WO2014173258A1 - 响应计划的可靠性分析方法及装置 - Google Patents

响应计划的可靠性分析方法及装置 Download PDF

Info

Publication number
WO2014173258A1
WO2014173258A1 PCT/CN2014/075730 CN2014075730W WO2014173258A1 WO 2014173258 A1 WO2014173258 A1 WO 2014173258A1 CN 2014075730 W CN2014075730 W CN 2014075730W WO 2014173258 A1 WO2014173258 A1 WO 2014173258A1
Authority
WO
WIPO (PCT)
Prior art keywords
psf
reliability
response plan
psfs
node
Prior art date
Application number
PCT/CN2014/075730
Other languages
English (en)
French (fr)
Inventor
张力
李鹏程
戴立操
胡鸿
蒋建军
陈青青
黄卫刚
戴忠华
王春辉
苏德颂
李晓蔚
Original Assignee
湖南工学院
南华大学
中广核核电运营有限公司
大亚湾核电运营管理有限责任公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 湖南工学院, 南华大学, 中广核核电运营有限公司, 大亚湾核电运营管理有限责任公司 filed Critical 湖南工学院
Publication of WO2014173258A1 publication Critical patent/WO2014173258A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00Administration; Management
    • G06Q10/04Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"

Definitions

  • a response plan refers to a decision process for formulating a course of action, method, or plan for solving an abnormal event.
  • the operator plans the actions to be performed after completing the status assessment.
  • alternative methods, strategies, and plans should be identified to evaluate them to select an optimal or feasible response plan.
  • the digital control system (DCS) response plan is a state-oriented procedure (SOP).
  • the reasons may include: dissipation (di S tm C tion), memory failure (wrong reasoning), excessive cause (excessive demand); specific antecedents may include: Neglect side, inadequate training, inadequate impact assessment, consequent, model error, Violation, Overlook precondition, short-sighted planning Too short planning horizon ).
  • the priority of the prior error is due to the wrong diagnosis and Flow failure specific standard legal ago because of a higher priority (legitimate higher priority) and contradictory (conflicting criteria).
  • IDAC causal model of operator problem-solving in 2007 The model argues that factors affecting action planning are: attention, cognitive bias, time pressure, perceived severity of the consequences, perceived responsibility for decision making, perceived task complexity, role and Responsibility, knowledge and experience, memory of input information, team factors (coordination, cooperation, effectiveness of communication, quality of communication, team composition, leadership) and human-machine interface.
  • the above method analyzes the factors affecting the response plan, and there is no effective quantitative model and data to support the quantitative calculation of the reliability of the response plan.
  • the probability analysis of the error response of the HRA method such as CREAM is basically the first to determine the basic error probability of the response plan error, and then consider the state of the Performance Shaping Factor (PSF) to correct the error probability of the response plan.
  • PSF Performance Shaping Factor
  • the problems in the related art are: lack of consideration of the characteristics of the digital human-machine system to study the reliability of the response plan. Lack of quantitative analysis techniques for reliability of nuclear power plant operator response plans considering PSF causality. Lack of digital simulator data to support quantification of response plans.
  • Embodiments of the present invention provide a reliability analysis method and apparatus for a response plan, to at least solve a plurality of PSFs that cannot be combined in a reliability analysis process of a response plan in the related art, and consider a causal relationship of the PSF.
  • the problem of analyzing the reliability of the response plan is provided.
  • the reliability analysis method of the response plan includes: determining a plurality of behavior forming factors PSF to be used in the reliability analysis process of the response plan, an association relationship between each PSF among the plurality of PSFs, and each PSF and Corresponding to the relationship between the reliability of the plan; establishing a reliability analysis model of the response plan according to the determined plurality of PSFs, the relationship between the PSFs of the plurality of PSFs, and the relationship between the PSFs and the reliability of the response plan, Analyze the reliability of the response plan.
  • the reliability analysis model of the response plan is a Bayesian network model, wherein each network node in the Bayesian network model respectively corresponds to one PSF or a response plan reliability node.
  • the reliability analysis model for establishing a response plan according to the determined plurality of PSFs, the relationship between the PSFs of the plurality of PSFs, and the relationship between the respective PSFs and the response plan reliability comprises: according to the determined PSFs
  • the relationship between the associations and the reliability of each PSF and the response plan reliability sets a part of the PSFs in the plurality of PSFs to the root node PSF of the Bayesian network model, wherein each PSF in the root node PSF is in a different state.
  • the statistical results of the prior probability distribution are independent of the remaining PSFs other than the PSF and the response plan reliability of the plurality of PSFs; the other nodes except the root node PSF in the Bayesian network model are set as the child nodes PSF and the response Planning a reliability node, wherein the conditional probability distribution statistics for each PSF and response plan reliability in the sub-PSF are in different states depending on one or more PSFs in the root node PSF, and/or depending on other parts of the PSF One or more PSFs in .
  • analyzing the reliability of the response plan comprises: calculating a prior probability distribution of each PSF of the root node PSF in different states in a preset period and determining a PSF and response plan reliability of the child nodes other than the root node PSF The conditional probability distribution of each node in the node in different states; according to the statistics of the prior probability statistics of the root node PSF in different states, the relationship between each PSF, and the relationship between the PSF and the reliability of the response plan, The conditional probability distribution of the child nodes PSF except the root node PSF in different states and the conditional probability distribution of the response plan reliability nodes in different states are obtained, and the reliability of the response plan is evaluated according to the causal analysis of the Bayesian network.
  • the analyzing the reliability of the response plan further comprises: comparing the acquired prior probability statistical result of the root node PSF in a bad state with a posterior probability result of the root node PSF in a bad state, determining the response plan The PSF with the greatest impact on reliability.
  • a reliability analysis apparatus for a response plan is provided.
  • the reliability analysis apparatus includes: a determination module, configured to determine a plurality of behavior forming factors PSF to be used in the reliability analysis process of the response plan, and association between each of the plurality of PSFs Relationship and relationship between each PSF and response plan reliability; an analysis module for correlating between the plurality of PSFs determined, the association between each PSF of the plurality of PSFs, and the reliability of each PSF and the response plan The relationship establishes a reliability analysis model of the response plan and analyzes the reliability of the response plan.
  • the reliability analysis model of the response plan is a Bayesian network model, wherein each network node in the Bayesian network model respectively corresponds to one PSF or a response plan reliability node.
  • the analysis module includes: a first setting unit, configured to set a part of the PSFs in the plurality of PSFs to Bayes according to the relationship between the determined PSFs and the relationship between the respective PSFs and the reliability of the response plan
  • the root node PSF of the network model wherein the statistical result of the prior probability distribution of each PSF in the root node PSF is independent of the remaining PSFs of the plurality of PSFs except the PSF and the reliability of the response plan
  • the partial nodes other than the root node PSF in the Bayesian network model are set as the child node PSF and the response plan reliability node, wherein the conditional probability distribution statistical result of each PSF and response plan reliability in the child node PSF Depends on one or more PSFs in the root node PSF, and/or depends on one or more PSFs pointed to by one or more PSFs in the root node PSF.
  • the analyzing module further comprises: a statistical unit, configured to calculate a prior probability distribution of each PSF of the root node PSF in different states in a preset period, and determine a PSF and a response plan reliability of the child node other than the root node PSF A conditional probability distribution of each node in the node in different states; an analyzing unit, configured to perform a correlation between the probability and the statistical result of the root node PSF and the relationship between the PSFs and the reliability of each PSF and the response plan reliability The relationship obtains the conditional probability distribution of the child nodes PSF except the root node PSF in different states and the conditional probability distribution of the response plan reliability nodes in different states, and evaluates the reliability of the response plan according to the causal analysis of the Bayesian network.
  • a statistical unit configured to calculate a prior probability distribution of each PSF of the root node PSF in different states in a preset period, and determine a PSF and a response plan reliability of the child node other than the root node PSF A condition
  • the analyzing module further comprises: a comparing unit, configured to compare the acquired prior probability statistical result of the root node PSF in a bad state with a posterior probability result of the root node PSF in a bad state, and determine a reliable response plan The most influential PSF.
  • a comparing unit configured to compare the acquired prior probability statistical result of the root node PSF in a bad state with a posterior probability result of the root node PSF in a bad state, and determine a reliable response plan The most influential PSF.
  • a plurality of PSFs to be used in the reliability analysis process of the response plan, an association relationship between each PSF of the plurality of PSFs, and an association relationship between each PSF and the reliability of the response plan are determined; Establishing a reliability analysis model of the response plan according to the determined plurality of PSFs, an association relationship between each PSF of the plurality of PSFs, and an association relationship between each PSF and a response plan reliability, and the response plan is configured
  • the reliability analysis is carried out to solve the problem that the reliability of the response plan is analyzed without considering the causal relationship between the multiple PSFs used in the reliability analysis process of the response plan in the related art, and then the digital control room is further analyzed.
  • the operator response plan reliability analysis provides qualitative and quantitative methods and tools to provide a countermeasure for the plant to reduce the operator's response plan failure probability.
  • BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are set to illustrate,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
  • FIG. 1 is a flow chart of a reliability analysis method of a response plan according to an embodiment of the present invention. As shown in FIG.
  • the method may include the following processing steps: Step S102: determining a plurality of behavior forming factors PSF to be used in a reliability analysis process of a response plan, association relationships between respective PSFs of the plurality of PSFs, and each Correlation relationship between PSF and response plan reliability; Step S104: According to the determined multiple PSFs, the association relationship between each PSF among the multiple PSFs, and each
  • the relationship between the PSF and the reliability of the response plan establishes a reliability analysis model of the response plan, and analyzes the reliability of the response plan.
  • the reliability of the response plan is analyzed without considering the causal relationship of the plurality of PSFs used in the reliability analysis process of the response plan. Using the method shown in FIG.
  • the control room operator responds to the plan reliability analysis by providing qualitative and quantitative methods and tools to provide a countermeasure for the plant to reduce the operator's response plan failure probability.
  • identifying the PSF factor affecting the operator's response plan reliability and its causal relationship establishing a Bayesian network model for qualitative analysis of the response plan, responding to the operator
  • the basis for quantifying the reliability of the plan based on the Bayesian network model of the established response plan, the prior probability and conditional probability of the network node are collected by the simulator experiment, and the data that is difficult to collect can be collected. Use event report analysis and expert judgment to obtain data.
  • a simple operator response plan can include: Response to an alarm.
  • the operator's behavior directly corresponds to the preset action, such as: Response to the emergency procedure can be handled directly according to the procedure without having to select path.
  • Complex response plans such as: no corresponding procedures, procedures, and rules can be followed, or existing procedures have been proven to be unable to meet actual work requirements, as well as operators needing to rebuild new ones Respond to the plan and evaluate the feasibility and effectiveness of the response plan.
  • the reliability analysis model of the foregoing response plan may be a Bayesian network model, wherein each network node in the Bayesian network model corresponds to one PSF or a response plan reliability node, respectively.
  • the nodes that represent the variables with finite states, the nodes can be any abstract problems, such as: device component status, test values, organizational factors, human Diagnosis results, etc.
  • the directed edge indicates the probability causal relationship between nodes.
  • the starting node i of the directed edge is the parent node of the ending node j, j is called the child node, and the node without the parent node and only the child node is called the root node.
  • DAG implies a conditional independence hypothesis: Given its set of parent nodes, each variable is independent of its non-children.
  • P is the quantitative part, which is the probability distribution on V. For the discrete case, it can be represented by the Conditional Probability Table (CPT), which is used to quantify the influence of the parent node on the child nodes.
  • CPT Conditional Probability Table
  • the probability distribution function of the root node is the edge probability distribution function. Since the probability of the node is not conditional on other nodes, the probability is the prior probability, and the other nodes are the conditional probability distribution functions.
  • the reliability analysis model for establishing a response plan according to the determined plurality of PSFs, the relationship between the PSFs among the plurality of PSFs, and the relationship between the respective PSFs and the response plan reliability may include the following Operation: Step S1: setting a part of PSFs in the plurality of PSFs as a root node PSF of the Bayesian network model according to the relationship between the determined PSFs and the relationship between the PSFs and the reliability of the response plans, where The prior probability distribution statistical result of each PSF in the root node PSF in different states is independent of the remaining PSFs in the plurality of PSFs except the PSF and the response plan reliability; Step S2: removing the root node from the Bayesian network model The other nodes except the PSF are set as the child node PSF and the response plan reliability node, wherein the conditional probability distribution statistical result of each PSF and the response plan reliability in the child node PSF is different depending on one of the root node PSF
  • analyzing the reliability of the response plan may include the following steps: Step S3: Statistic a priori probability distribution of each PSF of the root node PSF in a preset period in different states and determining a root node PSF The conditional probability distribution of the child node PSF and the response plan reliability node are in different states; Step S4: acquiring the relationship other than the root node PSF according to the relationship between the PSFs and the relationship between the PSF and the response plan reliability The conditional probability distribution of the child nodes PSF in different states and the conditional probability distribution of the response plan reliability nodes in different states, the reliability of the response plan is evaluated according to the causal analysis of the Bayesian network.
  • the above-described preferred implementation process is further described in conjunction with the fuzzy Bayesian method of data acquisition and response planning quantitative calculations. The first part, data collection
  • Each picture is evaluated to obtain the probability distribution (assuming a total of 100 sub-pictures, through the evaluation of the expert group, 90 pictures are satisfied with the preset conditions, 8 pictures are normal, and 2 pictures are poor, then the prior probability distribution of the human-machine interface is: 0.9, 0.08, 0.02, the same can be The prior probability distribution of other influencing factors).
  • Conditional probability distribution of network nodes The operator's knowledge and experience, stress levels, attitudes, etc. were assessed during the course of the experiment. This requires the testee to self-assess each time a critical task is completed, which requires seeking truth from facts. Statistical assessment results, knowledge and experience (assuming three levels: good, medium, poor), stress levels (assuming three levels: good, medium, poor), task complexity (assuming three levels: good, medium, Conditional probability distribution such as difference). At the same time, the conditional probability distribution of the reliability of the response plan is obtained by responding to the statistical distribution of the experimental results of the plan.
  • the level of training and communication affects the knowledge and experience of the operator
  • you can select the personnel of different training levels to conduct experiments which may include: training, training, training, training, training, training, and communication
  • the results of a group of operators with good levels (requires the knowledge and experience acquired by the operator)
  • the results of the test, the level of training and the level of communication are generally average (the operator needs to assess the knowledge and experience acquired) and the level of training and communication of general experimental results (required by the operator) Knowledge and experience are assessed) to obtain a conditional probability distribution of knowledge and part of experience.
  • a conditional probability distribution of all knowledge and experience can be obtained.
  • conditional probability distribution of the node variables such as the complexity of the task can be obtained.
  • model estimation can be performed by means of expert judgment or regression techniques.
  • the probability of each factor being usually in what state is usually discussed by expert discussion can be a triangular fuzzy number (ie the most likely value, the best value, the worst value) Evaluate organizational factors, such as: (0.1, 0.3, 0.6), or, in descriptive language, for example: high, medium, and low.
  • a fuzzy membership function can be introduced to determine the probability value that a factor is in a particular state.
  • ( 1 ) s is the fuzzy comprehensive score of the factor, which is a triangular fuzzy number: ( l Xc k )
  • Bayesian network analysis chain rule shows that a BN is a description of the joint distribution of all variables in the DAG, and the edge probability and conditional probability of each node in the network can be calculated.
  • i a collection of parent nodes.
  • the edge probability of ⁇ is:
  • Table 2 is a table of conditional probability information for a child node in accordance with a preferred embodiment of the present invention.
  • the conditional probability P of the child node "Knowledge and experience KE" (knowledge and experience I team exchange level, training level), as shown in Table 2: Table 2
  • Diagnostic analysis is a reason for inferring conclusions. It is a bottom-up analysis process. The purpose is to find the possibility of various causes of the results when the results have been known. It is known that some kind of result has occurred, and according to the Bayesian network calculation, the cause and the probability of occurrence of the result are obtained.
  • analyzing the reliability of the response plan may further include the following processing: Step S5: after the acquired root node PSF is in a bad state, the prior probability statistics result and the root node PSF are in a bad state. The probability results are compared to determine the PSF that has the greatest impact on the reliability of the response plan.
  • FIG. 2 is a schematic diagram of a Bayesian network model of an operator response plan in accordance with a preferred embodiment of the present invention.
  • the reliability of the response plan is mainly influenced by the mental state of the frontline operator, the information in the memory, and the inherent attributes of the personality.
  • the operator's knowledge and experience will recognize the specific power plant status and what response strategy or plan should be taken.
  • Knowledge and experience are mainly influenced by organizational training and team communication and cooperation; if the training is not enough, it will have a negative impact on the operator's knowledge and experience.
  • the team's communication and cooperation can make up for the lack of knowledge and experience of the individual operators.
  • the stress level will have a great impact on the formulation of the response plan.
  • the stress level is mainly affected by the severity of the event, the complexity of the task and the available time.
  • the complexity of the same task is mainly affected by the quality of the digital protocol design and digitization.
  • the impact of the human-machine interface design, the complexity of the tasks in the procedures, the complexity of the tasks that the operator needs to complete, the perfection of the procedures or programming is conducive to guiding the operator to respond to the plan, the design of the human-machine interface is flawed, It is difficult for the operator to obtain useful information that is conducive to the development of the response plan.
  • the response plan is also affected by the attitude of the operator.
  • the operator's attitude is good and the sense of responsibility is strong. It is difficult to violate the rules and concentrate. Among them, the attitude of the operator is mainly affected by the safety culture and management of the organization. If the safety culture is not deeply rooted in the hearts of the people, the operator's risk awareness is weak and the safety attitude is poor.
  • the response plan can be influenced by the team's level of communication and cooperation, training level, digital procedures, digital human-machine interface, the severity of the incident, the available time of accident handling, safety culture and organizational management level.
  • 3 is a block diagram showing the structure of a reliability analysis apparatus for a response plan according to an embodiment of the present invention. As shown in FIG.
  • the reliability analysis apparatus of the response plan may include: a determining module 10, configured to determine a plurality of forming factor PSFs to be used in a reliability analysis process of the response plan, and each of the plurality of PSFs An association relationship between the PSF and the reliability of the response plan; the analysis module 20, configured to determine, according to the determined plurality of PSFs, association relationships between PSFs of the plurality of PSFs, and respective PSFs A reliability analysis model of the response plan is established in association with the reliability of the response plan, and the reliability of the response plan is analyzed.
  • a determining module 10 configured to determine a plurality of forming factor PSFs to be used in a reliability analysis process of the response plan, and each of the plurality of PSFs An association relationship between the PSF and the reliability of the response plan
  • the analysis module 20 configured to determine, according to the determined plurality of PSFs, association relationships between PSFs of the plurality of PSFs, and respective PSFs A reliability analysis model of the response plan is established in association with
  • the reliability analysis model of the foregoing response plan may be a Bayesian network model, wherein each network node in the Bayesian network model corresponds to one PSF or a response plan reliability node, respectively.
  • each network node in the Bayesian network model corresponds to one PSF or a response plan reliability node, respectively.
  • the foregoing analysis module 20 may include: a first setting unit 200, configured to: according to the relationship between the determined PSFs and the relationship between each PSF and the reliability of the response plan
  • the partial PSFs in the PSFs are set to the root node PSF of the Bayesian network model, wherein the statistical results of the prior probability distribution of each PSF in the root node PSF are independent of the remaining PSFs and responses in the plurality of PSFs except the PSF.
  • the second setting unit 202 is configured to set other nodes except the root node PSF in the Bayesian network model as the child node PSF and the response plan reliability node of the Bayesian network model, wherein the child node PSF
  • the conditional probability distribution statistics that are in a different state from the response plan reliability depend on one or more PSFs in the root node PSF, and/or depend on one or more PSFs pointed to by one or more PSFs in the root node PSF .
  • the foregoing analysis module 20 may further include: a statistics unit 204, configured to collect a prior probability distribution of each PSF of the root node PSF in a preset period in different states, and determine a root node PSF.
  • the conditional probability distribution of each of the outer child node PSF and the response plan reliability node is in a different state; the analyzing unit 206 is configured to determine the relationship between the respective PSFs and the reliability of each PSF and the response plan according to the determination The relationship relationship obtains the conditional probability distribution of the child nodes PSF except the root node PSF in different states and the conditional probability distribution of the response plan reliability nodes in different states, and evaluates the reliability of the response plan according to the causal analysis of the Bayesian network. .
  • the analyzing unit 206 is configured to determine the relationship between the respective PSFs and the reliability of each PSF and the response plan according to the determination
  • the relationship relationship obtains the conditional probability distribution of the child nodes PSF except the root node PSF in different states and the conditional probability distribution of the response plan reliability nodes in different states, and evaluates the reliability of the response plan according to the causal analysis of the Bayesian network.
  • the foregoing analysis module 20 may further include: a comparison unit 208, configured to obtain a posterior probability result of the root node PSF in a bad state and a posterior probability result of the root node PSF being in a bad state. Compare and determine the PSF that has the greatest impact on the reliability of the response plan.
  • a comparison unit 208 configured to obtain a posterior probability result of the root node PSF in a bad state and a posterior probability result of the root node PSF being in a bad state. Compare and determine the PSF that has the greatest impact on the reliability of the response plan.
  • Room operator response planning reliability analysis provides qualitative and quantitative methods and tools to provide countermeasures for power plant reduction operator response plan failure probability; for nuclear power plant digital main control room operator reliability analysis (HRA) and probabilistic safety evaluation (PSA) Providing operator response plan reliability interface data and calculation tools, establishing a fuzzy Bayesian method for calculating the reliability of response plans considering PSF causality, can improve the accuracy of HRA and PSA analysis;
  • HRA nuclear power plant digital main control room operator reliability analysis
  • PSA probabilistic safety evaluation
  • the staff provides support for error prevention, training and scenario development; provides technical support and tools for reliability analysis and safety assessment of the master control room operator response plan or decision behavior in the digital industrial system.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Marketing (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本发明公开了一种响应计划的可靠性分析方法及装置,在上述方法中,确定在响应计划的可靠性分析过程中待使用的多个行为形成因子PSF、多个PSF中各个PSF之间的关联关系以及各个PSF与响应计划可靠性之间的关联关系;根据确定的多个PSF、多个PSF中各个PSF之间的关联关系以及各个PSF与响应计划可靠性之间的关联关系建立响应计划的可靠性分析模型,对响应计划的可靠性进行分析。根据本发明提供的技术方案,进而为数字化主控室操纵员响应计划可靠性分析提供定性与定量的方法与工具,为电厂降低操纵员响应计划失效概率提供对策。

Description

响应计划的可靠性分析方法及装:
技术领域 本发明涉及人因可靠性分析领域, 具体而言, 涉及一种响应计划的可靠性分析方 法及装置。 背景技术 响应计划是指为解决异常事件而制定行动方针、 方法或方案的决策过程。 操纵员 在完成状态评估之后对即将执行的行动加以计划。 在事故条件下, 操纵员为了识别合 适的方法来实现目标, 应该识别可供选择的备选方法、 策略和计划, 从而对它们进行 评估选择最优的或可行的响应计划。 通常而言, 在概率安全评价 (PSA) 始发事件情况下, 数字化控制系统 (DCS ) 的响应计划是执行状态导向的规程(SOP)。对于规程和规程中路径的选择是相对简单 的响应计划, 但对于某些事故, 如果没有合适的规程, 操纵员的认知负荷会变得很大, 对于工作记忆, 长期记忆和注意力资源要求极高。 此种情况下, 响应计划的制定和计 划的正确性等都会遇到困难, 响应计划的可靠性由此降低。 目前国内外文献对操纵员响应计划的研究甚少,主要集中在响应计划的失误模式、 影响因素等定性层面研究, 且研究对象均为基于模拟技术的核电厂主控室, 例如: Hollnagel于 1998年在其所著的(Cognitive Reliability and Error Analysis Method》书中 将响应计划失误模式分为 "不充分的计划"和 "优先性失误 (priority error) ", 并指出 引起不充分的计划的一般前因可以包括:注意力分散(diStmCtion)、记忆失效(memory failure )、 错误的推理 (wrong reasoning ) 过度的需求 (excessive demand); 具体的前 因可以包括: 目标错误 (error in goal)、 忽视副效应 (overlook side)、 不充分的培训 ( inadequate training ) 后果影响评价不充分 ( consequent ) 错误建模 (model error ) 违规 (Violation)、 忽视前提条件 (Overlook precondition )、 规划时目光短浅 (too short planning horizon )。 优先性失误的一般前因为错误的诊断和交流失效, 具体的前因为合 法的更高的优先级 (legitimate higher priority) 和矛盾的标准 (conflicting criteria )。
Kontogiannis于 1997年在 《A framework for the analysis of cognitive reliability in complex systems: a recovery centred approach》一文中指出影响响应计戈 ij的主要因素就 是培训实践 (training practices ) 和规程 (procedures)。 这些共同作用可能带来副作用 和花费更多的时间和资源。 规程和培训实践上的局限性可能引起忽视备选方案、 采用 不可能的方法以及中断方法的测试。 另外, 经验的缺乏和支持的缺乏可能引起计划序 列不合适, 错误的线索会影响计划的形成。 有限的时间窗口使得操纵员可能忽视对事 件进展的预计, 从而难以做出具体的合适的计划。 由于系统的设计方式使得恢复线索 的缺乏和操纵员警觉性的缺乏而不能纠正计划。 Chang禾口 Mosleh于 2007年发表的 Cognitive modeling and dynamic probabilistic simulation of operating crew response to complex system accidents— art 4. IDAC causal model of operator problem- solving response》 一文中在建立的操纵员问题索解的 IDAC 因果模型中认为, 影响行为计划 (action planning ) 的因素有: 注意力、 认知偏见、 时 间压力、 感知到的后果的严重度、 感知到的决策的责任、 感知到的任务的复杂性、 角 色与责任、 知识和经验、 对输入信息的记忆、 班组因素 (协调、 合作、 交流的有效性、 交流的质量、 班组构成、 领导) 以及人-机界面。 上述方法分析了影响响应计划的因素, 而没有有效的定量模型和数据来支持响应 计划可靠性的定量计算。 尽管像 CREAM等 HRA方法对响应计划的失误概率分析基 本上是先确定响应计划失误的基本失误概率, 再考虑行为形成因子 (Performance Shaping Factor,简称为 PSF)的状态对响应计划的失误概率进行修订,而没有考虑 PSF 之间交互作用, 从而可能带来重复计算其影响的可能, 对响应计算失误概率可能造成 错误的估计。 由此可见, 相关技术中存在的问题在于: 缺乏考虑数字化人-机系统的特征来研究 响应计划可靠性问题。 缺乏考虑 PSF因果关系的核电厂操纵员响应计划可靠性定量分 析技术。 缺乏数字化模拟机数据来支持响应计划的定量化。 发明内容 本发明实施例提供了一种响应计划的可靠性分析方法及装置, 以至少解决相关技 术中无法结合在响应计划的可靠性分析过程中所使用的多个 PSF并考虑 PSF的因果关 系对响应计划的可靠性进行分析的问题。 根据本发明实施例的一个方面, 提供了一种响应计划的可靠性分析方法。 根据本发明实施例的响应计划的可靠性分析方法包括: 确定在响应计划的可靠性 分析过程中待使用的多个行为形成因子 PSF、多个 PSF中各个 PSF之间的关联关系以 及各个 PSF与响应计划可靠性之间的关联关系; 根据确定的多个 PSF、 多个 PSF中各 个 PSF之间的关联关系以及各个 PSF与响应计划可靠性之间的关联关系建立响应计划 的可靠性分析模型, 对响应计划的可靠性进行分析。 优选地, 响应计划的可靠性分析模型为贝叶斯网络模型, 其中, 贝叶斯网络模型 中的每个网络节点分别对应一个 PSF或响应计划可靠性节点。 优选地,根据确定的多个 PSF、多个 PSF中各个 PSF之间的关联关系以及各个 PSF 与响应计划可靠性之间的关联关系建立响应计划的可靠性分析模型包括: 根据确定后 的各个 PSF之间的关联关系以及各个 PSF与响应计划可靠性之间的关联关系将多个 PSF中部分 PSF设置为贝叶斯网络模型的根节点 PSF,其中,根节点 PSF中的每个 PSF 处于不同状态的先验概率分布统计结果独立于多个 PSF中除该 PSF之外的其余 PSF 以及响应计划可靠性; 将贝叶斯网络模型中除根节点 PSF之外的其他部分节点设置为 子节点 PSF和响应计划可靠性节点,其中,子 PSF中的每个 PSF和响应计划可靠性处 于不同状态的条件概率分布统计结果依赖于根节点 PSF中的一个或多个 PSF, 和 /或, 依赖于其他部分 PSF中的一个或多个 PSF。 优选地, 对响应计划的可靠性进行分析包括: 统计根节点 PSF在预设周期内的每 个 PSF处于不同状态的先验概率分布以及确定根节点 PSF之外的子节点 PSF和响应计 划可靠性节点中的每个节点处于不同状态的条件概率分布; 根据统计出的根节点 PSF 处于不同状态的先验概率统计结果、各个 PSF之间的关联关系以及各个 PSF与响应计 划可靠性的关联关系,获取除根节点 PSF之外的子节点 PSF处于不同状态的条件概率 分布以及响应计划可靠性节点处于不同状态的条件概率分布, 依据贝叶斯网络的因果 分析对响应计划的可靠性进行评价。 优选地, 对响应计划的可靠性进行分析还包括: 将获取到的根节点 PSF处于不良 状态的先验概率统计结果与根节点 PSF处于不良状态的后验概率结果进行比较, 确定 对响应计划的可靠性影响最大的 PSF。 根据本发明实施例的另一方面, 提供了一种响应计划的可靠性分析装置。 根据本发明实施例的响应计划的可靠性分析装置包括: 确定模块, 用于确定在响 应计划的可靠性分析过程中待使用的多个行为形成因子 PSF、多个 PSF中各个 PSF之 间的关联关系以及各个 PSF与响应计划可靠性之间的关联关系; 分析模块, 用于根据 确定的多个 PSF、多个 PSF中各个 PSF之间的关联关系以及各个 PSF与响应计划可靠 性之间的关联关系建立响应计划的可靠性分析模型, 对响应计划的可靠性进行分析。 优选地, 响应计划的可靠性分析模型为贝叶斯网络模型, 其中, 贝叶斯网络模型 中的每个网络节点分别对应一个 PSF或响应计划可靠性节点。 优选地, 分析模块包括: 第一设置单元, 用于根据确定后的各个 PSF之间的关联 关系以及各个 PSF与响应计划可靠性之间的关联关系将多个 PSF中部分 PSF设置为贝 叶斯网络模型的根节点 PSF, 其中, 根节点 PSF中的每个 PSF的先验概率分布统计结 果独立于多个 PSF中除该 PSF之外的其余 PSF以及响应计划可靠性; 第二设置单元, 用于将贝叶斯网络模型中除根节点 PSF之外的其他部分节点设置为子节点 PSF和响应 计划可靠性节点, 其中, 子节点 PSF中的每个 PSF和响应计划可靠性的条件概率分布 统计结果依赖于根节点 PSF中的一个或多个 PSF,和 /或,依赖于根节点 PSF中的一个 或多个 PSF指向的一个或多个 PSF。 优选地, 分析模块还包括: 统计单元, 用于统计根节点 PSF在预设周期内的每个 PSF处于不同状态的先验概率分布以及确定除根节点 PSF之外的子节点 PSF和响应计 划可靠性节点中的每个节点处于不同状态的条件概率分布; 分析单元, 用于根据统计 出的根节点 PSF的概率统计结果以及各个 PSF之间的关联关系以及各个 PSF与响应计 划可靠性之间的关联关系获取除根节点 PSF之外的子节点 PSF处于不同状态的条件概 率分布以及响应计划可靠性节点处于不同状态的条件概率分布, 依据贝叶斯网络的因 果分析对响应计划的可靠性进行评价。 优选地, 分析模块还包括: 比较单元, 用于将获取到的根节点 PSF处于不良状态 的先验概率统计结果与根节点 PSF处于不良状态的后验概率结果进行比较, 确定对响 应计划的可靠性影响最大的 PSF。 通过本发明实施例,采用确定在响应计划的可靠性分析过程中待使用的多个 PSF、 所述多个 PSF中各个 PSF之间的关联关系以及每个 PSF与响应计划可靠性的关联关 系;根据确定的所述多个 PSF、所述多个 PSF中各个 PSF之间的关联关系以及每个 PSF 与响应计划可靠性的关联关系建立所述响应计划的可靠性分析模型, 对所述响应计划 的可靠性进行分析, 解决了相关技术中在响应计划的可靠性分析过程中没有考虑所使 用的多个 PSF之间的因果关系对响应计划的可靠性进行分析的问题, 进而为数字化主 控室操纵员响应计划可靠性分析提供定性与定量的方法与工具, 为电厂降低操纵员响 应计划失效概率提供对策。 附图说明 此处所说明的附图用来提供对本发明的进一步理解, 构成本申请的一部分, 本发 明的示意性实施例及其说明用于解释本发明, 并不构成对本发明的不当限定。 在附图 中: 图 1是根据本发明实施例的响应计划的可靠性分析方法的流程图; 图 2是根据本发明优选实施例的操纵员响应计划的贝叶斯网络模型的示意图; 图 3是根据本发明实施例的响应计划的可靠性分析装置的结构框图; 以及 图 4是根据本发明优选实施例的响应计划的可靠性分析装置的结构框图。 具体实 Ifi^式 下文中将参考附图并结合实施例来详细说明本发明。 需要说明的是, 在不冲突的 情况下, 本申请中的实施例及实施例中的特征可以相互组合。 图 1是根据本发明实施例的响应计划的可靠性分析方法的流程图。 如图 1所示, 该方法可以包括以下处理步骤: 步骤 S102:确定在响应计划的可靠性分析过程中待使用的多个行为形成因子 PSF、 多个 PSF中各个 PSF之间的关联关系以及各个 PSF与响应计划可靠性之间的关联关 系; 步骤 S104:根据确定的多个 PSF、多个 PSF中各个 PSF之间的关联关系以及各个
PSF与响应计划可靠性之间的关联关系建立响应计划的可靠性分析模型, 对响应计划 的可靠性进行分析。 相关技术中, 在响应计划的可靠性分析过程中没有考虑所使用的多个 PSF的因果 关系对响应计划的可靠性进行分析。 采用如图 1所示的方法, 确定在响应计划的可靠 性分析过程中待使用的多个 PSF、所述多个 PSF中各个 PSF之间的关联关系以及每个 PSF与响应计划可靠性的关联关系; 根据确定的所述多个 PSF、 所述多个 PSF中各个 PSF之间的关联关系以及每个 PSF与响应计划可靠性的关联关系建立所述响应计划的 可靠性分析模型, 对所述响应计划的可靠性进行分析, 解决了相关技术中在响应计划 的可靠性分析过程中没有考虑所使用的多个 PSF之间的因果关系对响应计划的可靠性 进行分析的问题, 进而为数字化主控室操纵员响应计划可靠性分析提供定性与定量的 方法与工具, 为电厂降低操纵员响应计划失效概率提供对策。 在优选实施例中, 基于数字化主控室操纵员的情境环境分析, 识别影响操纵员响 应计划可靠性的 PSF因子及其因果关系, 建立响应计划定性分析的贝叶斯网络模型, 为操纵员响应计划可靠性的量化奠定基础; 基于建立的响应计划的贝叶斯网络模型, 通过模拟机实验来收集网络节点的先验概率和条件概率, 对于难以收集的数据, 可采 用事件报告分析和专家判断来获取数据。 为确保数据和结果的有效性, 建立一种操纵 员响应计划可靠性评定的模糊贝叶斯方法, 以提高分析的精度。 需要说明的是, 如果没有合适的响应计划, 则需要重新构建新的响应计划。例如: 简单的操纵员响应计划可以包括: 对报警的响应, 当出现某个报警, 操纵员的行为直 接对应预设动作, 再如: 对应急规程的响应, 可以直接按照规程处理, 而无需选择路 径。 与简单的响应计划相比, 复杂的响应计划, 例如: 没有对应的规程、 程序和规则 可以遵循, 或者, 现有的规程已经被证实无法满足实际工作的需求, 同样需要操纵员 重新构建新的响应计划, 并对该响应计划的可行性与有效性进行评估。 在优选实施过程中, 上述响应计划的可靠性分析模型可以为贝叶斯网络模型, 其 中, 贝叶斯网络模型中的每个网络节点分别对应一个 PSF或响应计划可靠性节点。 贝叶斯网络 (BN) 是由节点和边组成的有向无环图 (Directed Acyclic Graph, 简 称为 DAG), 可以用 N=«V, E>, ?>来描述。 离散随机变量 V={Xi, X2, ..., Χημ 应的节点表示具有有限状态的变量, 节点可以是任何抽象的问题, 例如: 设备部件状 态、 测试值、 组织因素、 人的诊断结果等。 有向边 Ε表示节点间的概率因果关系, 有 向边的起始节点 i是终节点 j的父节点, j称为子节点, 没有父节点、 只有子节点的节 点称为根节点。 DAG蕴涵了一个条件独立假设: 给定其父节点集, 每一个变量独立于 它的非子孙节点。 P为定量部分, 是 V上的概率分布, 对于离散情况, 可用条件概率 表 (Conditional Probability Table, 简称为 CPT) 来表示, 用于定量说明父节点对子节 点的影响。 根节点的概率分布函数为边缘概率分布函数, 由于该类节点的概率不以其 它节点为条件, 故其概率为先验概率, 其它节点为条件概率分布函数。 优选地, 在步骤 S104中, 根据确定的多个 PSF、 多个 PSF中各个 PSF之间的关 联关系以及各个 PSF与响应计划可靠性之间的关联关系建立响应计划的可靠性分析模 型可以包括以下操作: 步骤 S1 : 根据确定后的各个 PSF之间的关联关系以及各个 PSF与响应计划可靠 性之间的关联关系将多个 PSF中部分 PSF设置为贝叶斯网络模型的根节点 PSF,其中, 根节点 PSF中的每个 PSF处于不同状态的先验概率分布统计结果独立于多个 PSF中除 该 PSF之外的其余 PSF以及响应计划可靠性; 步骤 S2: 将贝叶斯网络模型中除根节点 PSF之外的其他节点设置为子节点 PSF 和响应计划可靠性节点, 其中, 子节点 PSF中的每个 PSF和响应计划可靠性处于不同 状态的条件概率分布统计结果依赖于根节点 PSF中的一个或多个 PSF, 和 /或, 依赖于 其他部分 PSF中的一个或多个 PSF 。 优选地, 在步骤 S104中, 对响应计划的可靠性进行分析可以包括以下步骤: 步骤 S3: 统计根节点 PSF在预设周期内的每个 PSF处于不同状态的先验概率分 布以及确定根节点 PSF之外的子节点 PSF和响应计划可靠性节点处于不同状态的条件 概率分布; 步骤 S4: 根据各个 PSF之间的关联关系以及各个 PSF与响应计划可靠性的关联 关系,获取除根节点 PSF之外的子节点 PSF处于不同状态的条件概率分布以及响应计 划可靠性节点处于不同状态的条件概率分布, 依据贝叶斯网络的因果分析对响应计划 的可靠性进行评价。 在优选实施例中, 结合数据的获取和响应计划定量计算的模糊贝叶斯方法对上述 优选实施过程做进一步的描述。 第一部分、 数据采集
1 ) 基于模拟机实验的数据获取
( 1 ) 网络节点的先验概率分布。 选择典型的事故场景 (例如: SGTR、 LOCA、 全厂失电等) 进行实验, 对事故场景中关键点的任务所涉及的数字化人机界面、 数字 化规程、 任务的复杂性、 事故场景下的时间窗口、 交流水平、 培训水平等影响因子进 行评定, 识别主要影响因素的概率分布。 例如: 针对 SGTR事故的关键任务所涉及的 数字化人机界面按人机界面设计好坏的评定标准 (从信息搜集、 诊断和执行的容易度 方面) 各个画面进行评定, 获取概率分布 (假设共涉及 100副画面, 通过专家组的评 定, 得到 90幅画面满足预设条件, 8幅画面一般, 2幅画面差, 则得到人机界面的先 验概率分布为: 0.9, 0.08, 0.02, 同理可得其他影响因素的先验概率分布)。
(2)网络节点的条件概率分布。在实验过程中对操纵员的知识和经验、压力水平、 态度等进行评定。 这需要被测试者针对每完成一个关键的任务进行自我评定, 其要求 实事求是。 统计评定结果, 得到知识和经验(假设有三个水平: 好、 中、 差)、 压力水 平 (假设有三个水平: 好、 中、 差)、 任务的复杂性 (假设有三个水平: 好、 中、 差) 等的条件概率分布。 同时通过响应计划的实验结果的统计分布, 得到响应计划可靠性 的条件概率分布。 例如: 培训和交流水平影响操纵员的知识和经验, 则可以选择不同培训水平的人 员进行实验, 其中, 可以包括: 培训水平好、 中、 差以及交流水平一般的情况下分别 进行实验, 得到培训水平好的一组操纵员的实验结果 (需操纵员对其获取的知识和经 验进行评定)、培训水平中等与交流水平一般的情况下的实验结果(需操纵员对其获取 的知识和经验进行评定) 以及培训水平差和交流一般的实验结果 (需操纵员对其获取 的知识和经验进行评定), 从而得到知识和经验的一部分的条件概率分布, 同理, 控制 好交流不同水平的实验变量, 可得到所有知识和经验的条件概率分布。 控制好其他可 以控制的变量, 可得到任务的复杂性等节点变量的条件概率分布, 如果对于难以进行 实验或者难以进行控制的变量, 可采用专家判断的方法或回归技术等进行建模估计。 最终根据响应计划可靠性的测量结果, 得到响应计划可靠性的条件概率分布。
2) 基于专家判断的数据获取 对于难以进行实验或者难以进行控制的变量, 例如: 安全文化、 管理水平等节点 变量, 可以采用专家判断的方法来获取数据。 由于因素状态等级评定的复杂性和不确 定性以及专家知识、 能力、 经验的有限性, 使某些专家难以确定因素状态等级的确切 值, 由此导致专家可能用描述性语言或者用范围值来表达, 例如: "大约 7"、 "很可能 在 5-7这个范围"、 "(3, 5, 7)"等。 并且决策者认为, 模糊判断比确切值判断更可信, 更符合人们的真实思维, 因此, 本发明提出通过模糊方法对 PSF因子处于不同状态的 概率分布进行评价, 其评价程序如下:
( 1 )组建专家组。不同的专家由于知识背景和经验不同对组织因素的评价结果也 有所不同, 从而影响决策结果, 因此, 需组建专家组来消除上述影响, 并且每个专家 分配不同的权重。 假设有 m个专家组成的专家组, 且第 i个专家赋予的权重为
Ci , c; e [0,1] , f i = l。
i二 1 (2) 确定 PSF处于不同状态的概率 通过专家讨论将每个因素通常处于何种状态的概率可采用三角模糊数 (即最有可 能的值、 最好的值、 最差的值) 对组织因素进行评价, 例如: (0.1, 0.3, 0.6), 或者, 用描述性语言来表示, 例如: 高、 中、 低。 对于描述性语文可引入模糊隶属函数来确 定因素处于特定状态的概率值。 (3 ) 计算各因素的综合概率分布值并解模糊 依据专家权重和相应的因素状态概率, 可计算各因素的状态概率分布:
( 1 ) s,是因素 的模糊综合得分,它是一个三角模糊数:( l
Figure imgf000011_0001
xck)
二 二 二 为了将综合的三角模糊数转化为确切值, 可通过三角形重心解模糊的方法求解:
F =(U「 1 φΓ1 ^ …… (2)
1 3 1 其中, 表示最大可能值, A表示最可能值, 表示最小可能值。 第二部分、 贝叶斯网络的分析 链式法则表明一个 BN就是在 DAG中所有变量的联合分布的一种描述,并且网络 中每个节点的边缘概率和条件概率均可计算。 贝叶斯网络的分析原理是基于 Bayes概率理论, 分析过程实质上就是概率计算过 程。 主要根据下列三个方程进行分析计算: 联合概率 p{x^ ,xn}:
Figure imgf000011_0002
…… (3)
i二 为 ,父节点的集合。 ^的边缘概率为:
P(XI)= ∑P(U) …… (4)
; 贝叶斯网络的主要应用在于一个用于计算事件信念的推理机, 其任务是计算 "在 给定的证据 (或观察数据) 的条件下, 某些事件的发生概率"。 假设已知证据 e, 则有: p(U e) = = ^ C5
' P(e) ∑p(U,e)
u
1) 因果分析 因果分析由原因推知结论, 是一种自顶向下的分析。 在给定的原因或证据的条件 下, 使用贝叶斯网络分析计算, 求解结果发生的概率。 在正常情况下, 即各变量服从 专家组判断和基于模拟机实验数据得到的初始概率分布, 例如: 网络中的根节点的先 验概率可用表 1来表示。 假设得到的班组的交流与合作水平的处于不同状态的模糊先 验概率为 (0.09, 0.10, 0.11)、 (0.29, 0.30, 0.31)、 (0.59, 0.60, 0.61), 分别解模糊后 得到班组交流与合作水平处于不充分状态、可接受状态和充分状态的概率分布为(0.1, 0.3, 0.6), 同理可通过专家判断或者模拟机实验得到其他节点变量的概率分布。 表 1是根据本发明优选实施例的根节点的模糊先验概率统计表。 如表 1所示, 表 1
Figure imgf000012_0001
同理可得, 表 2是根据本发明优选实施例的子节点的条件概率信息表。子节点 "知 识和经验 KE"的条件概率 P (知识和经验 I班组交流合作水平,培训水平),如表 2所示: 表 2
Figure imgf000013_0002
则班组的交流与合作水平、 培训水平引起操纵员的知识和经验处于 "低"水平状 态的概率 (用 P ( KE=KE,!) 表示) 可根据公式 (4), 具体如下:
P(KE=KE,1)=P(CO=CO,I)X[P(TR=TR,1)XP(KE=KE,1ICO=C0,I,TR=TR,1)+P(TR=TR,2)XP(KE =KE,1ICO=CO,I,TR=TR,2)+P(TR=TR,3)XP(KE=KE,1ICO=CO,I,TR=TR,3)]+P(CO=CO,2)X [P(TR=TR,1 )XP(KE=KE,1ICO=CO,2,TR=TR,I)+P(TR=TR,2)XP(KE=KE,1ICO=CO,2,TR=TR,2)+P(TR=TR,3)XP(KE =KE,1ICO=CO,2,TR=TR,3)]+P(CO=CO,3)X[P(TR=TR,1)XP(KE=KE,1ICO=CO,3,TR=TR,I)+P(TR=TR,2 )XP(KE=KE,1ICO=CO,3,TR=TR,2)+P(TR=TR,3)XP(KE=KE,1ICO=CO,3,TR=TR,3)] 将获得的数据代入公式即可得出
Figure imgf000013_0001
同样可计算得出 ρ(κΕΕ,2)和 Ρ(ΚΕΕ,3)。 由此得到子节点 "知识和经验"的概率处于不同状态的概率分布。 同理可 计算得到其他节点变量的概率分布。 最终计算得到响应计划可靠性 Ρ »Ή
2 ) 诊断分析 诊断分析是由结论推知原因, 是一种自底向上的分析过程, 其目的是在已经获知 结果的情况下, 找寻产生该结果的各种原因的可能性。 已知发生了某种结果, 根据贝 叶斯网络计算, 得到造成该结果发生的原因和发生的概率。 在响应计划可靠性的模糊 贝叶斯网络模型中, 假设已发生响应计划失误, 则利用贝叶斯法则可计算出相应的后 验概率。 例如: 要计算 "班组交流与合作水平"处于 "不充分"状态的概率, 则根据 公式 (5 ) 可以得出: p(c0 =c01k = fiR 2) = P(Co = c^RR = RR'2)
02 Pd R^) 其中, = 1^2表示发生响应计划失误。 由公式(3 )可以计算出 P (C。 =。。,1 , 1^ = 1^2),由公式(4)可以计算出15 (1^ = 1^2 ), 从而可以获取需要计算的数值。 优选地, 在步骤 S104中, 对响应计划的可靠性进行分析还可以包括以下处理: 步骤 S5:将获取到的根节点 PSF处于不良状态的先验概率统计结果与根节点 PSF 处于不良状态的后验概率结果进行比较, 确定对响应计划的可靠性影响最大的 PSF。 识别最有可能引发响应计划失误的影响因素, 为失误的预防提供决策支持。 下面结合图 2所示的优选实施方式对上述优选实施过程做进一步的描述。 图 2是根据本发明优选实施例的操纵员响应计划的贝叶斯网络模型的示意图。 如 图 2所示, 响应计划的可靠性主要受一线操纵员的心理状态、 记忆中的信息以及个性 固有属性的影响。 操纵员的知识和经验丰富, 则会认识到特定的电厂状态对应该采取 何种响应策略或计划。 知识和经验主要受组织培训和班组的交流与合作的影响; 如果 培训不够, 则会对操纵员的知识和经验造成消极影响, 班组的交流与合作可以弥补操 纵员个体的知识和经验的不足。 此外, 压力水平对响应计划的制定也会有很大影响, 压力水平主要受事件的严重度、 任务的复杂性及可用时间的影响, 同样任务的复杂性 主要受数字化规程设计的好坏与数字化人-机界面设计的好坏的影响,规程中的任务复 杂则操纵员需要完成的任务复杂, 规程或程序设计的完美有利于指导操纵员做出响应 计划,人 -机界面的设计存在缺陷,则操纵员难以获取有利于响应计划制定的有用信息。 再者, 响应计划还要受到操纵员态度的影响, 操纵员的态度好、 责任心强, 则难以违 规, 注意力集中, 其中, 操纵员的态度主要受组织的安全文化和管理好坏的影响, 如 果安全文化没有深入人心, 则操纵员的风险意识淡薄、安全态度较差。通过上述分析, 响应计划可以受到班组的交流与合作水平、培训水平、数字化规程、数字化人机界面、 事件的严重度、 事故处置的可用时间、 安全文化与组织管理水平等因素的影响。 图 3是根据本发明实施例的响应计划的可靠性分析装置的结构框图。如图 3所示, 该响应计划的可靠性分析装置可以包括: 确定模块 10, 用于确定在响应计划的可靠性 分析过程中待使用的多个形成因子 PSF、所述多个 PSF中各个 PSF之间的关联关系以 及各个 PSF与响应计划可靠性之间的关联关系; 分析模块 20, 用于根据确定的所述多 个 PSF、所述多个 PSF中各个 PSF之间的关联关系以及各个 PSF与响应计划可靠性之 间的关联关系建立所述响应计划的可靠性分析模型, 对所述响应计划的可靠性进行分 析。 采用如图 3所示的装置, 解决了相关技术中在响应计划的可靠性分析过程中没有 考虑所使用的多个 PSF的因果关系对响应计划的可靠性进行分析的问题, 进而为数字 化主控室操纵员响应计划可靠性分析提供定性与定量的方法与工具, 为电厂降低操纵 员响应计划失效概率提供对策。 在优选实施过程中, 上述响应计划的可靠性分析模型可以为贝叶斯网络模型, 其 中, 贝叶斯网络模型中的每个网络节点分别对应一个 PSF或响应计划可靠性节点。 优选地, 如图 4所示, 上述分析模块 20可以包括: 第一设置单元 200, 用于根据 确定后的各个 PSF之间的关联关系以及各个 PSF与响应计划可靠性之间的关联关系将 多个 PSF中部分 PSF设置为贝叶斯网络模型的根节点 PSF, 其中, 根节点 PSF中的每 个 PSF的先验概率分布统计结果独立于多个 PSF中除该 PSF之外的其余 PSF以及响 应计划可靠性; 第二设置单元 202, 用于将贝叶斯网络模型中除根节点 PSF之外的其 他节点设置为贝叶斯网络模型的子节点 PSF和响应计划可靠性节点,其中,子节点 PSF 和响应计划可靠性处于不同状态的条件概率分布统计结果依赖于根节点 PSF中的一个 或多个 PSF, 和 /或, 依赖于根节点 PSF中的一个或多个 PSF指向的一个或多个 PSF。 优选地, 如图 4所示, 上述分析模块 20还可以包括: 统计单元 204, 用于统计根 节点 PSF在预设周期内的每个 PSF处于不同状态的先验概率分布以及确定除根节点 PSF之外的子节点 PSF和响应计划可靠性节点中的每个节点处于不同状态的条件概率 分布; 分析单元 206, 用于根据确定的各个 PSF之间的关联关系以及各个 PSF与响应 计划可靠性之间的关联关系获取除根节点 PSF之外的子节点 PSF处于不同状态的条件 概率分布以及响应计划可靠性节点处于不同状态的条件概率分布, 依据贝叶斯网络的 因果分析对响应计划的可靠性进行评价。 优选地, 如图 4所示, 上述分析模块 20还可以包括: 比较单元 208, 用于获取到 的根节点 PSF处于不良状态的先验概率统计结果与根节点 PSF处于不良状态的后验概 率结果进行比较, 确定对响应计划的可靠性影响最大的 PSF。 从以上的描述中, 可以看出, 上述实施例实现了如下技术效果 (需要说明的是这 些效果是某些优选实施例可以达到的效果): 采用本发明所提供的技术方案, 为数字化 主控室操纵员响应计划可靠性分析提供定性与定量的方法与工具, 为电厂降低操纵员 响应计划失效概率提供对策; 为核电厂数字化主控室操纵员可靠性分析 (HRA) 与概 率安全评价(PSA)提供操纵员响应计划可靠性接口数据与计算工具,建立的考虑 PSF 因果关系的响应计划可靠性计算的模糊贝叶斯方法, 可以提高 HRA和 PSA分析的精 度; 为核电厂数字化主控室操纵员的人因失误预防、 培训与场景开发提供支持; 为数 字化工业系统中主控室作业人员响应计划或者决策行为的可靠性分析与安全评估提供 技术支持与工具。 显然, 本领域的技术人员应该明白, 上述的本发明的各模块或各步骤可以用通用 的计算装置来实现, 它们可以集中在单个的计算装置上, 或者分布在多个计算装置所 组成的网络上, 可选地, 它们可以用计算装置可执行的程序代码来实现, 从而, 可以 将它们存储在存储装置中由计算装置来执行, 并且在某些情况下, 可以以不同于此处 的顺序执行所示出或描述的步骤, 或者将它们分别制作成各个集成电路模块, 或者将 它们中的多个模块或步骤制作成单个集成电路模块来实现。 这样, 本发明不限制于任 何特定的硬件和软件结合。 以上所述仅为本发明的优选实施例而已, 并不用于限制本发明, 对于本领域的技 术人员来说, 本发明可以有各种更改和变化。 凡在本发明的精神和原则之内, 所作的 任何修改、 等同替换、 改进等, 均应包含在本发明的保护范围之内。

Claims

权 利 要 求 书
1. 一种响应计划的可靠性分析方法, 包括:
确定在响应计划的可靠性分析过程中待使用的多个行为形成因子 PSF、 所 述多个 PSF中各个 PSF之间的关联关系以及所述各个 PSF与响应计划可靠性之 间的关联关系;
根据确定的所述多个 PSF、所述多个 PSF中各个 PSF之间的关联关系以及 所述各个 PSF与所述响应计划可靠性之间的关联关系建立所述响应计划的可靠 性分析模型, 对所述响应计划的可靠性进行分析。
2. 根据权利要求 1所述的方法, 其中, 所述响应计划的可靠性分析模型为贝叶斯 网络模型, 其中, 所述贝叶斯网络模型中的每个网络节点分别对应一个 PSF或 响应计划可靠性节点。
3. 根据权利要求 2所述的方法, 其中, 根据确定的所述多个 PSF、 所述多个 PSF 中各个 PSF之间的关联关系以及所述各个 PSF与所述响应计划可靠性之间的关 联关系建立所述响应计划的可靠性分析模型包括:
根据确定后的所述各个 PSF之间的关联关系以及所述各个 PSF与所述响应 计划可靠性之间的关联关系将所述多个 PSF中部分 PSF设置为所述贝叶斯网络 模型的根节点 PSF, 其中, 所述根节点为没有被其他节点指向的节点, 所述根 节点 PSF中的每个 PSF处于不同状态的先验概率分布统计结果独立于所述多个 PSF中除该 PSF之外的其余 PSF和响应计划可靠性;
将所述多个 PSF中除所述根节点 PSF之外的其他部分 PSF设置为所述贝叶 斯网络模型的子节点 PSF, 其中, 所述子节点 PSF为被其他节点指向的 PSF, 所述子节点 PSF中的每个 PSF和响应计划可靠性节点处于不同状态的条件概率 分布统计结果依赖于所述根节点 PSF中的一个或多个 PSF, 和 /或, 依赖于所述 其他部分 PSF中的一个或多个 PSF。
4. 根据权利要求 3所述的方法, 其中, 对所述响应计划的可靠性进行分析包括: 统计所述根节点 PSF在预设周期内的每个 PSF处于不同状态的先验概率分 布以及确定所述根节点 PSF之外的子节点 PSF和响应计划可靠性节点中的每个 节点处于不同状态的条件概率分布; 根据统计出的所述根节点 PSF处于不同状态的先验概率统计结果、所述各 个 PSF之间的关联关系以及所述各个 PSF与所述响应计划可靠性的关联关系, 获取除所述根节点 PSF之外的各子节点 PSF 处于不同状态的条件概率分布以 及响应计划可靠性节点 PSF处于不同状态的条件概率分布, 依据贝叶斯网络的 因果分析对所述响应计划的可靠性进行评价。
5. 根据权利要求 4所述的方法,其中,对所述响应计划的可靠性进行分析还包括:
将获取到的所述根节点 PSF处于不良状态的先验概率统计结果与所述根节 点 PSF处于不良状态的后验概率结果进行比较, 确定对所述响应计划的可靠性 影响最大的 PSF。
6. 一种响应计划的可靠性分析装置, 包括:
确定模块, 用于确定在响应计划的可靠性分析过程中待使用的多个行为形 成因子 PSF、所述多个 PSF中各个 PSF之间的关联关系以及所述各个 PSF与响 应计划可靠性之间的关联关系;
分析模块, 用于根据确定的所述多个 PSF、所述多个 PSF中各个 PSF之间 的关联关系以及所述各个 PSF与所述响应计划可靠性之间的关联关系建立所述 响应计划的可靠性分析模型, 对所述响应计划的可靠性进行分析。
7. 根据权利要求 6所述的装置, 其中, 所述响应计划的可靠性分析模型为贝叶斯 网络模型, 其中, 所述贝叶斯网络模型中的每个网络节点分别对应一个 PSF或 响应计划可靠性节点。
8. 根据权利要求 7所述的装置, 其中, 所述分析模块包括: 第一设置单元, 用于根据确定后的所述各个 PSF之间的关联关系以及所述 各个 PSF与所述响应计划可靠性之间的关联关系将所述多个 PSF中部分 PSF 设置为所述贝叶斯网络模型的根节点 PSF, 其中, 子节点 PSF为被其他节点指 向的 PSF,所述根节点 PSF中的每个 PSF的先验概率分布统计结果独立于所述 多个 PSF中除该 PSF之外的其余 PSF和响应计划可靠性;
第二设置单元, 用于将所述贝叶斯网络模型中除根节点 PSF之外的其他部 分节点设置为子节点 PSF和响应计划可靠性节点, 其中, 所述子节点 PSF中的 每个 PSF和响应计划可靠性节点处于不同状态的条件概率分布统计结果依赖于 所述根节点 PSF中的一个或多个 PSF, 和 /或, 依赖于所述根节点 PSF中的一 个或多个 PSF指向的一个或多个 PSF。 根据权利要求 8所述的装置, 其中, 所述分析模块还包括:
统计单元,用于统计所述根节点 PSF在预设周期内的每个 PSF处于不同状 态的先验概率分布以及确定所述除根节点 PSF之外的子节点 PSF和响应计划可 靠性节点中的每个节点处于不同状态的条件概率分布;
分析单元, 用于根据统计出的所述根节点 PSF的概率统计结果以及所述各 个 PSF之间的关联关系以及所述各个 PSF与响应计划可靠性之间的关联关系获 取所述除根节点 PSF之外的子节点 PSF处于不同状态的条件概率分布以及响应 计划可靠性节点处于不同状态的条件概率分布, 依据贝叶斯网络的因果分析对 所述响应计划的可靠性进行评价。 根据权利要求 9所述的装置, 其中, 所述分析模块还包括:
比较单元, 用于将获取到的所述根节点 PSF处于不良状态的先验概率统计 结果与所述根节点 PSF处于不良状态的后验概率结果进行比较, 确定对所述响 应计划的可靠性影响最大的 PSF。
PCT/CN2014/075730 2013-04-23 2014-04-18 响应计划的可靠性分析方法及装置 WO2014173258A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2013101426007A CN103198438A (zh) 2013-04-23 2013-04-23 响应计划的可靠性分析方法及装置
CN201310142600.7 2013-04-23

Publications (1)

Publication Number Publication Date
WO2014173258A1 true WO2014173258A1 (zh) 2014-10-30

Family

ID=48720955

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2014/075730 WO2014173258A1 (zh) 2013-04-23 2014-04-18 响应计划的可靠性分析方法及装置

Country Status (2)

Country Link
CN (1) CN103198438A (zh)
WO (1) WO2014173258A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109345121A (zh) * 2018-10-09 2019-02-15 湖南工学院 一种用于核电厂运行事件分析的方法和装置

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103218689B (zh) * 2013-04-23 2017-09-22 湖南工学院 操作员状态评估的可靠性分析方法及装置
CN103198438A (zh) * 2013-04-23 2013-07-10 湖南工学院 响应计划的可靠性分析方法及装置
CN105989435B (zh) * 2015-02-06 2020-01-07 中国石油天然气股份有限公司 一种基于rcm理论估算设备维护周期的方法
CN109447494A (zh) * 2018-11-08 2019-03-08 南华大学 核电厂班组情景意识可靠性评估方法、装置、设备及介质
CN109447495A (zh) * 2018-11-08 2019-03-08 南华大学 核电厂组织行为可靠性评估方法、装置及设备
CN110188973A (zh) * 2019-03-26 2019-08-30 苏州热工研究院有限公司 一种核电站火灾安全综合评估方法
CN113191084B (zh) * 2021-05-07 2024-04-19 哈尔滨工业大学 基于贝叶斯网络理论的既有铁路桥梁可靠度综合评估方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060004682A1 (en) * 2004-06-30 2006-01-05 Mats Danielson Method for decision and risk analysis in probabilistic and multiple criteria situations
CN101915234A (zh) * 2010-07-16 2010-12-15 西安交通大学 一种基于贝叶斯网络的压缩机关联故障诊断方法
CN103198438A (zh) * 2013-04-23 2013-07-10 湖南工学院 响应计划的可靠性分析方法及装置
CN103218689A (zh) * 2013-04-23 2013-07-24 湖南工学院 操作员状态评估的可靠性分析方法及装置

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060004682A1 (en) * 2004-06-30 2006-01-05 Mats Danielson Method for decision and risk analysis in probabilistic and multiple criteria situations
CN101915234A (zh) * 2010-07-16 2010-12-15 西安交通大学 一种基于贝叶斯网络的压缩机关联故障诊断方法
CN103198438A (zh) * 2013-04-23 2013-07-10 湖南工学院 响应计划的可靠性分析方法及装置
CN103218689A (zh) * 2013-04-23 2013-07-24 湖南工学院 操作员状态评估的可靠性分析方法及装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109345121A (zh) * 2018-10-09 2019-02-15 湖南工学院 一种用于核电厂运行事件分析的方法和装置
CN109345121B (zh) * 2018-10-09 2021-12-07 湖南工学院 一种用于核电厂运行事件分析的方法和装置

Also Published As

Publication number Publication date
CN103198438A (zh) 2013-07-10

Similar Documents

Publication Publication Date Title
WO2014173258A1 (zh) 响应计划的可靠性分析方法及装置
WO2014173257A1 (zh) 操作员状态评估的可靠性分析方法及装置
Zwirglmaier et al. Capturing cognitive causal paths in human reliability analysis with Bayesian network models
KR102118670B1 (ko) Ict 인프라 관리 시스템 및 이를 이용한 ict 인프라 관리 방법
CN110942086B (zh) 数据预测优化方法、装置、设备及可读存储介质
US9799007B2 (en) Method of collaborative software development
Li et al. An assessment method of operator’s situation awareness reliability based on fuzzy logic-AHP
US20170169342A1 (en) System and method for diagnosing at least one component requiring maintenance in an appliance and/or installation
Di Pasquale et al. Methodology for the analysis and quantification of human error probability in manufacturing systems
Nan et al. Developing an agent-based hierarchical modeling approach to assess human performance of infrastructure systems
CN112379325A (zh) 一种用于智能电表的故障诊断方法及系统
Li et al. Study on operator’s SA reliability in digital NPPs. Part 3: A quantitative assessment method
CN103198232A (zh) 数字化主控室工作人员人因可靠性的确定方法及装置
Liu et al. Methodology for dynamic reliability assessment of team situation awareness of digital nuclear power plants
Zahedi et al. Modeling the interplay between human trust and monitoring
Perkusich et al. A Bayesian network approach to assist on the interpretation of software metrics
Zhang et al. The simulator experimental study on the operator reliability of Qinshan nuclear power plant
Jang et al. A proactive alarm reduction method and its human factors validation test for a main control room for SMART
KR20150081839A (ko) 사이버-물리 시스템의 목표 평가 장치 및 방법
Rozaki Design and implementation for automated network troubleshooting using data mining
Raderius et al. Assessing system availability using an enterprise architecture analysis approach
Groth et al. Development and use of a Bayesian network to estimate human error probability
MÉRAND et al. ARGUinDSM–A model of argumentation in team problem-solving situation: An application to nuclear control room teams
Angelopoulou et al. The system dynamics architecture of UTASiMo: a simulation-based task analysis tool to predict human error probability
CN109636164B (zh) 一种用于工业系统的人因安全评估方法、装置及存储介质

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14787818

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 14787818

Country of ref document: EP

Kind code of ref document: A1