Bayesian inference method for reliability of high voltage power supply under common cause failure effect

Abstract

A bayesian inference method for reliability of high-voltage power supply under common cause failure influence belongs to the technical field of high-voltage power supply reliability analysis. A dynamic fault tree of the high-voltage power supply is constructed; the proportion of common cause failure occurrence in each pair of bottom event failure is obtained, a fuzzy matrix is constructed by collecting evaluation opinions for the proportion of bottom events; the common cause failure factor of each pair of bottom events is determined; the dynamic fault tree of the common cause failure is converted into a discrete-time bayesian network, and the bayesian network graph, the conditional probability table and the marginal probability table are obtained respectively; the bidirectional inference is carried out on the dynamic reliability of the high-voltage power supply, and the reliability and maintainability indexes of the high-voltage power supply in each time period are calculated. The common cause failure factor is determined through the fuzzy theory, and the dynamic fault tree considering the common cause failure is mapped into the bayesian network for reliability analysis, so that the problems of difficult determination of the common cause failure factor and large calculation amount of the reliability analysis are solved, and the precision and calculation efficiency of the high-voltage power supply reliability analysis are effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of high-voltage power supply reliability analysis technology, specifically relating to a Bayesian inference method for high-voltage power supply reliability under the influence of common cause failure. Background Technology

[0002] A high-voltage power supply is a device that converts low voltage into high voltage to drive high-voltage devices. Compared to ordinary power supplies, high-voltage power supplies have higher energy density and conversion efficiency, and are widely used in industrial equipment, communication electronics, aerospace, and other fields. High-voltage power supplies operate in extremely complex environments; their components must withstand harsh conditions such as high temperature, high humidity, and mechanical vibration. This makes them more susceptible to various failures, and these failures often lead to major accidents. Therefore, practical applications typically have specific requirements for the reliability of high-voltage power supplies, making reliability analysis and evaluation essential.

[0003] Dynamic fault tree (FLAP) is a commonly used reliability analysis method. Traditional fault tree analysis often assumes that failures between bottom events are independent of each other. However, due to the complexity of the high-voltage power supply's operating environment, common causes of failure among bottom events often exist, such as environmental stress, abnormal input, design defects, and human error. This makes traditional analysis methods prone to errors when analyzing dynamic fault trees.

[0004] Currently, scholars both domestically and internationally have conducted numerous studies on reliability analysis methods considering common-cause failures. One commonly used method is to decompose the underlying events using common-cause failure factors. However, the selection of common-cause failure factors is often a major challenge in reliability analysis, as the determination of factor values ​​directly affects the final reliability analysis results.

[0005] Meanwhile, high-voltage power supply systems are large and complex, with numerous underlying events and common cause failures. Traditional Markov chain analysis methods are prone to combinatorial explosion when analyzing such large systems, which makes computation difficult. Summary of the Invention

[0006] To overcome the aforementioned problems, this invention provides a Bayesian inference method for the reliability of high-voltage power supplies under the influence of common-cause failures, belonging to the field of high-voltage power supply reliability analysis technology. The method involves constructing a dynamic fault tree for the high-voltage power supply; obtaining the proportion of common-cause failures in each pair of bottom-event failures, collecting evaluation opinions based on the bottom-event proportions, and constructing a fuzzy matrix; determining the common-cause failure factor for each pair of bottom-event common-cause failures; transforming the dynamic fault tree of common-cause failures into a discrete-time Bayesian network, obtaining the Bayesian network graph, conditional probability table, and marginal probability table; and performing bidirectional inference on the dynamic reliability of the high-voltage power supply to calculate the reliability and maintainability indicators of the high-voltage power supply at each time period. This invention determines the common-cause failure factor through fuzzy theory and maps the dynamic fault tree considering common-cause failures to a Bayesian network for reliability analysis, thereby solving the problems of difficulty in determining the common-cause failure factor and excessive computational load in reliability analysis, effectively improving the accuracy and computational efficiency of high-voltage power supply reliability analysis.

[0007] A Bayesian inference method for the reliability of high-voltage power supplies under the influence of common-cause failure, such as Figure 1 As shown, it includes the following steps:

[0008] Step 1: Construct a dynamic fault tree for the high-voltage power supply;

[0009] The high-voltage power supply includes three modules: a high-voltage module, a control module, and an input protection module.

[0010] The dynamic fault tree of the high-voltage power supply includes the dynamic fault tree of the high-voltage module, the dynamic fault tree of the control module, and the dynamic fault tree of the input protection module. The top events of the three module dynamic fault trees are connected by an OR gate to obtain the dynamic fault tree of the high-voltage power supply. The output event of the three module dynamic fault trees connected by an OR gate is the top event of the dynamic fault tree of the high-voltage power supply, representing the final fault of the high-voltage power supply.

[0011] Analyze the bottom events in the dynamic fault tree of the high-voltage power supply that have a common failure cause and list them in pairs; a pair of bottom events with a common failure cause is defined as a pair of common cause failure events;

[0012] Step 2: Construct a comment set based on the proportion of common cause failures and collect qualitative evaluation opinions;

[0013] The set of comments for the qualitative evaluation opinions for:

[0014] ;

[0015] Among them, weak common causes This indicates that common cause failures account for 0.05-0.35% of all failures; weaker common cause failures. This indicates that common cause failures account for 0.25-0.55% of all failures; strong common cause. This indicates that common-cause failures account for 0.45-0.75% of all failures; strong common-cause failures... This indicates that the proportion of common cause failures among all failures is 0.65-0.95; the proportions mentioned in various qualitative evaluation opinions are fuzzy proportions, which will be constructed and defuzzified in step 3;

[0016] For each pair of common cause failure events, five qualitative evaluation opinions were collected for comprehensive assessment based on the proportion of common cause failures.

[0017] Step 3: Determine the common cause failure factor for each pair of bottom event common cause failures;

[0018] Five qualitative evaluation opinions are fuzzy quantized to construct a fuzzy matrix; a decision matrix is ​​constructed based on the indicators of the source person corresponding to the evaluation opinions, and the final evaluation weight of each of the five qualitative evaluation opinions is calculated; the five evaluation opinions are aggregated according to the final evaluation weight; the aggregation result is defuzzified to obtain the common cause failure factor of each pair of basic event common cause failures;

[0019] Step 4: Transform the dynamic fault tree of common cause failure into a discrete-time Bayesian network to obtain the Bayesian network graph, conditional probability table and marginal probability table respectively.

[0020] Step 5: Perform bidirectional reasoning on the dynamic reliability of the high-voltage power supply and calculate the reliability and maintainability indicators of the high-voltage power supply for each time period.

[0021] Furthermore, in step 1, the process of constructing the dynamic fault tree of the high-voltage power supply is as follows:

[0022] Step 1.1: Query the high-voltage power supply structure and collect historical fault records, maintenance reports, and failure mode and effects analysis reports of the system;

[0023] Step 1.2: Based on the query results in Step 1.1, the final failure of each of the high-voltage module, control module, and input protection module is regarded as the top event of each module. Then, the top events of each module are decomposed level by level, and the dependency relationship between the upper-level event and the lower-level event in the decomposition process is described by static gates and dynamic gates according to the failure logic. The decomposition is carried out level by level until it is decomposed into the bottom event that cannot be further divided. The top events of the dynamic fault trees of the three modules are connected by OR gates to obtain the dynamic fault tree of the high-voltage power supply.

[0024] The static gates include AND gates and OR gates; the dynamic gates include priority AND gates (PAND), cold spare parts gates (CSP), sequential dependent gates (SEQ), and function dependent gates (FDEP).

[0025] The indivisible basic events are the smallest functional units or the most basic failure modes that constitute a high-voltage power supply system. There are no more detailed "failure causes," and they are natural basic events. Indivisible basic events include inherent failures of components, external environmental interference, and human error.

[0026] The higher-level events are the output events of static gates and dynamic gates, and the lower-level events are the decomposition events of the output events;

[0027] Step 1.3: Determine the failure distribution function and parameter values ​​for each bottom event:

[0028] Based on the query results in 1.1, use historical fault records to draw a probability paper chart or trend chart for each basic event to determine the distribution type fitted by the data points;

[0029] Then, the maximum likelihood estimation method is used to obtain the parameter values ​​of the distribution function corresponding to each basic event distribution type;

[0030] Step 1.4: Based on the query results of 1.1, identify the common causes of failure in the underlying events, including environmental stress, input anomalies, design defects, and human error. List the underlying events that cause common causes of failure and their causes.

[0031] Furthermore, in step 3, the process for determining the common cause failure factor is as follows:

[0032] Step 3.1: Construct a fuzzy matrix by performing fuzzy quantization on the qualitative evaluation opinions. :

[0033] A trapezoidal membership function is used to fuzzify the linguistic variables of different qualitative evaluation opinions. Each linguistic variable corresponds to a set of trapezoidal fuzzy numbers. The range of common cause failure factors is defined as follows: The blurring process is as follows:

[0034] ;

[0035] The trapezoidal fuzzy numbers of the qualitative evaluation opinions for each pair of common cause failures are combined to obtain the fuzzy matrix for each pair of common cause failures. fuzzy matrix Each element That is, the first The first evaluation A number of trapezoidal fuzzy numbers:

[0036] ;

[0037] Step 3.2: Obtain the final evaluation weights for each of the five qualitative evaluation comments:

[0038] First, based on three indicators—experience years (assigned as years of work experience, higher values ​​indicate richer practical experience), education level (assigned as: PhD = 5, Master's = 4, Bachelor's = 3, Associate's = 2, below Associate's = 1, higher values ​​indicate a more solid theoretical foundation) and work relevance (1-5 points, 5 points representing long-term engagement in target chemical process safety analysis, 1 point representing unrelated field)—a decision matrix is ​​constructed according to the indicator values. Decision matrix Each element That is, the first The first source of the evaluation opinion The scores of each indicator; Defined as the first The first source of the evaluation opinion Projected probabilities of scores for each indicator:

[0039] ;

[0040] Then, obtain the dispersion of the three indicators: years of experience, education level, and job relevance. :

[0041] ;

[0042] Next, based on the dispersion The weights of the three indicators—years of experience, education level, and job relevance—are calculated. :

[0043] ;

[0044] Next, the preliminary rating weights of the five qualitative rating opinions were calculated. :

[0045] ;

[0046] Finally, the weights Normalization was performed to obtain the final weights of the five qualitative comments;

[0047] Step 3.3: Aggregate the five qualitative comments based on their respective final comment weights:

[0048] First, the final evaluation weights are compared with the fuzzy matrices of each pair of common cause failures. A weighted average is performed row by row to obtain the trapezoidal fuzzy number after aggregating the five evaluation opinions;

[0049] Then, based on the trapezoidal fuzzy numbers, the aggregated trapezoidal membership functions for each pair of common cause failures are obtained. :

[0050]

[0051] in, These are the trapezoidal fuzzy numbers obtained from aggregation;

[0052] Step 3.4: Defuzzify the aggregation results to obtain the common cause failure factor:

[0053] Using the central area method Perform deblurring transformation:

[0054] ;

[0055] in, for The independent variable represents the common cause failure factor; This represents the defuzzification result, which is the common cause failure factor for each pair of bottom events.

[0056] Furthermore, in step 4, the process of obtaining the Bayesian network graph, conditional probability table, and marginal probability table is as follows:

[0057] Step 4.1, set the high-voltage power supply operating time. Divided into equal parts Time period:

[0058] The duration of each time period is... , recorded as ;exist Based on the existing time periods, a time period in which no high-voltage power supply failure occurs is introduced, then there are a total of The period before The time period represents the high-voltage power supply at... The fault occurred during the [number] time period. Each time period represents a period where the high-voltage power supply is not faulty. The process for dividing these time periods is as follows:

[0059] ;

[0060] Step 4.2: Map the dynamic fault tree diagram of the high-voltage power supply, which considers common-cause failures, to a Bayesian network diagram:

[0061] First, the top event of the high-voltage power supply is regarded as the leaf node in the Bayesian network graph; the output events of each logic gate are regarded as the middle node in the Bayesian network graph; and the bottom event is regarded as the root node in the Bayesian network graph.

[0062] Then, the root node corresponding to each pair of common cause failure events is decomposed into three nodes in the Bayesian diagram. The three nodes represent one common cause failure event and two independent failure events, respectively.

[0063] Finally, directed line segments are connected between the events to form a Bayesian network graph;

[0064] Step 4.3, obtain the edge probability table of the root node:

[0065] First, based on each bottom event probability density function or distribution function Obtain the root node edge probability table of the bottom event at different time periods;

[0066] Bottom Event In the The probability of failure within a certain time period for:

[0067] ;

[0068] Then, the marginal probabilities of the base events in the common cause failure decomposition are decomposed according to the common cause failure factor obtained in step 3, so as to obtain the final root node marginal probability table of each base event in different time periods after decomposition.

[0069] Step 4.4, obtain the conditional probability table of intermediate nodes:

[0070] Based on the logical relationships of different logic gates, obtain the conditional probability table of the intermediate nodes corresponding to the output events of each logic gate;

[0071] Step 4.4.1: Based on the logic gate, determine whether the sub-event fails in different time periods and output whether the event fails in each time period, thereby obtaining the value of the intermediate node corresponding to the sub-node under various value combinations.

[0072] Step 4.4.2: Obtain the conditional probability table of intermediate nodes corresponding to the input events of the function-related gates. :

[0073] ;

[0074] In the formula, Indicates input event The probability density function; To trigger the event; Indicates the triggering of the event The period of failure; Indicates input event The period of failure;

[0075] Step 4.4.3: Obtain the conditional probability table of the intermediate nodes corresponding to the cold spare parts gate spare parts event. :

[0076] ;

[0077] In the formula, Indicates input event The probability density function; Indicates a spare parts event The probability density function; Indicates input event The period of failure; Indicating spare parts events Expiration period;

[0078] Step 4.4.4: Update the conditional probability table of intermediate nodes corresponding to events containing common cause failure sub-events accordingly;

[0079] When the common cause failure event does not occur, the value of the intermediate node conditional probability table is determined by the child node corresponding to the independent failure event under the intermediate node, and is independent of the child node corresponding to the common cause failure event.

[0080] When a common cause failure event occurs, the intermediate node conditional probability table is determined by referring to the values ​​of the corresponding child nodes when two independent failure events occur simultaneously.

[0081] Furthermore, in step 5, a two-way inference is performed on the dynamic reliability of the high-voltage power supply, specifically as follows:

[0082] Step 5.1: Perform forward reasoning of the probability of upper-level nodes in the Bayesian network from bottom to top using the law of total probability.

[0083] Let the bottom events be in order. Each intermediate event is High-voltage power supply top incident No. The probability of failure within a certain time period for:

[0084] ;

[0085] In the formula, This indicates the failure status of each underlying event at different time periods. This indicates the failure status of each intermediate event at different time periods. Indicates the top event at the 1st Failure scenarios within a time period; n represents the total number of events. Represents the total number of intermediate events;

[0086] Based on the failure probability of the top event in different time periods Calculate the high-voltage power supply at different time periods reliability :

[0087] ;

[0088] In the formula, if the time period is taken as Then, the reliability of the high-voltage power supply over the entire operating time is obtained. :

[0089] ;

[0090] Step 5.2: When the high-voltage power supply top event occurs, reverse reasoning of the occurrence probability of each bottom event node is performed from top to bottom in the Bayesian network using Bayes' theorem.

[0091] When the high-voltage power supply top event In the When a time period expires, calculate any event. At any time period failure probability for:

[0092] ;

[0093] Maintenance personnel use a posterior probability table to determine the probability of occurrence of each underlying event when a high-voltage power supply top event occurs. They then systematically check each underlying event for faults, from highest to lowest probability. Simultaneously, to improve maintenance efficiency, maintenance is conducted in stages: first, the probability of occurrence of the top events of the three modules when a high-voltage power supply top event occurs is determined; then, the faulty module is identified based on the probability of occurrence of the top events of the three modules; finally, the probability of occurrence of each underlying event under each module is determined; and then, each underlying event is systematically checked for faults, from highest to lowest probability.

[0094] An electronic device includes: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to perform the Bayesian inference method described above.

[0095] A computer-readable storage medium storing program code that can be invoked by a processor to execute the Bayesian inference method described above.

[0096] The technical advantages of this invention compared to the prior art are as follows:

[0097] This invention uses fuzzy theory to determine common cause failure factors and maps the dynamic fault tree that takes into account common cause failures into a Bayesian network for reliability analysis, thereby solving the problems of difficulty in determining common cause failure factors and excessive computational load, and effectively improving the accuracy and computational efficiency of high voltage power supply reliability analysis. Attached Figure Description

[0098] Figure 1This is an overall flowchart of the present invention;

[0099] Figure 2 shows the dynamic fault tree of the high-voltage power supply of the present invention;

[0100] Figure 2(a) shows the dynamic fault tree of the high-voltage module; Figure 2(b) shows the dynamic fault tree of the control module; Figure 2(c) shows the dynamic fault tree of the input protection module.

[0101] Figure 3(a) shows the discrete-time Bayesian network of the high-voltage module; Figure 3(b) shows the discrete-time Bayesian network of the control module; Figure 3(c) shows the discrete-time Bayesian network of the input protection module. Detailed Implementation

[0102] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0103] Step 1: The high-voltage power supply consists of three modules: a high-voltage module, a control module, and an input protection module. The final failure of the high-voltage power supply is the top event. Dynamic fault trees are constructed for the high-voltage module, the control module, and the input protection module respectively. The high-voltage power supply is regarded as a complex system composed of three sub-parts: the high-voltage part, the control part, and the input part. With the final failure of the high-voltage power supply as the top event, dynamic fault trees are constructed for the three sub-parts respectively. The constructed dynamic fault trees are shown in Figure 2.

[0104] Based on the bottom events of the dynamic fault trees for the high-voltage module, control module, and input protection module, the possible common cause failures in the bottom events are analyzed and listed in pairs, as shown in Table 1:

[0105] Table 1. Possible common-cause failure events and their causes.

[0106]

[0107] The meanings of each event in the dynamic fault tree shown in Figure 2 are shown in Table 2.

[0108] Table 2. Meaning of each event in the dynamic fault tree

[0109]

[0110]

[0111]

[0112]

[0113] Step 2: Construct a comment set based on the proportion of common cause failures and collect qualitative evaluation opinions;

[0114] The set of comments for the qualitative evaluation opinions for:

[0115] ;

[0116] Among them, weak common causes This indicates that common cause failures account for 0.05-0.35% of all failures; weaker common cause failures. This indicates that common cause failures account for 0.25-0.55% of all failures; strong common cause. This indicates that common-cause failures account for 0.45-0.75% of all failures; strong common-cause failures... This indicates that the proportion of common cause failures among all failures is 0.65-0.95; the proportions mentioned in various qualitative evaluation opinions are fuzzy proportions, which will be constructed and defuzzified in step 3;

[0117] For each pair of common-cause failure events, five qualitative evaluation opinions were collected for comprehensive assessment.

[0118] Step 3: Construct a fuzzy matrix by fuzzy quantization of the five qualitative evaluation opinions; construct a decision matrix based on the indicators of the source person corresponding to the evaluation opinions and calculate the final evaluation weight of each of the five qualitative evaluation opinions; aggregate the five evaluation opinions according to the final evaluation weight; defuzzify the aggregation result to obtain the common cause failure factor of each pair of basic event common cause failures;

[0119] With P 30 P 31 Taking the common-cause failure as an example, a trapezoidal membership function is used to fuzzify the linguistic variables of different qualitative evaluation opinions. Each linguistic variable corresponds to a set of trapezoidal fuzzy numbers, resulting in P. 30 P 31 The fuzzy matrix of the common cause failure of the bottom-end event. for:

[0120]

[0121] Based on three indicators—the experience, education level, and job relevance of the evaluation opinions—a decision matrix is ​​constructed according to the indicator values. :

[0122]

[0123] Subsequently, the dispersion of the three indicators of years of experience, education level, and job relevance was found to be 0.079, 0.011, and 0.044, respectively.

[0124] Then, based on the dispersion of the three indicators of years of experience, education level, and job relevance, the weights of the three indicators were calculated to be 0.590, 0.082, and 0.328, respectively.

[0125] Then, the preliminary comment weights of the five qualitative comments were calculated, and the weights were normalized to obtain the final comment weights of the five qualitative comments as 0.256, 0.134, 0.362, 0.162 and 0.086 respectively.

[0126] Then, the fuzzy matrix is ​​weighted row-wise according to the weight index, and the five evaluation opinions are aggregated to obtain the final trapezoidal fuzzy number. The trapezoidal membership function obtained by aggregating trapezoidal fuzzy numbers:

[0127] ;

[0128] Finally, calculate the deblurring result. That is, P 30 P 31 Common failure factors among them.

[0129] Calculate the common cause failure factor for each pair of common cause failure events in the manner described above.

[0130] Step 4: The dynamic fault tree considering common cause failures is transformed into a discrete-time Bayesian network to obtain the Bayesian structure diagram, conditional probability table and marginal probability table. The obtained Bayesian structure diagram is shown in Figure 3.

[0131] Nodes in the diagram Meaning and The meaning is the same as in Table 2.

[0132] The black box indicates that there is a common cause failure decomposition. The root node corresponding to each pair of common cause failure events is decomposed into three nodes in the Bayesian diagram. The meaning of the nodes in the box is shown in Table 3.

[0133] Table 3. Meaning of nodes in the box

[0134]

[0135] The common cause failures of each pair of common cause failure events are shown in Table 1.

[0136] These are additional intermediate nodes. These nodes do not have specific physical meanings, but they have certain logical meanings, corresponding to the outputs of specific logic gates. The meanings are shown in Table 4.

[0137] Table 4 The meaning of a node

[0138]

[0139] Step 5: Perform bidirectional reasoning on the dynamic reliability of the high-voltage power supply, calculate the reliability and maintainability indicators of the three sub-parts of each time period and the final top event of the high-voltage power supply, where the reliability indicators are shown in Table 5 and the maintainability indicators are shown in Table 6.

[0140] Table 5 Failure Probability and Reliability

[0141]

[0142] According to Table 5, the probability of a final peak event in the high-voltage power supply within 5000 working hours is approximately [missing information]. The reliability is 9.98661 × 10⁻⁶. -01 It exhibits good reliability. Looking at different time periods, the failure rate of the high-voltage power supply during the first operating period is approximately... The failure rate during the second working period was approximately The probability of failure is higher in the second time period than in the first, meaning the reliability is lower in the second time period than in the first. Therefore, maintenance personnel should pay closer attention to the operating status of the high-voltage power supply during the second time period and monitor its normal operation. Looking at different components, the input section has a higher probability of failure; improving power supply reliability can begin with addressing this input section.

[0143] Based on the principle of graded inspection and maintenance, and according to calculations, the posterior probabilities of the three sub-modules of the high-voltage power supply when a top-level fault occurs are shown in Table 6.

[0144] Table 6. Posterior probabilities of the three sub-modules

[0145]

[0146] As shown in Table 6, when a high-voltage power supply top event occurs, the probabilities of the three sub-top events occurring, ranked from highest to lowest, are as follows: Therefore, when the high-voltage power supply outputs abnormally during the working period, it should be handled in the following order: The investigation proceeds in the following order. Once the sub-part that caused the failure is identified, the posterior probability table of the underlying events corresponding to the top event of the failure sub-part can be queried, and the underlying events can be investigated one by one according to the magnitude of the posterior probability until the underlying event that caused the failure is found.

[0147] Meanwhile, a comparative experiment was set up without considering common-cause failure. Calculations showed that, without considering common-cause failure, the probability of the final peak event in the high-voltage power supply within 5000 hours is approximately [missing information]. Compared to considering common-cause failures, the failure rate decreased by approximately 6.05%, demonstrating the impact of common-cause failures on the reliability analysis of high-voltage power supplies and illustrating the engineering application value of this method.

Claims

1. A Bayesian inference method for reliability of high voltage power supply under common cause failure effect, characterized in that, The method comprises the following steps: Step 1, constructing a dynamic fault tree of the high-voltage power supply; The high-voltage power supply comprises a high-voltage module, a control module and an input protection module; The dynamic fault tree of the high-voltage power supply comprises a high-voltage module dynamic fault tree, a control module dynamic fault tree and an input protection module dynamic fault tree; the top events of the three module dynamic fault trees are connected by an OR gate, that is, the high-voltage power supply dynamic fault tree is obtained; the output event of the top events of the three module dynamic fault trees connected by the OR gate is the top event of the high-voltage power supply dynamic fault tree, representing the final failure of the high-voltage power supply; Step 2, constructing an evaluation set for the proportion of common cause failure and collecting qualitative evaluation opinions; For the proportion of common cause failure in each pair of common cause failure bottom events, five qualitative evaluation opinions are collected for comprehensive evaluation; the set of comments of the qualitative evaluation opinion is: = weak co-factor = less strong co-factor = stronger co-factor = strong co-factor ; wherein, weak common cause represents the proportion of common cause failure in all failures is 0.05-0.35; relatively weak common cause represents the proportion of common cause failure in all failures is 0.25-0.55; relatively strong common cause represents the proportion of common cause failure in all failures is 0.45-0.75; strong common cause represents the proportion of common cause failure in all failures is 0.65-0.95; wherein the proportions described by various qualitative evaluation opinions are fuzzy proportions, and the construction and defuzzification conversion of a fuzzy matrix will be performed in step 3. Step 3, determining the common cause failure factor of each pair of bottom event common cause failure; The five qualitative evaluation opinions are quantified to construct a fuzzy matrix; a decision matrix is constructed according to the index of the source of the evaluation opinions and the final evaluation weight of the five qualitative evaluation opinions is calculated; the five evaluation opinions are aggregated according to the final evaluation weight; the aggregation result is de-fuzzified to obtain the common cause failure factor of each pair of bottom event common cause failure; Step 4, converting the dynamic fault tree of the common cause failure into a discrete-time Bayesian network to obtain a Bayesian network graph, a conditional probability table and an edge probability table; Step 5, bidirectional reasoning of the dynamic reliability of the high-voltage power supply to calculate the reliability and maintainability indexes of the high-voltage power supply in each time period. In step 1, the process of constructing the dynamic fault tree of the high-voltage power supply is as follows: 2.The Bayesian inference method for reliability of high-voltage power supply under common cause failure effect according to claim 1, characterized in that, Step 1.1, querying the structure of the high-voltage power supply and collecting records of system historical failures, maintenance reports and failure mode and effects analysis reports; Step 1.2, according to the query results in step 1.1, regarding the final failure of the high-voltage module, the control module and the input protection module as the top events of the respective modules; then, the top events of the modules are decomposed level by level, and the dependency between the upper-level events and the lower-level events in the decomposition process is described through static gates and dynamic gates according to the failure logic, and the decomposition is performed level by level until the bottom events that cannot be further decomposed are obtained; the top events of the three module dynamic fault trees are connected by an OR gate, and the high-voltage power supply dynamic fault tree is obtained; The static gate comprises an AND gate and an OR gate; the dynamic gate comprises a priority AND gate, a cold spare gate, a sequence-related gate and a function-related gate; The bottom event that cannot be further decomposed is the smallest functional unit or the most basic failure form constituting the high-voltage power supply system, and there is no finer "failure cause", which is a natural bottom event; the bottom event that cannot be further decomposed includes inherent failure of a component, external environmental disturbance and human operation failure; The upper-level event is the output event of the static gate and the dynamic gate, and the lower-level event is the decomposition event of the output event; Step 1.3, determining the failure distribution function of each bottom event and the parameter value of the distribution function: According to the query results of 1.1, the probability paper diagram or the trend diagram of each bottom event is drawn by using the historical failure records to determine the distribution type fitted by the data points; ​ Then the distribution function parameters corresponding to each bottom event distribution type are obtained by maximum likelihood estimation method; Step 1.4, according to the query result of 1.1, determine the common cause failure caused by environmental stress, input abnormality, design defect and human operation in the bottom event, list the bottom event and the common cause failure reason of the common cause failure. 3.The Bayesian inference method for reliability of high-voltage power supply under common cause failure effect according to claim 1, characterized in that, In step 3, the determination process of the common cause failure factor is as follows: Step 3.

1. Constructing the fuzzy matrix by fuzzy quantifying the qualitative evaluation opinions : The trapezoidal membership function is used to fuzz the language variables of different qualitative evaluation opinions. Each language variable corresponds to a set of trapezoidal fuzzy numbers. The range of the common cause failure factor is defined as The fuzzing process is as follows: ； The trapezoidal fuzzy numbers of the qualitative evaluation opinions for each pair of common cause failures are combined to obtain the fuzzy matrix for each pair of common cause failures. Each element in the fuzzy matrix is ​​the first element. The first trapezoidal fuzzy number of the evaluation: ； Step 3.2, obtain the final comment weight of each of the five qualitative comment opinions: Firstly, according to the three indexes of experience years, education level and work relevance of the source of qualitative evaluation opinions, a decision matrix is constructed according to the index values. Each element in the decision matrix is the score of the first index of the source of the first evaluation opinion corresponding to the first evaluation opinion; the projection probability of the score of the first index of the source of the first evaluation opinion corresponding to the first evaluation opinion is defined as: the first index of the source of the first evaluation opinion corresponding to the first evaluation opinion.​ ； Then, the dispersion of the three indexes of experience, education level and work relevance is obtained: ； Then, according to the dispersion, the weights of the three indexes of experience, education level and work relevance are calculated: ； Thereafter, the preliminary evaluation weight of each of the five qualitative comment opinions is calculated : ； Finally, the weights are normalized to obtain the final comment weight of each of the five qualitative comment opinions; Step 3.3, aggregate the five evaluation opinions according to the final comment weight of each of the five qualitative comment opinions: Firstly, the final comment weight and the fuzzy matrix of each pair of common cause failure are weighted and averaged by row to obtain the trapezoidal fuzzy number after aggregation of the five evaluation opinions; Then, the aggregated trapezoidal membership function of each pair of common cause failure is obtained according to the trapezoidal fuzzy number: ； Wherein, and are the aggregated trapezoidal fuzzy numbers respectively; Step 3.4, defuzzification transformation is performed on the aggregation result to obtain the common cause failure factor: The center area method is used to De-mangling transformation is performed: ； where is the argument of, representing the common cause failure factor; represents the deblurring result, i.e., the common cause failure factor for each pair of bottom events. 4.The Bayesian inference method for reliability of high-voltage power supply under common cause failure effect according to claim 1, wherein, In step 4, the acquisition process of the Bayesian network diagram, conditional probability table and marginal probability table is as follows: Step 4.1, length of high voltage power supply operation equally divided into time periods: The length of each time period is , denoted as ; on the basis of time periods, a time period in which the high-voltage power supply does not fail is introduced, so that there are time periods in total, the first time periods represent that the high-voltage power supply fails in time periods, and the time period represents that the high-voltage power supply does not fail, and the time period division process is as follows: ； Step 4.2, map the high-voltage power supply dynamic fault tree diagram considering common cause failure into a Bayesian network diagram: Firstly, the high-voltage power supply top event is regarded as a leaf node in the Bayesian network diagram; the output events of each logic gate are regarded as intermediate nodes in the Bayesian network diagram; and the bottom events are regarded as root nodes in the Bayesian network diagram; Then, the root node corresponding to each pair of common cause failure events is decomposed into three nodes in the Bayesian diagram, and the three nodes respectively represent one common cause failure event and two independent failure events; Finally, the events are connected by directed line segments to form a Bayesian network diagram; Step 4.3, obtain the marginal probability table of the root node: First, according to the probability density function or distribution function of each base event, the root node edge probability table of the base event in different time periods is obtained; Bottom event In the first Time period The probability of failure is: ； Then, the marginal probability of the decomposed bottom event of the common cause failure is decomposed according to the common cause failure factor obtained in step 3 to obtain the final root node marginal probability table of each decomposed bottom event in different time periods; Step 4.4, obtain the intermediate node conditional probability table: According to the logical relationship of different logic gates, the conditional probability table of the intermediate node corresponding to each logic gate output event is obtained; Step 4.4.1, according to the logic judgment of the logic gate, whether the output event is faulty in each time period when the sub-event is faulty in different time periods, so as to obtain the value of the intermediate node under various value combinations of the sub-nodes corresponding to the intermediate node; Step 4.4.2, obtaining the conditional probability table of the intermediate node corresponding to the function-related gate input event : ； wherein denotes the probability density function of the input event denotes the probability density function of the input event; denotes the triggering event; denotes the failure period of the triggering event; j B denotes the failure period of the input event Step 4.4.3, obtaining the conditional probability table of the intermediate node corresponding to the cold spare parts door spare parts event : ； wherein represents the probability density function of the input event ; represents the probability density function of the spare event ; represents the failure period of the input event event ; represents the failure period of the spare event ; Step 4.4.4, update the intermediate node conditional probability table corresponding to the event containing the common cause failure sub-event; When the common cause failure event does not occur, the value of the intermediate node conditional probability table is determined by the sub-nodes corresponding to the independent failure events under the intermediate node, which is irrelevant to the sub-nodes corresponding to the common cause failure event; When the common cause failure event occurs, the intermediate node conditional probability table is determined by referring to the corresponding sub-node values when the two independent failure events occur simultaneously.

5. The Bayesian inference method for reliability of high voltage power supply under common cause failure effect according to claim 1, characterized in that, In step 5, bidirectional reasoning is performed for the dynamic reliability of the high-voltage power supply, specifically: Step 5.1, forward reasoning of the upper node probability in the Bayesian network from bottom to top through the total probability formula; Let each bottom event be in turn, each intermediate event is, high-voltage power supply top event The first Probability of failure of the first time period Is: ； In the formula, This indicates the failure status of each underlying event at different time periods. This indicates the failure status of each intermediate event at different time periods. Indicates the top event at the 1st Failure scenarios within a time period; n represents the total number of events. Represents the total number of intermediate events; According to the top event failure probability of different time periods , the reliability of the high-voltage power supply in different time periods is calculated : ； In the formula, if the time period is taken then the reliability of the high voltage power supply over the entire operating time is obtained : ； Step 5.2, when the top event of the high-voltage power supply occurs, backward reasoning of the occurrence probability of each bottom event node in the Bayesian network from top to bottom through the Bayesian formula; When the high voltage power supply top event In the first time period failure, calculate the probability of any event failure at any time period is: ​ 。 6. An electronic device, comprising: one or more processors; a memory; one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to perform the Bayesian inference method of any one of claims 1-5.

7. A computer readable storage medium, the computer readable storage medium storing program code, the program code being executable by a processor to perform the Bayesian inference method of any one of claims 1-5.

Images