed, and two sets of missing humidity data were supplemented to form a complete historical equipment status parameter dataset.

[0016] Specifically, a parameter correlation analysis is performed on the historical equipment status parameters to obtain correlation parameters, including: The historical equipment status parameters are standardized to obtain standard hydroelectric equipment parameters; Calculate the Pearson correlation coefficient between any two standardized parameters as the correlation value; By integrating the correlation values ​​among all parameters, a correlation parameter is formed.

[0017] First, the historical equipment status parameters are standardized to obtain standard hydropower equipment parameters. Standardization refers to converting the original historical equipment status parameters of different magnitudes and units into standard parameters of a uniform magnitude and without units, eliminating the impact of dimensional differences and ensuring the accuracy and comparability of subsequent correlation calculations.

[0018] Specifically, the z-score standardization method is used to standardize the preprocessed historical equipment state parameters for each category. The calculation formula is: z = (x - μ) / σ, where z is the standardized parameter value, x is the original historical equipment state parameter value, μ is the mean of all original values ​​for that category of parameters, and σ is the standard deviation of all original values ​​for that category of parameters. After the calculation is completed, the standard hydropower equipment parameters for all parameters are obtained.

[0019] For example, Z-score standardization is performed on the electrical, mechanical, operating condition, and environmental parameters in the historical equipment status parameters. For instance, the stator current parameter has a mean μ = 1000A and a standard deviation σ = 141.42A. Standardizing one set of original values ​​x = 800A yields z = (800 - 1000) / 141.42 ≈ -1.41. Similarly, the same method is used to standardize all the collected parameters such as voltage, speed, pressure, temperature, and humidity to obtain standard hydroelectric equipment parameters.

[0020] Secondly, the Pearson correlation coefficient between any two standardized parameters is calculated as the correlation strength value. The Pearson correlation coefficient is a statistic that measures the degree of linear correlation between two variables, with a value ranging from -1 to 1. A Pearson correlation coefficient closer to 1 indicates a strong positive correlation between the two parameters; a value closer to -1 indicates a strong negative correlation; and a value close to 0 indicates no significant linear correlation. For all standard hydropower equipment parameters, pairwise parameters are used, and the correlation strength value between each pair of parameters is calculated using the Pearson correlation coefficient formula, which is the product of the covariance of the two variables divided by their respective standard deviations.

[0021] For example, by pairing the standardized stator current and output voltage and substituting them into the Pearson formula, the correlation coefficient between the stator current and output voltage is approximately 1.0, indicating a strong positive correlation between them. Then, by selecting the standardized stator current and the computer room temperature for calculation, a correlation coefficient of approximately 0.15 is obtained, indicating no significant linear correlation between the stator current and temperature. The same method is used to calculate the correlation coefficients between all pairwise parameters.

[0022] Finally, the correlation values ​​between all parameters are integrated to form the correlation parameters. The correlation parameters are a set formed by integrating the correlation values ​​of all pairwise device status parameters in matrix form. The rows and columns of the matrix correspond to various types of device status parameters, and each element in the matrix is ​​the correlation coefficient between the corresponding row and column parameters, which can intuitively present the distribution of the strength of the correlation between all parameters.

[0023] Specifically, all historical device status parameters are analyzed, and the parameter names are used as rows and columns of the correlation matrix. The correlation value of each pair of parameters is filled into the corresponding position in the matrix to form a complete correlation matrix, which is the correlation parameter. If there are parameter pairs whose correlation has not been calculated, such as the same parameter itself, then 0 is filled in to indicate no correlation.

[0024] For example, based on eight parameters—stator current, output voltage, rotor speed, bearing oil pressure, switch open / closed status, protection action signal, machine room temperature, and machine room humidity—correlation values ​​are integrated to form an 8×8 correlation matrix. Some key elements are as follows: stator current and output voltage correlation value 1.0, stator current and rotor speed correlation value 0.85, stator current and machine room temperature correlation value 0.15, rotor speed and bearing oil pressure correlation value 0.92, machine room temperature and machine room humidity correlation value 0.76, switch open / closed status and protection action signal correlation value 0.88. The correlation values ​​of other parameters with no obvious correlation are all below 0.2. After filling all correlation values ​​into the matrix, the complete correlation parameters are obtained.

[0025] In this embodiment of the invention, by comprehensively collecting multi-dimensional historical state parameters of hydropower equipment, covering four dimensions—electrical, mechanical, operating conditions, and environment—the problem of partial parameter collection and omission of key parameters in traditional methods is avoided, providing comprehensive basic data for subsequent analysis. Through z-score standardization, the dimensional differences between different parameters are effectively eliminated, solving the problem of inaccurate correlation calculations caused by different parameter magnitudes. The Pearson correlation coefficient is used to accurately quantify the degree of correlation between parameters, replacing the subjective method of traditional manual judgment of parameter correlation, ensuring the scientific and objective nature of the correlation analysis. The finally integrated correlation parameters clearly present the distribution of the strength of correlations among the state parameters of each device, accurately screening out correlation parameters that have a significant impact on equipment failure and eliminating irrelevant parameters. This provides accurate data support for subsequent FTA fault tree bottom-level node extraction, node weight allocation, and logic gate configuration, reducing the interference of invalid data on subsequent analysis and improving the overall operational efficiency of the method.

[0026] S200: Obtain the historical fault event database and extract the fault status parameters corresponding to the fault events as the bottom-level nodes of the FTA fault tree.

[0027] In this embodiment of the invention, a historical fault event database is obtained, and the fault state parameters corresponding to the fault events are extracted as the bottom-level nodes of the FTA fault tree. The bottom-level nodes of the FTA fault tree are the foundation of fault analysis and prediction. The accuracy and comprehensiveness of the bottom-level nodes directly determine the rationality of the fault tree construction and the accuracy of fault prediction. In traditional FTA fault tree construction, the bottom-level nodes are mostly selected based on manual experience, without combining the extraction with fault data throughout the equipment's life cycle. This easily leads to problems such as missing bottom-level nodes, mismatch with the actual root causes of faults, and fuzzy node representations, resulting in chaotic subsequent fault propagation logic and large deviations in fault prediction. Combining the correlation parameters obtained in step S100, it is necessary to first collect relevant records of faults throughout the life cycle of hydropower equipment, integrate them to form a historical fault event database, and accurately extract the fault state parameters corresponding to the root causes of faults as bottom-level nodes, ensuring that the bottom-level nodes are highly consistent with the actual fault scenario.

[0028] Step S200 in the method provided in this embodiment of the invention includes: Collect fault occurrence records, fault investigation records, and fault handling records throughout the entire life cycle of hydropower equipment, and integrate them to form a historical fault event database. The historical fault event database includes information on the root cause of the fault, the status of the corresponding equipment parameters, the time of the fault occurrence, and the scope of the fault's impact. Extract the abnormal values ​​of the hydropower equipment status parameters corresponding to each fault root cause from the historical fault event database, and take the union of each fault root cause as the fault status parameter. Each fault state parameter is used as an independent FTA fault tree bottom node, and each bottom node uniquely corresponds to the parameter representation of the root cause of a type of hydropower equipment fault.

[0029] First, records of fault occurrence, troubleshooting, and handling throughout the entire lifecycle of hydropower equipment are collected and integrated to form a historical fault event database. This database includes information on the root cause of the fault, the corresponding equipment parameter status, the time of occurrence of the fault, and the scope of its impact. The historical fault event database is a dataset that integrates all complete fault-related records from the commissioning of hydropower equipment to its current operating cycle; it serves as the data source for extracting the root cause of the fault and fault status parameters.

[0030] The fault occurrence record refers to the basic information recorded when the fault occurred, such as the time, operating conditions, and initial symptoms. The fault investigation record is a detailed record of the maintenance personnel's investigation of the fault cause and location after the fault occurred. The fault handling record is a record of the repair measures taken, the handling results, and the fault reproduction status after the fault investigation was completed. The root cause of the fault refers to the most fundamental reason for the failure of the hydroelectric equipment. The fault-corresponding equipment parameter status refers to the specific values ​​of various status parameters of the hydroelectric equipment at the time of the fault. The scope of the fault's impact refers to the degree of impact of the fault on the hydroelectric equipment itself, surrounding equipment, and the operation of the entire hydroelectric system.

[0031] Specifically, the system compiles and organizes data such as operation and maintenance records, monitoring system logs, fault reporting records, and maintenance reports throughout the entire lifecycle of hydropower equipment. It comprehensively collects fault occurrence records, fault investigation records, and fault handling records. The collected records are preprocessed to remove duplicate, invalid, and incomplete records and supplement missing key information. All preprocessed records are sorted by fault occurrence time, and the root causes of faults, corresponding equipment parameter status, fault occurrence time, and fault impact range from various records are integrated to form a structured historical fault event database, which facilitates subsequent extraction and retrieval.

[0032] For example, taking a hydropower station turbine generator unit as the research object, we collected fault-related records for the unit throughout its entire life cycle over the past three years. Specifically, for example: Fault occurrence record: On May 10, 2023, abnormal vibration and a sudden increase in stator current occurred during unit operation, initially identified as an electrical fault; Fault investigation record: After investigation, the root cause of the fault was determined to be poor contact in the stator winding. During the investigation, the stator current was recorded as 1350A at the time of the fault, exceeding the normal range of 800-1200A, and the output voltage was 11.2kV, exceeding the normal range of 9.5-10.5kV; Fault handling record: The stator winding was rewired and tightened, the fault was eliminated, and it did not recur in subsequent operation. The fault affected the unit's single-unit reduced load operation and did not affect other equipment. In this way, for example, 12 fault-related records of the unit in the past 3 years were collected. Two sets of incomplete records without root causes were removed, and two sets of missing fault impact range information were supplemented. After sorting by the fault occurrence time, they were integrated to form a historical fault event database containing 10 complete fault events. Each event includes the fault root cause, the status of the corresponding equipment parameters, the fault occurrence time, and the fault impact range.

[0033] Secondly, abnormal values ​​of hydropower equipment status parameters corresponding to each root cause of failure are extracted from the historical failure event database, and the union of these parameters for each root cause is calculated as the failure status parameter. The failure status parameter refers to the equipment status parameter that is in an abnormal state and can characterize a specific root cause of failure in the hydropower equipment; that is, a parameter that deviates from the normal value range, and is a quantitative representation of the root cause of failure. An abnormal parameter value refers to a value of the equipment status parameter that exceeds the normal value range collected in S100. The union is the set of all unique abnormal parameters obtained by integrating the abnormal parameter values ​​corresponding to multiple root causes and removing duplicate abnormal parameters.

[0034] Specifically, each fault event in the historical fault event database is retrieved one by one, and all abnormal values ​​of equipment status parameters corresponding to the root cause of the fault in that event are extracted. According to the root cause of the fault, all abnormal values ​​of parameters corresponding to the same root cause are grouped into one category. The union of the abnormal values ​​of parameters corresponding to each category of root cause is calculated, that is, duplicate abnormal parameters under the same root cause are removed, and each unique abnormal parameter is retained. The union of all the root causes of the fault is integrated to obtain a complete set of fault status parameters, ensuring that each fault status parameter can correspond to at least one category of root cause and that there are no duplicate parameters.

[0035] For example, the historical fault event database is searched, and abnormal parameter values ​​are extracted for three typical fault root causes and their union is calculated: Stator winding poor contact: the corresponding abnormal parameter values ​​are stator current 1350A and output voltage 11.2kV, and the union is {abnormal stator current, abnormal output voltage}; Bearing wear: the corresponding abnormal parameter values ​​are bearing oil pressure 0.2MPa, lower than the normal range of 0.3-0.5MPa, and rotor speed 3100r / min, higher than the normal range of 2950-3050r / min, and the union is {abnormal bearing oil pressure, abnormal rotor speed}; Excessive temperature and humidity in the computer room: the corresponding abnormal parameter values ​​are computer room temperature 38℃, higher than the normal range of 15-35℃, and computer room humidity 65%, higher than the normal range of 40%-60%, and the union is {abnormal computer room temperature, abnormal computer room humidity}. In this way, the abnormal values ​​of the parameters corresponding to the 8 types of root causes of all 10 fault events are extracted, and the union of each parameter is calculated to obtain a complete set of fault status parameters: {stator current abnormality, outlet voltage abnormality, bearing oil pressure abnormality, rotor speed abnormality, machine room temperature abnormality, machine room humidity abnormality, switch open / closed status abnormality, protection action signal abnormality}. There are no duplicate parameters in this set, and each parameter corresponds to at least one type of root cause of fault.

[0036] Furthermore, each fault state parameter is treated as an independent bottom-level node in the FTA fault tree. Each bottom-level node uniquely corresponds to the parameter representation of a specific type of hydropower equipment fault root cause. The bottom-level node of the FTA fault tree, also known as the basic event node, is the starting point for fault propagation in the fault tree. It directly corresponds to the parameter manifestation of a specific fault root cause in the hydropower equipment and cannot be further decomposed downwards. Unique correspondence means that each bottom-level node represents only one type of fault root cause parameter anomaly. Different bottom-level nodes correspond to different fault root causes or different parameter anomalies of the same fault root cause, avoiding node confusion.

[0037] Specifically, the fault state parameter set is sorted out, and each fault state parameter is treated as an independent bottom-level node of the FTA fault tree. A unique identifier is assigned to each bottom-level node, such as bottom-level node 1 - stator current anomaly. The parameter representation of the fault root cause corresponding to each bottom-level node is clarified to ensure that each bottom-level node corresponds to only one type of fault root cause and one type of parameter anomaly, with no overlap or duplication. All bottom-level nodes are classified and organized according to parameter type to facilitate subsequent association with intermediate nodes and complete the fault tree construction.

[0038] For example, based on the fault state parameter set obtained above, eight independent FTA fault tree bottom-level nodes are constructed. Each node uniquely corresponds to the parameter representation of a type of fault root cause, as follows: bottom-level node 1 - stator current anomaly, bottom-level node 2 - outlet voltage anomaly, ..., bottom-level node 8 - protection action signal anomaly. All bottom-level nodes are categorized and organized according to electrical parameters, mechanical parameters, operating condition parameters, and environmental parameters to ensure that each node uniquely corresponds to the parameter representation of the fault root cause, with no overlap or duplication.

[0039] In this embodiment of the invention, by collecting fault records throughout the entire lifecycle of hydropower equipment and integrating them to form a complete historical fault event database, a comprehensive and authentic data source is provided for fault state parameter extraction. By extracting the parameter outliers corresponding to the root cause of the fault and taking the union, parameters that can characterize the fault are accurately selected. Each fault state parameter is treated as an independent bottom-level node, and each node uniquely corresponds to the parameter representation of the root cause of the fault, ensuring that the bottom-level nodes are both consistent with the actual fault scenario and have data support. The final constructed bottom-level node system provides a precise and standardized underlying foundation for the subsequent extraction of middle-level nodes, node weight allocation, and logic gate configuration in the FTA fault tree, effectively improving the rationality and standardization of fault tree construction, ensuring the accuracy of subsequent fault prediction, and enhancing the practicality and operability of the method.

[0040] S300: Extract the equipment fault type from the historical fault event database as the middle node of the FTA fault tree, and extract the global fault result as the top node of the FTA fault tree.

[0041] In this embodiment of the invention, equipment fault types are extracted from the historical fault event database as mid-level nodes of the FTA fault tree, and global fault results are extracted as top-level nodes of the FTA fault tree. The mid-level nodes of the FTA fault tree serve as the connecting link between the bottom-level and top-level nodes, responsible for reflecting the transmission process from the root cause of the fault to the global fault; the top-level node is the final analysis target of the fault tree, representing the most severe consequences of hydropower equipment failure. Traditional FTA fault tree construction cannot accurately reflect the transmission path of a fault from its root cause to its final consequence, thus affecting the accuracy of fault prediction. Combining the historical fault event database and bottom-level nodes constructed in step S200, equipment-level fault types need to be extracted as mid-level nodes, and global fault results as top-level nodes, clarifying the hierarchical structure of the fault tree from bottom-level to mid-level to top-level, and establishing a complete fault transmission logic.

[0042] Step S300 in the method provided in this embodiment of the invention includes: Extract equipment-level hydropower equipment failure types caused by one or more root causes from the historical failure event database, and use them as independent FTA fault tree mid-level nodes. Extract the global failure results of hydropower equipment caused by one or more equipment-level failures from the historical failure event database, and use them as the top-level node of the FTA fault tree.

[0043] First, equipment-level hydropower equipment fault types caused by one or more root causes are extracted from the historical fault event database and used as independent mid-level nodes in the FTA fault tree. Equipment-level hydropower equipment fault types refer to fault classifications for a specific functional module or the entire hydropower equipment, caused by one or more root causes. They serve as an intermediate carrier between the root cause and the global fault outcome, possessing a clear fault characterization and scope, and can be further decomposed into root causes corresponding to the bottom-level nodes. Mid-level nodes in the FTA fault tree, also known as intermediate event nodes, are located above the bottom-level nodes and below the top-level nodes. They are used to integrate equipment-level faults caused by similar root causes, achieving the transition and propagation of fault levels. Each mid-level node corresponds to an independent equipment-level fault type.

[0044] Specifically, the historical fault event database constructed in step S200 is retrieved, and the root cause and fault characteristics of each fault event are analyzed one by one. Faults caused by the same or similar root causes and with consistent fault manifestations and impact ranges are classified into the same equipment-level fault type. Duplicate and redundant equipment-level fault types are eliminated to ensure that each type of equipment-level fault is unique and independent. Each independent equipment-level fault type is treated as a mid-level node in the FTA fault tree. The root cause of each mid-level node is identified, i.e., the associated bottom-level node. A unique identifier is assigned to each mid-level node, and the nodes are classified and organized according to the fault impact range and type to facilitate subsequent association with bottom-level and top-level nodes.

[0045] For example, based on the historical fault event database constructed in step S200, which contains 10 fault events and 8 types of root causes, equipment-level fault types are extracted as intermediate-level nodes, as follows: Analysis of the fault events reveals that stator winding contact problems, overcurrent faults, and overvoltage faults all cause abnormalities in the unit's electrical system. These faults are characterized by abnormal current and voltage parameters and are classified as electrical system faults, thus becoming intermediate-level node 1, i.e., Intermediate-level node 1 - Electrical system fault. Similarly, intermediate-level nodes 2 - Mechanical transmission fault, intermediate-level node 3 - Environmental adaptation fault, and intermediate-level node 4 - Control and protection fault are obtained. These four intermediate-level nodes are all independent equipment-level fault types, each corresponding to a clear root cause, with no duplication or redundancy. Classified and organized according to fault type, the intermediate-level node extraction is completed.

[0046] Secondly, global failure results of hydropower equipment caused by one or more equipment-level failures are extracted from the historical failure event database and used as the top-level node of the FTA fault tree. A global failure result of hydropower equipment refers to a serious failure consequence caused by one or more equipment-level failures, affecting the normal operation of hydropower equipment and even the stability of the entire hydropower system. It is the ultimate goal of fault tree analysis, possessing globality, severity, and comprehensiveness, and cannot be further decomposed upwards. The top-level node of the FTA fault tree, also known as the top event node, is the highest-level node in the fault tree, uniquely corresponding to the global failure result of the hydropower equipment. All intermediate nodes directly or indirectly point to this top-level node, reflecting the complete transmission path of the failure from its root cause to its final consequence.

[0047] Specifically, the historical fault event database is searched to analyze the final failure consequences caused by equipment-level faults in each fault event. The focus is on screening for serious fault consequences that affect the overall operation of the equipment, cover the entire equipment system, or are related to the hydropower system. Localized or minor fault consequences are eliminated to ensure the extracted fault results are global. If multiple equipment-level faults cause the same global fault consequence, only that unique global fault result is retained. This global fault result is used as the top-level node of the FTA fault tree, and the relationship between the top-level node and each intermediate node is clarified to ensure that the top-level node can cover the fault propagation endpoints of all intermediate nodes.

[0048] For example, based on the historical fault event database of S200 and the four mid-level nodes extracted above, the final consequences of equipment-level faults are analyzed: Mid-level node 1 - a severe electrical system fault will cause the stator of the unit to burn out and fail to generate electricity normally; Mid-level node 2 - a severe mechanical transmission fault will cause the unit rotor to jam and shut down; Mid-level node 3 - a long-term environmental adaptation fault will accelerate equipment aging, indirectly causing electrical and mechanical faults, ultimately leading to unit shutdown; Mid-level node 4 - a severe control and protection fault will cause the unit control to malfunction and fail to start and stop normally, ultimately leading to shutdown. The analysis reveals that all four mid-level nodes ultimately lead to the global fault consequence of turbine generator unit shutdown. This consequence affects the overall operation of the unit and is a global, severe fault result, with no other more serious global fault consequences. Therefore, "hydrogenerator unit shutdown" is extracted as the top-level node of the FTA fault tree, clarifying that this top-level node is jointly or individually caused by mid-level nodes 1, 2, 3, and 4, thus completing the top-level node extraction.

[0049] In this embodiment of the invention, based on the historical fault event database constructed using S200, equipment-level fault types are extracted as mid-level nodes, achieving effective connection between bottom-level and top-level nodes. By classifying equipment-level faults caused by similar root causes, mid-level nodes have clear fault representations and scopes, ensuring clear fault propagation logic while reducing redundancy and improving the simplicity of the fault tree. By extracting global fault results as top-level nodes, the core risks of hydropower equipment operation are focused on, clarifying the final analysis objective of the fault tree. The final constructed three-level node system (bottom-middle-top) improves the hierarchical structure of the FTA fault tree, providing a clear hierarchical foundation for subsequent node weight allocation and logic gate configuration. This ensures that the fault tree accurately reflects the propagation path of the fault from its root cause to its final consequence, further improving the rationality of the fault tree construction and guaranteeing the accuracy of subsequent fault prediction. It also avoids subjective errors from manually dividing node levels, improving the standardization and operability of the method.

[0050] S400: Based on the correlation parameters and the historical fault event database, obtain the node fault contribution parameters, and assign weights to the bottom-level nodes and the middle-level nodes.

[0051] In this embodiment of the invention, based on the correlation parameters and the historical fault event database, node fault contribution parameters are obtained, and weights are assigned to the bottom-level nodes and the middle-level nodes. The node weights of the FTA fault tree directly determine the accuracy of the fault propagation probability calculation. The influence of bottom-level nodes on middle-level nodes and middle-level nodes on top-level nodes differs. If nodes are not scientifically weighted, and only equal weights or subjective manual assignment are used, the predicted fault probability will deviate significantly from the actual fault occurrence pattern, failing to accurately reflect the influence weights of different fault root causes and different equipment-level faults. Combining the correlation parameters obtained in step S100 and the historical fault event database constructed in step S200, node fault contribution parameters need to be obtained through quantitative statistics and normalization processing. Reasonable weights are then assigned to the bottom-level nodes and middle-level nodes, enabling the fault tree to accurately quantify the fault impact of different nodes and improve the accuracy of fault prediction.

[0052] Step S400 in the method provided in this embodiment of the invention includes: The number of times a fault state parameter corresponding to each bottom node triggers a fault type in the middle node device is counted from the historical fault event database. The ratio of this number of occurrences to the total number of occurrences in the corresponding middle node is calculated to obtain the middle-level trigger probability. The mid-layer trigger probability is normalized and used as the node fault contribution parameter of the bottom-layer node. The number of times a device fault type corresponding to each mid-level node causes a global fault result in the top-level node is counted from the historical fault event database. The ratio of this number of occurrences to the total number of occurrences in the top-level node is calculated to obtain the top-level trigger probability. The top-level trigger probability is normalized and used as the node fault contribution parameter of the middle-level node. The node fault contribution parameters of each bottom-level node and middle-level node are used as the weight values ​​of the corresponding nodes to complete the weight allocation of the bottom-level nodes and middle-level nodes.

[0053] First, the number of times each fault state parameter corresponding to a bottom-level node triggers a fault type in a mid-level node is statistically analyzed from the historical fault event database. The ratio of this occurrence to the total number of occurrences for the corresponding mid-level node is then calculated to obtain the mid-level trigger probability. The mid-level trigger probability refers to the relative frequency of a bottom-level node's occurrence among all historical fault events corresponding to its mid-level nodes, reflecting the degree of local contribution of that bottom-level node to the fault of that mid-level node. The mid-level trigger probability is calculated based on the total number of occurrences under that mid-level node and is only comparable within the same mid-level node.

[0054] Specifically, the historical fault event database is traversed. For each mid-level node, all historical events that caused the failure of that mid-level node are retrieved. The occurrence count of each bottom-level node under that mid-level node is counted, and the total occurrence count of all bottom-level nodes under that mid-level node is calculated. The occurrence count of each bottom-level node is divided by the total occurrence count to obtain the mid-level trigger probability of that bottom-level node relative to that mid-level node.

[0055] For example, statistics were performed on a historical database of 10 fault events built by S200 and 4 intermediate nodes extracted by S300. Intermediate node 1 experienced 4 events: stator current abnormality 4 times, output voltage abnormality 3 times, and protection action signal abnormality 2 times, for a total of 9 occurrences. The calculated trigger probabilities for these events were 0.444, 0.333, and 0.222, respectively. Intermediate node 2 experienced 3 events: bearing oil pressure abnormality 3 times and rotor speed abnormality 2 times, for a total of 5 occurrences, with probabilities of 0.6 and 0.4, respectively. Intermediate node 3 experienced 2 events: machine room temperature abnormality 2 times and machine room humidity abnormality 1 time, for a total of 3 occurrences, with probabilities of 0.667 and 0.333, respectively. Intermediate node 4 experienced 1 event: switch open / closed status abnormality 1 time and protection action signal abnormality 1 time, for a total of 2 occurrences, with a probability of 0.5 for both. Among them, the abnormal protection action signal appeared in both middle layer node 1 and middle layer node 4, with corresponding middle layer trigger probabilities of 0.222 and 0.5, respectively. Based on different statistical benchmarks, they cannot be directly compared across middle layers.

[0056] Secondly, the mid-level trigger probability is normalized and used as a parameter for the node fault contribution of the bottom-level nodes. Normalization refers to mapping the mid-level trigger probabilities calculated based on different statistical benchmarks to a unified [0,1] numerical range through mathematical transformation, eliminating benchmark differences, making the contributions of different mid-levels comparable, and facilitating subsequent fusion analysis with the correlation parameters obtained in step S100. This embodiment uses the Min-Max normalization method.

[0057] Specifically, collect the mid-level trigger probability values ​​of all bottom-level nodes for all mid-level nodes, and find the minimum and maximum values ​​in the set. Apply the Min-Max normalization formula to each mid-level trigger probability to obtain the normalized node fault contribution parameter, with a value range of [0,1].

[0058] For example, all intermediate-level trigger probability values ​​are collected, with a minimum of 0.222, a maximum of 0.667, and a range of 0.445. After Min-Max normalization calculation, the contribution parameter of stator current anomaly to intermediate-level node 1 is 0.499, output voltage anomaly is 0.249, protection action signal anomaly to intermediate-level node 1 is 0, while its contribution to intermediate-level node 4 is 0.625; bearing oil pressure anomaly to intermediate-level node 2 is 0.849, rotor speed anomaly is 0.400; machine room temperature anomaly to intermediate-level node 3 is 1, machine room humidity anomaly is 0.249; switch open / closed status anomaly to intermediate-level node 4 is 0.625. The normalized parameters are all located in the [0,1] interval, which makes the weights of the bottom nodes under different middle-level nodes comparable. For example, the contribution of abnormal protection action signals to electrical system faults is 0, while the contribution to control protection faults is 0.625, which is consistent with the difference in their actual roles in the two types of faults.

[0059] Next, the number of times each device fault type corresponding to a mid-level node triggers a global fault at the top-level node is statistically analyzed from the historical fault event database. The ratio of this occurrence to the total number of faults triggered by the top-level node is calculated to obtain the top-level trigger probability. The top-level trigger probability refers to the relative frequency of a mid-level node in all historical fault events corresponding to its top-level node, reflecting the degree of local contribution of the mid-level node to the fault of the top-level node.

[0060] Specifically, the historical fault event database is traversed. For each top-level node, all historical events that caused that top-level node to occur are retrieved. The occurrence count of each mid-level node is counted, and the total occurrence count of all mid-level nodes is calculated. The occurrence count of each mid-level node is divided by the total occurrence count to obtain the top-level trigger probability of that mid-level node relative to that top-level node.

[0061] For example, based on the top-level node extracted from S300, all 10 fault events ultimately led to unit shutdown. The occurrence frequency of each mid-level node in the events was counted: mid-level node 1 appeared 4 times, mid-level node 2 appeared 3 times, mid-level node 3 appeared 2 times, and mid-level node 4 appeared 1 time, for a total of 10 occurrences. The calculated top-level trigger probabilities for each mid-level node were 0.4, 0.3, 0.2, and 0.1, respectively.

[0062] Furthermore, the top-level trigger probability is normalized and used as the node fault contribution parameter for the middle-level nodes. When there are multiple top-level nodes, the middle-level nodes may correspond to multiple different top-level nodes, and their top-level trigger probabilities are based on different statistical benchmarks, requiring normalization to eliminate benchmark differences. If there is only one top-level node, the top-level trigger probability itself is the normalized contribution parameter. If there are multiple top-level nodes, Min-Max normalization is performed using the method described above. If there is only one top-level node, the top-level trigger probability is directly used as the contribution parameter for the middle-level nodes.

[0063] For example, this embodiment has only one top-level node. Therefore, the node failure contribution parameter of the middle-level node is the top-level trigger probability calculated above. That is, the contribution of middle-level node 1 to middle-level node 4 is 0.4, 0.3, 0.2 and 0.1 respectively.

[0064] Finally, the node fault contribution parameters of each bottom-level and mid-level node are used as the weight values ​​for their respective nodes to complete the weight allocation for the bottom-level and mid-level nodes. A weight value is a quantified numerical value assigned to each node in the fault tree, representing the importance of that node in the faults of its upper-level nodes. The bottom-level node contribution parameters are assigned to the corresponding bottom-level nodes; for different mid-level nodes, the same bottom-level node may have multiple weight values. The mid-level node contribution parameters are then assigned to the corresponding mid-level nodes.

[0065] For example, the normalized contribution parameters are assigned to the corresponding nodes. At the bottom-level nodes, for instance, abnormal protection action signals have two weights: 0 for middle-level node 1 and 0.625 for middle-level node 4; abnormal stator current has a weight of 0.499 for middle-level node 1; abnormal bearing oil pressure has a weight of 0.849 for middle-level node 2; abnormal machine room temperature has a weight of 1 for middle-level node 3; and abnormal switch open / close status has a weight of 0.625 for middle-level node 4, etc. At the middle-level nodes, the weights of middle-level nodes 1 through 4 to the top-level node are 0.4, 0.3, 0.2, and 0.1, respectively.

[0066] In this embodiment of the invention, by combining the correlation parameters of S100 and the historical fault event database of S200, the triggering probability is calculated and normalized by quantifying the number of triggers from bottom-level nodes to middle-level nodes and from middle-level nodes to top-level nodes, thus achieving precise quantification of the node fault contribution parameter. Using the fault contribution parameter as the node weight value, the weight allocation of bottom-level and middle-level nodes is completed, clarifying the influence weight of different nodes in the fault propagation process. This ensures that the weight allocation highly matches the actual fault occurrence pattern and closely corresponds to the three-level node system of bottom-middle-top constructed by S300. Through weight allocation, a scientific and precise quantitative basis is provided for the subsequent logic gate configuration and fault probability calculation of the FTA fault tree, effectively improving the accuracy of fault propagation probability calculation and thus improving the accuracy of fault prediction. Simultaneously, the quantified weight system replaces subjective human judgment, reducing human error and improving the objectivity and standardization of the method.

[0067] S500: Retrieve node correlation characteristics from the historical fault event database, and configure the logic gates of the FTA fault tree based on the node correlation characteristics to complete the construction of the FTA fault tree.

[0068] In this embodiment of the invention, node correlation characteristics are retrieved from the historical fault event database, and the logic gates of the FTA fault tree are configured based on these characteristics to complete the construction of the FTA fault tree. The logic gates of the FTA fault tree determine how bottom-level nodes combine to trigger middle-level nodes, and how middle-level nodes combine to trigger top-level nodes, which is crucial to the logical structure of the fault tree. The triggering relationships of hydroelectric equipment faults are complex, including situations where a single parameter anomaly independently triggers a fault, and situations where multiple parameter anomalies work together to cause a fault. Furthermore, the correlation between parameters reveals that some parameters, although not directly appearing in the fault records, are highly correlated with the fault-triggered parameters and may exist as indirect causes. Therefore, it is necessary to comprehensively obtain node correlation characteristics based on the actual triggering relationships between nodes in the historical fault event database, combined with parameter correlation analysis, and configure scientific logic gates accordingly to ensure that the logical structure of the fault tree accurately reflects the actual fault propagation patterns.

[0069] Step S500 in the method provided in this embodiment of the invention includes: The direct fault triggering relationships between the bottom-level nodes, middle-level nodes, and top-level nodes are retrieved from the historical fault event database and used as basic node correlation characteristics. These basic node correlation characteristics include correlation forms such as a single node directly triggering a superior node and multiple nodes jointly and directly triggering a superior node. Based on the aforementioned correlation parameters, other underlying nodes whose correlation values ​​with the underlying nodes that directly trigger the upper-level nodes meet the preset correlation requirements are selected as indirectly related underlying nodes. The correlation characteristics of the basic nodes and the correlation relationships of the indirectly related underlying nodes are integrated to obtain the node correlation characteristics. If the node association characteristic is that a single directly triggered node or any indirectly associated node can independently trigger the parent node, then an OR gate logic is configured between all corresponding lower-level nodes and the parent node. If the node association characteristic requires multiple directly triggered nodes to satisfy simultaneously, or if a directly triggered node and at least one indirectly associated node must satisfy together to trigger the upper-level node, then an AND gate logic is configured between all corresponding lower-level nodes and the upper-level node. Based on the configured logic gates, weighted bottom-level nodes, middle-level nodes, and top-level nodes, the FTA fault tree is constructed.

[0070] Wherein, the middle-layer node is the parent node of the bottom-layer node, and the top-layer node is the parent node of the middle-layer node.

[0071] First, the direct fault triggering relationships between bottom-level nodes, middle-level nodes, and top-level nodes are retrieved from the historical fault event database and used as basic node correlation characteristics. These basic node correlation characteristics include correlation forms where a single node directly triggers a higher-level node and multiple nodes jointly trigger a higher-level node. Basic node correlation characteristics refer to the actual correlation relationships between lower-level nodes and higher-level nodes that are directly extracted from the historical fault event database and are recorded. They reflect the actual fault propagation paths that occurred historically, including two basic forms: independent triggering by a single node and joint triggering by multiple nodes.

[0072] Specifically, each fault event in the historical fault event database is traversed, and the correspondence between the root cause of the fault and the type of equipment-level fault it causes, as well as the correspondence between the type of equipment-level fault and the final global fault result, is analyzed. For each parent node, all historical records of its triggering by its child nodes are counted, and the set of child nodes that directly trigger the parent node and their triggering forms are summarized. The above direct triggering relationships are recorded as basic node correlation characteristics.

[0073] For example, the direct triggering relationships between nodes are analyzed based on a historical database of 10 fault events. For instance, among the 3 events of mid-level node 2, there are cases where abnormal bearing oil pressure is triggered alone, and cases where abnormal bearing oil pressure and abnormal rotor speed are triggered together; while the 4 events of mid-level node 1 are all triggered jointly by multiple bottom-level nodes, with no record of a single node triggering them independently; and the 10 events of the top-level node are all triggered independently by a single mid-level node, with no cases of multiple mid-level nodes triggering them together. This yields the basic correlation characteristics of each node.

[0074] Secondly, based on the aforementioned correlation parameters, other bottom-level nodes whose correlation values ​​with the directly triggering upper-level nodes meet preset correlation requirements are selected as indirectly related bottom-level nodes. The correlation characteristics of the basic nodes and the correlation relationships of the indirectly related bottom-level nodes are integrated to obtain the node correlation characteristics. Indirectly related bottom-level nodes refer to nodes that, although not directly recorded as the cause of triggering a certain upper-level node in the historical fault event database, are found through the correlation parameter analysis in step S100 to have a strong correlation with the directly triggering node and may act as potential synergistic factors influencing the occurrence of the fault. Integrating indirectly related nodes can compensate for the limited historical samples, making the fault tree logic more complete.

[0075] Specifically, for each parent node, all its directly triggering lower-level nodes are obtained. For each directly triggering lower-level node, other lower-level nodes with a correlation value greater than a preset correlation threshold are retrieved from the correlation parameter matrix obtained in step S100 and added to the set of lower-level nodes of the parent node as indirectly related lower-level nodes. The directly triggering nodes and indirectly related nodes are integrated to form the complete node correlation characteristics of the parent node.

[0076] For example, a preset correlation threshold of 0.7 is used for filtering based on the correlation parameter matrix of S100. For instance, the correlation values ​​between abnormal protection action signals and abnormal stator current, abnormal output voltage, and abnormal switch opening / closing status are 0.88, 0.88, and 0.88 respectively, all exceeding the threshold. Therefore, although the abnormal switch opening / closing status does not directly appear in the historical events of intermediate node 1, it is added to the lower-level node set of intermediate node 1 as an indirectly correlated node because of its strong correlation with abnormal protection action signals. Similarly, abnormal stator current and abnormal output voltage, due to their strong correlation with abnormal protection action signals, are also added to the lower-level node set of intermediate node 4 as indirectly correlated nodes because of their strong correlation with abnormal protection action signals. The complete correlation characteristics of each intermediate node are thus determined after integration.

[0077] Furthermore, if the node association characteristic is that a single directly triggered node or any indirectly associated node can independently trigger the parent node, then OR gate logic is configured between all corresponding lower-level nodes and the parent node. OR gate logic means that the parent node will occur as long as at least one of the lower-level nodes occurs. This is suitable for scenarios where a fault can be independently caused by any one of multiple reasons.

[0078] Specifically, for each parent node, its complete node association characteristics are analyzed. If there is a record in the historical events where a single node independently triggers the parent node, and the indirect nodes supplemented through association analysis also meet any of the required characteristics, then an OR gate is configured between the parent node and all its child nodes.

[0079] For example, because there is a history of abnormal bearing oil pressure triggering a fault on its own, and there is no requirement for its subordinate nodes to meet the same condition, an OR gate is configured for its subordinate nodes; because there is a history of abnormal computer room temperature triggering a fault on its own, an OR gate is also configured for its subordinate nodes; because the top-level node is triggered independently by a single middle-level node each time, an OR gate is configured for its four subordinate middle-level nodes.

[0080] Furthermore, if the node association characteristic requires multiple directly triggering nodes to satisfy simultaneously, or requires a direct triggering node to be satisfied in conjunction with at least one indirectly associated node to trigger the parent node, then AND gate logic is configured between all corresponding lower-level nodes and the parent node. The AND gate logic indicates that all lower-level nodes must occur simultaneously for the parent node to occur. This is suitable for scenarios where a fault requires multiple conditions to be met before it can be triggered.

[0081] Specifically, for each parent node, if all trigger records in the historical events are triggered by multiple nodes, and the indirect nodes supplemented through correlation analysis are also considered to be necessary, then an AND gate is configured between the parent node and all its child nodes.

[0082] For example, all historical events of the middle-level node 1 are triggered by multiple lower-level nodes, with no single node triggering independently. Based on its fault mechanism, it is determined that an electrical system fault requires multiple electrical parameter anomalies to be simultaneously characterized. Therefore, an AND gate is configured for its lower-level nodes. The only historical event of the middle-level node 4 is triggered by both the abnormal switch opening / closing status and the abnormal protection action signal. Furthermore, a control and protection fault requires both to occur simultaneously. Therefore, an AND gate is configured for its directly triggered nodes, while indirectly related nodes are not included in the logic gate.

[0083] Finally, based on the configured logic gates and the weighted bottom, middle, and top nodes, the FTA fault tree is constructed. The FTA fault tree refers to connecting the defined top, middle, and bottom nodes according to the configured logic gates and adding the node weights assigned by S400 to form a complete and quantifiable fault tree model.

[0084] Specifically, the top-level node is used as the root of the tree, and four intermediate nodes are connected to it via OR gates, with each intermediate node's weight relative to the top-level node labeled. Under each intermediate node, its corresponding bottom-level node is connected according to a defined logic gate, with each bottom-level node's weight relative to that intermediate node also labeled. For intermediate nodes with indirectly related nodes, their roles are reflected in the logic gates. This ultimately forms a complete fault tree for subsequent fault prediction.

[0085] For example, the completed FTA fault tree is as follows: The top-level node is connected to an OR gate, with inputs from middle-level nodes 1 to 4, with weights of 0.4, 0.3, 0.2, and 0.1 respectively; Middle-level node 1 is connected to an AND gate, with inputs for stator current anomaly (0.499), output voltage anomaly (0.249), protection action signal anomaly (0), and switch open / closed state anomaly (0.625); Middle-level node 2 is connected to an OR gate, with inputs for bearing oil pressure anomaly (0.849) and rotor speed anomaly (0.400); Middle-level node 3 is connected to an OR gate, with inputs for computer room temperature anomaly (1.000) and computer room humidity anomaly (0.249); Middle-level node 4 is connected to an AND gate, with inputs for switch open / closed state anomaly (0.625) and protection action signal anomaly (0.625). Thus, based on historical data and correlation analysis, the logic gate configuration and overall construction of the FTA fault tree are completed.

[0086] In this embodiment of the invention, based on the direct triggering relationships between nodes in the historical fault event database and combined with the parameter correlation features obtained in step S100, potential indirectly related nodes are mined, forming a more complete node correlation feature. Furthermore, according to different triggering forms, OR and AND gate logic is configured for the fault tree, enabling the fault tree to accurately reflect various modes of actual fault propagation. The logic gate configuration is based on real historical events, avoiding the subjectivity of relying entirely on expert experience in traditional methods; by introducing correlation parameters to filter indirectly related nodes, the limitations of historical samples are compensated for, making the logical relationships of the fault tree more complete and capable of covering potential collaborative fault factors; the clear logic gate structure provides a clear logical foundation for the subsequent quantitative calculation of fault probabilities, ensuring that the fault prediction results conform to the actual operating rules of the equipment.

[0087] S600: Input the current status parameters of the hydropower equipment into the FTA fault tree to obtain the predicted fault type and predicted fault probability.

[0088] In this embodiment of the invention, the current status parameters of the hydropower equipment are input into the FTA fault tree to obtain the predicted fault type and predicted fault probability. The ultimate goal of the FTA fault tree is to predict potential equipment failures using real-time monitoring data. This step compares the current equipment status parameters with the bottom-level nodes of the constructed fault tree to determine the operating status of each bottom-level node. Then, based on the logic gate rules configured by S500 and the node weights allocated by S400, the probability of occurrence of middle-level and top-level nodes is derived layer by layer upwards, finally outputting a quantified prediction result. This process combines real-time data with historical patterns, achieving a precise mapping from parameter anomalies to fault types, providing clear early warning information for maintenance personnel.

[0089] Step S600 in the method provided in this embodiment of the invention includes: Obtain the current status parameters of the hydropower equipment; The status parameters of the hydropower equipment are compared with the fault status parameters of the bottom-level nodes of the FTA fault tree to determine whether the operating status of the bottom-level nodes is normal or abnormal. The node fault contribution parameters corresponding to the bottom-level nodes in abnormal states are summed and calculated. Combined with the association rules of logic gates, the intermediate-level predicted fault probability of the corresponding intermediate-level nodes is derived. The node fault contribution parameters corresponding to the middle-level nodes whose trigger probability exceeds the preset association threshold are summed and calculated. Combined with the logic gate association rules, the global fault result occurrence probability of the corresponding top-level node is derived as the top-level predicted fault probability. By integrating the predicted device-level fault types, global fault results, mid-level predicted fault probabilities, and top-level predicted fault probabilities, the predicted fault types and predicted fault probabilities are obtained.

[0090] First, obtain the current status parameters of the hydropower equipment. The current status parameters of the hydropower equipment refer to the multi-dimensional data reflecting the current operating status of the equipment collected from the real-time monitoring system of hydropower equipment, including electrical parameters, mechanical parameters, operating condition parameters and environmental parameters, which are consistent with the types of historical parameters collected in S100.

[0091] Specifically, real-time equipment operation data is acquired through an online monitoring system or sensor network for the hydropower equipment. The data acquisition range should cover all monitoring points corresponding to the fault status parameters extracted from S200 to ensure comparison with the underlying nodes. After acquisition, necessary data cleaning is performed to remove acquisition noise.

[0092] For example, taking a hydroelectric generator unit in a hydropower station as an example, at a certain time t, the following current status parameters are obtained through the monitoring system: stator current is 1150A, output voltage is 10.8kV, bearing oil pressure is 0.25MPa, rotor speed is 3080r / min, generator room temperature is 36℃, generator room humidity is 58%, switch status is closed, and protection action signal is not activated. Among these, the bearing oil pressure is below the lower limit of normal, the rotor speed is above the upper limit of normal, the generator room temperature is above the upper limit of normal, and the remaining parameters are all within the normal range. Other relevant parameters are also obtained to form the current status parameter set.

[0093] Next, the status parameters of the hydropower equipment are compared with the fault status parameters of the bottom-level nodes of the FTA fault tree to determine whether the operating status of the bottom-level nodes is normal or abnormal. The operating status of a bottom-level node refers to whether the status parameters of the equipment corresponding to that bottom-level node are currently in an abnormal range. If the parameter value exceeds the normal range, such as the threshold defined in S200 or the normal fluctuation range of historical data in S100, then the status of the bottom-level node is determined to be abnormal; otherwise, it is normal.

[0094] Specifically, for each underlying node, the definition of its corresponding fault status parameter is extracted. The currently acquired parameter value is compared with the normal range: if it exceeds the normal range, the underlying node is marked as abnormal; otherwise, it is normal. For switch parameters, the determination is based on whether its logical value is consistent with the normal state.

[0095] For example, after comparing the current parameters with the underlying nodes one by one, the result is that the three underlying nodes with abnormal bearing oil pressure, abnormal rotor speed, and abnormal machine room temperature are abnormal, while the rest of the underlying nodes are normal.

[0096] Next, the node fault contribution parameters corresponding to the bottom-level nodes in the abnormal state are summed and calculated. Combined with the association rules of the logic gates, the predicted fault probability of the corresponding middle-level node is derived. The predicted fault probability of the middle-level node refers to the probability value that the middle-level node may occur, calculated based on the current abnormal bottom-level node and its contribution to the middle-level node, combined with the logic gate rules under that middle-level node. For an OR gate, the probability is the sum of the contributions of each abnormal bottom-level node; for an AND gate, the probability is the product of the contributions of each abnormal bottom-level node.

[0097] Specifically, each intermediate node is traversed to obtain all its underlying nodes and the logic gate types configured for that node. Based on the determined set of abnormal underlying nodes, the abnormal nodes related to that intermediate node and their corresponding contribution parameters are extracted. The failure probability of that intermediate node is calculated according to the logic gate rules: if it is an OR gate, the probability is the sum of the contributions of all related abnormal nodes, but not exceeding 1; if it is an AND gate, the probability is the product of the contributions of each node only when all necessary underlying nodes under that intermediate node are abnormal, otherwise the probability is 0.

[0098] For example, according to the logic gates configured in S500, the middle-level node 2 is configured as an OR gate, and its related abnormal nodes are abnormal bearing oil pressure and abnormal rotor speed, with contribution values ​​of 0.849 and 0.400 respectively. The calculated failure probability of the middle-level node 2 is the sum of the two. The middle-level node 3 is configured as an OR gate, and its related abnormal node is abnormal computer room temperature, with a contribution value of 1.000. The calculated failure probability of the middle-level node 3 is 1.0. The middle-level node 1 is configured as an AND gate, and its required nodes include abnormal stator current, abnormal output voltage, abnormal protection action signal, and abnormal switch opening / closing status. However, all nodes are currently normal, so the failure probability of the middle-level node 1 is 0. The middle-level node 4 is configured as an AND gate, and its required nodes are abnormal switch opening / closing status and abnormal protection action signal. Currently, both are normal, so the failure probability of the middle-level node 4 is 0.

[0099] Furthermore, the node fault contribution parameters corresponding to the mid-level nodes whose trigger probabilities exceed the preset association threshold are summed. Combined with the logic gate association rules, the global fault occurrence probability of the corresponding top-level node is derived, serving as the top-level predicted fault probability. The preset association threshold is a threshold used to select which mid-level nodes participate in the top-level fault probability calculation; it is typically set to 0.5 or adjusted according to actual needs. Only mid-level nodes whose predicted fault probabilities exceed this threshold are considered candidate nodes that may trigger a top-level fault. The top-level predicted fault probability is the probability that a top-level node may occur, calculated based on the contribution of the candidate mid-level nodes and their logic gate rules.

[0100] Specifically, a mid-level failure probability threshold is set. Mid-level nodes with a predicted failure probability greater than this threshold are selected. For top-level nodes, whose logic gates are OR gates, the predicted failure probability of the top-level node is the sum of the contribution parameters of the mid-level nodes. The contribution parameters are the weights of the S400 mid-level nodes to the top-level node.

[0101] For example, a preset association threshold is set to 0.5. The failure probabilities of middle-layer nodes 2 and 3 both exceed 0.5, while those of middle-layer nodes 1 and 4 are both below the threshold. The weights of the middle-layer nodes to the top-layer node are: 0.3 for middle-layer node 2 and 0.2 for middle-layer node 3. The top-layer node is configured as an OR gate, therefore the calculated failure probability of the top-layer node is the sum of 0.3 and 0.2, which is 0.5.

[0102] Finally, the predicted device-level fault types, global fault results, mid-level predicted fault probabilities, and top-level predicted fault probabilities are integrated to obtain the predicted fault types and predicted fault probabilities. Predicted fault types refer to the predicted device-level and global faults based on the current state, typically listing the fault types with higher probabilities. Predicted fault probabilities include the probabilities of each mid-level node and the top-level node.

[0103] Specifically, the above calculation results are summarized to form a prediction output, which includes: the name and predicted probability of each intermediate node, the name and predicted probability of the top-level node, and information on the main abnormal bottom-level nodes causing the failure of each intermediate node. The predicted probabilities are divided into different risk levels, for example: a predicted probability > 0.8 is high risk, and immediate inspection is recommended; 0.5 ≤ predicted probability < 0.8 is medium risk, and enhanced monitoring is recommended; a predicted probability < 0.5 is low risk, and normal operation is recommended.

[0104] For example, the integrated prediction results are as follows: Regarding equipment-level fault types, the probability of mechanical transmission faults is 1.0, reaching the output upper limit. This indicates that the current combination of abnormal bearing oil pressure and abnormal rotor speed highly matches historical fault patterns. According to the preset grading standard, a predicted probability ≥ 0.8 is classified as a high-risk level, requiring immediate shutdown for inspection. The predicted probability of environmental adaptation faults is 1.0, caused by abnormal machine room temperature, also reaching a high-risk level, requiring a thorough investigation into the cause of the temperature anomaly. Regarding global fault results, the predicted probability of hydro-generator unit shutdown is 0.5. According to the grading standard, a predicted probability between 0.5 and 0.8 is classified as a medium-risk level, indicating the need to strengthen monitoring of equipment operation trends and prepare for emergencies, but immediate shutdown is not required at this time. Based on this, clear early warning information and corresponding graded handling suggestions can be provided to maintenance personnel.

[0105] In this embodiment of the invention, real-time monitoring data is deeply integrated with the constructed FTA fault tree, enabling quantitative prediction of fault probability from parameter anomalies. By comparing the status of the bottom-level nodes in real time, current parameter anomalies can be detected in a timely manner, and their contribution to mid-level and top-level faults can be objectively evaluated using historical weights and logic gate rules, avoiding the isolation of traditional threshold alarms. Multi-level probability calculation clearly shows the fault propagation path, improving the convenience of fault tracing. The final output of predicted fault types and probabilities provides a quantitative basis for equipment condition-based maintenance and operation decisions, helping to take early intervention measures, reduce the risk of unplanned downtime, and improve the reliability and safety of hydropower equipment operation.

[0106] Through the specific implementation methods described above, the embodiments of the present invention achieve the following technical effects: This invention provides a fault prediction method and system based on FTA (Fault Tree Analysis). By analyzing the correlation of historical equipment state parameters, the inherent coupling relationships between parameters are extracted. Fault state parameters are extracted as bottom-level nodes based on a historical fault event database, and equipment-level fault types and global fault results are summarized to construct a complete fault tree hierarchical structure. In node weight allocation, historical trigger frequencies are statistically analyzed and normalized to eliminate dimensional differences caused by different statistical benchmarks, making weight allocation more objective. The logic gate configuration integrates historical direct trigger relationships and parameter correlation characteristics, making the fault propagation logic more closely aligned with actual mechanisms. After inputting real-time state parameters into the fault tree, the quantified probabilities of equipment-level and global faults can be calculated layer by layer, achieving a precise mapping from parameter anomalies to fault results. This invention effectively avoids the limitations of traditional fault tree construction relying on subjective experience, improves the accuracy and interpretability of fault prediction, and provides reliable data support and decision-making basis for intelligent operation and maintenance of hydropower equipment.

[0107] Example 2, as Figure 2As shown, this invention provides a fault prediction system based on FTA fault trees, the system comprising: The parameter correlation analysis module 11 is used to obtain historical device status parameters and perform parameter correlation analysis on the historical device status parameters to obtain correlation parameters. The bottom node extraction module 12 is used to obtain a historical fault event database and extract the fault status parameters corresponding to the fault events as the bottom nodes of the FTA fault tree. The middle and top-level node extraction module 13 is used to extract equipment fault types from the historical fault event database as middle-level nodes of the FTA fault tree, and extract global fault results as top-level nodes of the FTA fault tree. The node weight allocation module 14 is used to obtain the node fault contribution parameter based on the correlation parameter and the historical fault event database, and to allocate weights to the bottom-level node and the middle-level node. The logic gate configuration module 15 is used to retrieve node correlation characteristics from the historical fault event database, and configure the logic gates of the FTA fault tree based on the node correlation characteristics to complete the construction of the FTA fault tree. The fault prediction output module 16 is used to input the current status parameters of the hydropower equipment into the FTA fault tree to obtain the predicted fault type and the predicted fault probability.

[0108] In one embodiment, the parameter correlation analysis module 11 is further configured to: Among them, obtaining historical device status parameters includes: The electrical parameters, mechanical parameters, operating condition parameters, and environmental parameters of the hydropower equipment are collected as historical equipment status parameters. The electrical parameters include at least current and voltage, the mechanical parameters include at least speed and pressure, the operating condition parameters include at least switch open / closed status and protection action signals, and the environmental parameters include at least temperature and humidity.

[0109] Specifically, a parameter correlation analysis is performed on the historical equipment status parameters to obtain correlation parameters, including: The historical equipment status parameters are standardized to obtain standard hydroelectric equipment parameters; Calculate the Pearson correlation coefficient between any two standardized parameters as the correlation value; By integrating the correlation values ​​among all parameters, a correlation parameter is formed.

[0110] In one embodiment, the underlying node extraction module 12 is further configured to: Collect fault occurrence records, fault investigation records, and fault handling records throughout the entire life cycle of hydropower equipment, and integrate them to form a historical fault event database. The historical fault event database includes information on the root cause of the fault, the status of the corresponding equipment parameters, the time of the fault occurrence, and the scope of the fault's impact. Extract the abnormal values ​​of the hydropower equipment status parameters corresponding to each fault root cause from the historical fault event database, and take the union of each fault root cause as the fault status parameter. Each fault state parameter is used as an independent FTA fault tree bottom node, and each bottom node uniquely corresponds to the parameter representation of the root cause of a type of hydropower equipment fault.

[0111] In one embodiment, the middle and top layer node extraction module 13 is further configured to: Extract equipment-level hydropower equipment failure types caused by one or more root causes from the historical failure event database, and use them as independent FTA fault tree mid-level nodes. Extract the global failure results of hydropower equipment caused by one or more equipment-level failures from the historical failure event database, and use them as the top-level node of the FTA fault tree.

[0112] In one embodiment, the node weight allocation module 14 is further configured to: The number of times a fault state parameter corresponding to each bottom node triggers a fault type in the middle node device is counted from the historical fault event database. The ratio of this number of occurrences to the total number of occurrences in the corresponding middle node is calculated to obtain the middle-level trigger probability. The mid-layer trigger probability is normalized and used as the node fault contribution parameter of the bottom-layer node. The number of times a device fault type corresponding to each mid-level node causes a global fault result in the top-level node is counted from the historical fault event database. The ratio of this number of occurrences to the total number of occurrences in the top-level node is calculated to obtain the top-level trigger probability. The top-level trigger probability is normalized and used as the node fault contribution parameter of the middle-level node. The node fault contribution parameters of each bottom-level node and middle-level node are used as the weight values ​​of the corresponding nodes to complete the weight allocation of the bottom-level nodes and middle-level nodes.

[0113] In one embodiment, the logic gate configuration module 15 is further configured to: The direct fault triggering relationships between the bottom-level nodes, middle-level nodes, and top-level nodes are retrieved from the historical fault event database and used as basic node correlation characteristics. These basic node correlation characteristics include correlation forms such as a single node directly triggering a superior node and multiple nodes jointly and directly triggering a superior node. Based on the aforementioned correlation parameters, other underlying nodes whose correlation values ​​with the underlying nodes that directly trigger the upper-level nodes meet the preset correlation requirements are selected as indirectly related underlying nodes. The correlation characteristics of the basic nodes and the correlation relationships of the indirectly related underlying nodes are integrated to obtain the node correlation characteristics. If the node association characteristic is that a single directly triggered node or any indirectly associated node can independently trigger the parent node, then an OR gate logic is configured between all corresponding lower-level nodes and the parent node. If the node association characteristic requires multiple directly triggered nodes to satisfy simultaneously, or if a directly triggered node and at least one indirectly associated node must satisfy together to trigger the upper-level node, then an AND gate logic is configured between all corresponding lower-level nodes and the upper-level node. Based on the configured logic gates, weighted bottom-level nodes, middle-level nodes, and top-level nodes, the FTA fault tree is constructed.

[0114] Wherein, the middle-layer node is the parent node of the bottom-layer node, and the top-layer node is the parent node of the middle-layer node.

[0115] In one embodiment, the fault prediction output module 16 is further configured to: Obtain the current status parameters of the hydropower equipment; The status parameters of the hydropower equipment are compared with the fault status parameters of the bottom-level nodes of the FTA fault tree to determine whether the operating status of the bottom-level nodes is normal or abnormal. The node fault contribution parameters corresponding to the bottom-level nodes in abnormal states are summed and calculated. Combined with the association rules of logic gates, the intermediate-level predicted fault probability of the corresponding intermediate-level nodes is derived. The node fault contribution parameters corresponding to the middle-level nodes whose trigger probability exceeds the preset association threshold are summed and calculated. Combined with the logic gate association rules, the global fault result occurrence probability of the corresponding top-level node is derived as the top-level predicted fault probability. By integrating the predicted device-level fault types, global fault results, mid-level predicted fault probabilities, and top-level predicted fault probabilities, the predicted fault types and predicted fault probabilities are obtained.

[0116] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A fault prediction method based on FTA fault tree, characterized in that, include: Historical device status parameters are obtained, and parameter correlation analysis is performed on the historical device status parameters to obtain correlation parameters; Obtain the historical fault event database and extract the fault status parameters corresponding to the fault events as the bottom-level nodes of the FTA fault tree; The equipment fault types are extracted from the historical fault event database and used as the middle-level nodes of the FTA fault tree, and the global fault results are extracted and used as the top-level nodes of the FTA fault tree. Based on the correlation parameters and the historical fault event database, the node fault contribution parameters are obtained, and weights are assigned to the bottom-level nodes and the middle-level nodes. Node correlation characteristics are retrieved from the historical fault event database, and the logic gates of the FTA fault tree are configured based on the node correlation characteristics to complete the construction of the FTA fault tree. Input the current status parameters of the hydropower equipment into the FTA fault tree to obtain the predicted fault type and predicted fault probability.

2. The fault prediction method based on FTA fault tree according to claim 1, characterized in that, Obtain historical device status parameters, including: The electrical parameters, mechanical parameters, operating condition parameters, and environmental parameters of the hydropower equipment are collected as historical equipment status parameters. The electrical parameters include at least current and voltage, the mechanical parameters include at least speed and pressure, the operating condition parameters include at least switch open / closed status and protection action signals, and the environmental parameters include at least temperature and humidity.

3. The fault prediction method based on FTA fault tree according to claim 1, characterized in that, Perform parameter correlation analysis on the historical equipment status parameters to obtain correlation parameters, including: The historical equipment status parameters are standardized to obtain standard hydroelectric equipment parameters; Calculate the Pearson correlation coefficient between any two standardized parameters as the correlation value; By integrating the correlation values ​​among all parameters, a correlation parameter is formed.

4. The fault prediction method based on FTA fault tree according to claim 1, characterized in that, Obtain the historical fault event database and extract the fault status parameters corresponding to the fault events as the bottom-level nodes of the FTA fault tree, including: Collect fault occurrence records, fault investigation records, and fault handling records throughout the entire life cycle of hydropower equipment, and integrate them to form a historical fault event database. The historical fault event database includes information on the root cause of the fault, the status of the corresponding equipment parameters, the time of the fault occurrence, and the scope of the fault's impact. Extract the abnormal values ​​of the hydropower equipment status parameters corresponding to each fault root cause from the historical fault event database, and take the union of each fault root cause as the fault status parameter. Each fault state parameter is used as an independent FTA fault tree bottom node, and each bottom node uniquely corresponds to the parameter representation of the root cause of a type of hydropower equipment fault.

5. The fault prediction method based on FTA fault tree according to claim 1, characterized in that, Equipment fault types are extracted from the historical fault event database and used as mid-level nodes in the FTA fault tree. Global fault results are also extracted and used as top-level nodes in the FTA fault tree, including: Extract equipment-level hydropower equipment failure types caused by one or more root causes from the historical failure event database, and use them as independent FTA fault tree mid-level nodes. Extract the global failure results of hydropower equipment caused by one or more equipment-level failures from the historical failure event database, and use them as the top-level node of the FTA fault tree.

6. The fault prediction method based on FTA fault tree according to claim 1, characterized in that, Based on the aforementioned correlation parameters and the historical fault event database, node fault contribution parameters are obtained, and weights are assigned to the bottom-level nodes and the middle-level nodes, including: The number of times a fault state parameter corresponding to each bottom node triggers a fault type in the middle node device is counted from the historical fault event database. The ratio of this number of occurrences to the total number of occurrences in the corresponding middle node is calculated to obtain the middle-level trigger probability. The mid-layer trigger probability is normalized and used as the node fault contribution parameter of the bottom-layer node. The number of times a device fault type corresponding to each mid-level node causes a global fault result in the top-level node is counted from the historical fault event database. The ratio of this number of occurrences to the total number of occurrences in the top-level node is calculated to obtain the top-level trigger probability. The top-level trigger probability is normalized and used as the node fault contribution parameter of the middle-level node. The node fault contribution parameters of each bottom-level node and middle-level node are used as the weight values ​​of the corresponding nodes to complete the weight allocation of the bottom-level nodes and middle-level nodes.

7. The fault prediction method based on FTA fault tree according to claim 1, characterized in that, Node correlation characteristics are retrieved from the historical fault event database, and the logic gates of the FTA fault tree are configured based on the node correlation characteristics to complete the construction of the FTA fault tree, including: The direct fault triggering relationships between the bottom-level nodes, middle-level nodes, and top-level nodes are retrieved from the historical fault event database and used as basic node correlation characteristics. These basic node correlation characteristics include correlation forms such as a single node directly triggering a superior node and multiple nodes jointly and directly triggering a superior node. Based on the aforementioned correlation parameters, other underlying nodes whose correlation values ​​with the underlying nodes that directly trigger the upper-level nodes meet the preset correlation requirements are selected as indirectly related underlying nodes. The correlation characteristics of the basic nodes and the correlation relationships of the indirectly related underlying nodes are integrated to obtain the node correlation characteristics. If the node association characteristic is that a single directly triggered node or any indirectly associated node can independently trigger the upper-level node, then an OR gate logic is configured between all corresponding lower-level nodes and the upper-level node. If the node association characteristic requires multiple directly triggered nodes to satisfy simultaneously, or if a directly triggered node and at least one indirectly associated node must satisfy together to trigger the upper-level node, then an AND gate logic is configured between all corresponding lower-level nodes and the upper-level node. Based on the configured logic gates, weighted bottom-level nodes, middle-level nodes, and top-level nodes, the FTA fault tree is constructed.

8. The fault prediction method based on FTA fault tree according to claim 7, characterized in that, The middle-level node is the parent node of the bottom-level node, and the top-level node is the parent node of the middle-level node.

9. The fault prediction method based on FTA fault tree according to claim 1, characterized in that, Input the current status parameters of the hydropower equipment into the FTA fault tree to obtain the predicted fault type and predicted fault probability, including: Obtain the current status parameters of the hydropower equipment; The status parameters of the hydropower equipment are compared with the fault status parameters of the bottom-level nodes of the FTA fault tree to determine whether the operating status of the bottom-level nodes is normal or abnormal. The node fault contribution parameters corresponding to the bottom-level nodes in abnormal states are summed and calculated. Combined with the association rules of logic gates, the intermediate-level predicted fault probability of the corresponding intermediate-level nodes is derived. The node fault contribution parameters corresponding to the middle-level nodes whose trigger probability exceeds the preset association threshold are summed and calculated. Combined with the logic gate association rules, the global fault result occurrence probability of the corresponding top-level node is derived as the top-level predicted fault probability. By integrating the predicted device-level fault types, global fault results, mid-level predicted fault probabilities, and top-level predicted fault probabilities, the predicted fault types and predicted fault probabilities are obtained.

10. A fault prediction system based on FTA fault tree, characterized in that, The fault prediction method based on FTA fault tree as described in any one of claims 1-9 includes: The parameter correlation analysis module is used to obtain historical device status parameters and perform parameter correlation analysis on the historical device status parameters to obtain correlation parameters. The bottom-level node extraction module is used to obtain a historical fault event database and extract the fault status parameters corresponding to the fault events as the bottom-level nodes of the FTA fault tree. The middle and top-level node extraction module is used to extract equipment fault types from the historical fault event database as middle-level nodes of the FTA fault tree, and extract global fault results as top-level nodes of the FTA fault tree. The node weight allocation module is used to obtain the node fault contribution parameter based on the correlation parameter and the historical fault event database, and to allocate weights to the bottom-level nodes and the middle-level nodes. The logic gate configuration module is used to retrieve node correlation characteristics from the historical fault event database, and configure the logic gates of the FTA fault tree based on the node correlation characteristics to complete the construction of the FTA fault tree. The fault prediction output module is used to input the current status parameters of the hydropower equipment into the FTA fault tree to obtain the predicted fault type and the predicted fault probability.

Images