WO2019076026A1 - 一种扩展的处理不确定因果关系类信息的智能系统的构造方法 - Google Patents

一种扩展的处理不确定因果关系类信息的智能系统的构造方法 Download PDF

Info

Publication number
WO2019076026A1
WO2019076026A1 PCT/CN2018/085539 CN2018085539W WO2019076026A1 WO 2019076026 A1 WO2019076026 A1 WO 2019076026A1 CN 2018085539 W CN2018085539 W CN 2018085539W WO 2019076026 A1 WO2019076026 A1 WO 2019076026A1
Authority
WO
WIPO (PCT)
Prior art keywords
event
variable
state
combination
degree
Prior art date
Application number
PCT/CN2018/085539
Other languages
English (en)
French (fr)
Inventor
张勤
张湛
Original Assignee
北京清睿智能科技有限公司
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 北京清睿智能科技有限公司 filed Critical 北京清睿智能科技有限公司
Publication of WO2019076026A1 publication Critical patent/WO2019076026A1/zh
Priority to US16/841,065 priority Critical patent/US20200234169A1/en

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00Computing arrangements based on specific mathematical models
    • G06N7/02Computing arrangements based on specific mathematical models using fuzzy logic
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/02Knowledge representation; Symbolic representation
    • G06N5/022Knowledge engineering; Knowledge acquisition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/04Inference or reasoning models
    • G06N5/041Abduction
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00Computing arrangements based on specific mathematical models
    • G06N7/01Probabilistic graphical models, e.g. probabilistic networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/04Inference or reasoning models
    • G06N5/045Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence

Definitions

  • the invention relates to an intelligent information processing technology, and is an authorized patent "a construction method for processing only uncertain causality information” (patent number: ZL 2006 8 0055266.X), "METHOD FOR CONSTRUCTING AN INLLIGENT SYSTEM” PROCESSING UNCERTAIN CAUSAL RELATIONSHIP INFORMATION (Patent No.: US 8,255,353 B2) and the technical solution described in "Method for Abnormality of Heuristic Detection System Based on Dynamic Uncertain Causal Graph” (Patent No.: ZL 2016 1 0282052.1) Further expansion.
  • the causal knowledge expression and operation ability called Dynamic Uncertain Causality Graph can be further improved by computer processing, so that it can more satisfy the actual needs and be more convenient and accurate for diagnosis.
  • the cause of the abnormality of the object system is beneficial to people taking effective measures in case of abnormality of the object system, so that they can return to normal as soon as possible.
  • B k , B kj , BX k and BX kj can be graphically used respectively It is shown that the difference between B k and BX k is that B k is the root cause variable, there is no input, and BX k has input, which can be affected by other factors, and BX k is the affected B k .
  • kj can be separated by commas (the same below).
  • the state of B k and BX k cannot be directly detected in most cases, or is difficult to detect directly. It is a variable that the DUCG intelligent system needs to infer whether it is in an abnormal state.
  • the logic gate variable which can be represented by G i , has at least two input variables and one output variable.
  • G ij represents the j state of G i .
  • G i is used to express a logical combination of various states of interest of the input variables, and these logical combinations are illustrated by the logic gate description table LGS i .
  • G ij G ij' 0 (empty set, where j ⁇ j'), that is, the different states of G are mutually exclusive.
  • G i and G ij can be respectively used with graphical symbols or Said.
  • the default reason variable which can be represented by D, is one of the reason variables of the corresponding X variable.
  • D 4 is the default reason variable for X 4 .
  • D i can be used graphics Said.
  • the B, X, BX, D, and G variables can be referred to as nodes, and the physical meaning of the variables themselves and their various states can be defined according to the described objects.
  • B, X, BX, D, and G are direct cause variables, they are called parent variables, and can be uniformly expressed by V, V ⁇ B, X, BX, D, G ⁇ , and the subscripts are unchanged.
  • the resulting variable can only be an X or BX variable.
  • a state of a variable is an event. For example, X yg , B kj , BX kj , G ij , H kj or V ij are all events.
  • DUCG is composed of the above variables and the causal relationship between variables or uncertainties, usually represented by graphical symbols.
  • a DUCG example is shown in Figure 1, where B variables or events are represented by rectangles; X variables or events are represented by circles; BX variables or events are represented by double circles; G variables or events are represented by logic gates, and directed arcs are used
  • the input variable is connected to G i ; the D variable is represented by a pentagon.
  • the action variable F n;i is a causal relationship matrix between the parent variable V i and the child variable X n or BX n , which can be used with a directed arc Or other graphical symbol representation, from the cause to the result, the elements in the matrix are called the action event F nk; ij , the causal relationship between the parent event V ij and the sub-event X nk or BX nk , F nk; i express the parent variable The causal relationship between V i and the sub-event X nk , F n; ij expresses the causal relationship between the parent event V ij and the sub-variant X n or BX n .
  • f nk; ij is the contribution of V ij to the weight probability of X nk , satisfying F nk; ij , f nk; ij , A nk; ij , a nk; ij are elements of the matrix F n; i , f n; i , A n; i , a n; i , respectively.
  • v ij Pr ⁇ V ij ⁇ , v ⁇ ⁇ b, x, bx, d, g ⁇ , and V ij and v ij are elements of the event vector V i and the parameter vector v i , respectively.
  • the expression of other causal variables is similar.
  • F nk; ij can be a conditional event, with a dotted arc Said.
  • the conditional action event expresses a conditional relationship between the cause event V ij and the result event X nk or BX nk , that is, whether F nk; ij is established according to whether the conditional event Z nk; ij is satisfied.
  • Z nk; ij X 1,2 , when X 1,2 is true, Z nk; ij satisfies, F nk; ij holds; when X 1, 2 is not true, Z nk; ij does not satisfy, F nk ; ij does not hold.
  • Rule 8 When a directed arc has no parent or no children, remove the directed arc from the DUCG.
  • Rule 12 The above rules may be used, used in combination, and reused in any order.
  • the simplified DUCG diagram it can be further assumed that when one of the B variables is established, the other B variables are not established, thereby decomposing the simplification map into individual sub-simplification diagrams, and further simplifying according to the above-mentioned simplification rules.
  • the B or BX variables After some subgraphs are simplified, the B or BX variables have no evidence of downstream anomalies, and the subgraphs are deleted.
  • the abnormal state of the B or BX variable in the remaining subgraph constitutes the possible cause event space S H of the system exception.
  • the normalized sorting probability can be found: according to with People can know the possible causes and ordering of system anomalies, so as to take correct countermeasures and make the system return to normal as soon as possible.
  • ⁇ k is independent of the subscript j, that is, independent of the abnormal state of the root cause variable. But in fact, the degree of attention of different abnormal states of the root cause variable should be different.
  • the present invention proposes an expanded technical solution to solve the above problems.
  • US invention patent "Method for constructing an intelligent system processing uncertain causal relationship information"; patent number: US 8255353 B2; authorization date: August 28, 2012; rights holder: Zhang Zhan; inventor: Zhang Qin, Zhang Zhan.
  • Chinese invention patent “A method for the abnormal cause of heuristic detection system based on dynamic uncertain causal map”; patent number: ZL 2016 1 0282052.1; authorization date: February 2015; right holder: Beijing Qingrui Intelligent Technology Co., Ltd.; inventor: Zhang Qin.
  • the present invention discloses a technical solution, and further expands the DUCG technical solution disclosed in the Chinese patents ZL 2006 8 0055266.X, ZL 2016 1 0282052.1, US Patent No. 8255353 B2 and the above documents.
  • a computer readable storage medium characterized in that: the computer medium program is stored on the storage medium, and when the computer program is executed, an intelligent system that processes uncertain causal relationship information can be executed as follows
  • the construction and reasoning method based on the existing DUCG technical scheme, adds a method for expressing and inferring the abnormal cause B k of the target system.
  • the added content includes: (1) using a new logic gate SG k and new The action variable SA k;k expresses the direct influence of the evidence X yg and its combination on the probability of occurrence of B k states.
  • the affected B k is represented by BX k , X y and B k are the inputs of SG k , and the event matrix SA k;k is the output of SG k , pointing to BX k ,SA k; the member event of k is SA kj; kn ; (2) using reverse logic gate RG i to express the state of each cause variable and the state of more than one result variable
  • the logical relationship between the combinations, and the state of the reverse logic gate is determined according to the meaningful combination of state evidence of the result variable, and the DUCG reasoning is performed according to it; (3) the variable SX y is used to express an abnormality with the specific B variable.
  • the state of the X-type variable state is more helpful.
  • the inverse logic gate RG i may be represented by a graphical symbol having at least one input variable connected by an F-type directed arc pointing from the input variable to RG i
  • RG in is the state of RG i labeled n, representing the state combination of the output variable labeled n, with events Combined n expression
  • Inference, RG in is DUCO logic expansion as an X type event
  • J k is the number of abnormal states of B k
  • S iG (y) is the set of all or part of the relevant state subscripts of X ig at time y.
  • Figure 3 shows the new DUCG diagram in Example 1
  • Figure 5 shows the situation after receiving evidence in Example 2.
  • FIG. 6 shows the results of Figure 5 in accordance with claims 2-9);
  • FIG. 15 Another example of a reverse logic gate
  • SG k uses graphical symbols Expression, SG kj with graphical symbols Expression, its input is used with a directed arc Expression, SA k; k with directed arc expression.
  • SA k a directed arc Expression
  • B kj and BX kj j ⁇ ⁇ 0, 1, 2 ⁇ .
  • X 1, X 2 and X 3 are B k constitute the input variables SG k of
  • BX k is the SG k of the output variables, by the SA k; k is connected, in the FIG. 3
  • LGS k Table 1
  • SA k;k and the parameters of B k are embedded.
  • B k and SG k0 and their input and output directed arcs are deleted, as shown in Fig. 6.
  • X 1,0 be the normal state of X 1 .
  • B k is deleted, that is, B k
  • the abnormal state does not exist and no further calculations are required.
  • the state probability distribution of BX k is exactly the same as B k .
  • B k is exactly equivalent to BX k , so B k k can be replaced by B k . That is, FIG. 9 can be further reduced to FIG.
  • E causes Z k;k to be established, SA k;k is deleted, and Fig. 11 becomes Fig. 12.
  • SA k;k is deleted, and Fig. 11 becomes Fig. 12.
  • the whole figure 12 (including B k ) is also deleted, and no further calculation is needed.
  • This example is equivalent to Example 2 except that different expressions are used, so the results are the same.
  • n 3, 2, 1, 0, and the others are the same as in Example 1.
  • E X 1,1 X 2,1 X 4, 1 X 5, 1 RG i3 .
  • Figure 18 has two more evidences X 6, 2 and X 7,1 that cannot be interpreted by B or BX variables.
  • a 1,1;1D 0.4
  • a 2,1; 2D 0.5
  • the degree of attention ⁇ is scored according to the percentage system
  • S 1 ⁇ X 1,1 , X 2,1 , X 4,1 , X 5,1 ⁇
  • S 2 ⁇ X 6,2 , X 7,1 ⁇ .

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)
  • Algebra (AREA)
  • Fuzzy Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Automation & Control Theory (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Debugging And Monitoring (AREA)

Abstract

一种计算机可读存储介质,所述存储介质上存储有计算机程序,所述计算机程序被执行时,能够执行处理不确定因果关系类信息的智能系统的方法,该方法在既有动态不确定因果图技术的基础上,增加了如下技术内容:用一种新的逻辑门和新的作用变量来表达证据及其组合对根原因事件发生概率的影响;用反向逻辑门来表达原因事件与一个以上结果事件组合之间的逻辑关系,进行推理;用特异性事件来表达其与根原因事件的对应关系,当其被观测到时,直接确定原因;用结果事件的关注度表达其不能被推理结果解释而导致推理结果成立可能性下降的程度,并参与概率计算;用根原因事件的危险度来表达其对系统的危害程度,并参与推理计算。

Description

一种扩展的处理不确定因果关系类信息的智能系统的构造方法 技术领域
本发明涉及智能信息处理技术,是对已授权专利《一种处理不确定因果关系类信息的只能系统的构造方法》(专利号:ZL 2006 8 0055266.X)、《METHOD FOR CONSTRUCTING AN INLLIGENT SYSTEM PROCESSING UNCERTAIN CAUSAL RELATIONSHIP INFORMATION》(专利号:US 8,255,353 B2)和《一种基于动态不确定因果图的启发式检测系统异常原因的方法》(专利号:ZL 2016 1 0282052.1)中所记载的技术方案的进一步扩展。使用本发明提出的技术方案,通过计算机处理,可进一步提升被称为动态不确定因果图DUCG(Dynamic Uncertain Causality Graph)的因果知识表达和运用能力,使之更加满足实际需求、更加方便准确地诊断对象系统异常的原因,从而有利于人们在对象系统异常的情况下采取有效措施,使之尽快恢复正常。
背景技术
如已授权专利《一种处理不确定因果关系类信息的只能系统的构造方法》、《METHOD FOR CONSTRUCTING AN INLLIGENT SYSTEM PROCESSING UNCERTAIN CAUSAL RELATIONSHIP INFORMATION》和《一种基于动态不确定因果图的启发式检测系统异常原因的方法》所记载,工业系统、社会系统和生物系统(简称对象系统)中存在大量导致系统异常的原因事件,例如线圈短路、泵故障停转、零部件失效、子系统失灵、传导通路阻塞、异物进入、某组织或机体发生变异、坏死、被污染、感染、伤害或自然失效等,称为事件变量,可用B i或BX k表示,k为事件变量标号、B kj或BX kj为变量B i或BX k的j状态。如图1和图2所示,B k、B kj、BX k和BX kj可分别用图形
Figure PCTCN2018085539-appb-000001
Figure PCTCN2018085539-appb-000002
表示,B k和BX k的区别在于B k为根原因变量,没有输入,而BX k有输入,可受其他因素影响,BX k是受影响后的B k。通常j=0表示B k或BX k处于正常状态;j=1、2、3…表示B k或BX k处于各种异常状态。
如果图形中只有一个数字,该数字代表变量标号,变量状态未知。为方便起见,kj之间可用逗号隔开(下同)。
B k和BX k的状态多数情况下不能直接检测,或难以直接检测,是DUCG智能系统需要推断其是否处于异常状态的变量。
此外,系统中还存在大量与B k或BX k有确定或不确定因果关系的变量,如温度、压力、流量、速度、频率、开关状态、各种化验或物理试验结果、调查结果、影像学结果、感觉、症状、体征、所处地域、时间、环境、季节、宗教、肤色、经历、血缘关系、嗜好、性格、居住条件、工作条件等,称为中间或结果变量,可以用X y表示,y=0、1、2、…。X yg为X y的g状态。通常g=0表示X y处于正常状态,g≠0表示X y处于标号为g的异常状态。X变量有至少一个输入(原因)变量,可以有或没有输出(结果)变量。如图1和图2所示,X y和X yg可分别用 图形符号
Figure PCTCN2018085539-appb-000003
表示。
采用DUCG技术方案,人们可通过检测X类型变量的状态来获取证据E,以推断系统异常的原因B kj或BX kj(j≠0),从而及时采取有效措施,使系统恢复正常。E由至少一个X变量的已知状态X yg构成,例如E=X 1,2X 2,3X 3,1X 4,0X 5,0
DUCG智能推理就是求解Pr{H kj|E}=Pr{H kjE}/Pr{E},其中H kj为假设事件,通常为DUCG中定义的待求的变量状态组合,例如H 1,2=B 1,2、H 2,1=BX 2,1、H 3,2=BX 3,2X 4,1等等,H kj中的下标k表示变量组合,例如H 1=B 1、H 2=BX 2、H 3=BX 3X 4、等等;H kj中的下标j表示H k中变量的状态组合,如上例。将所有在E条件下的可能假设事件H kj的集合用S H表示,则有H kj∈S H
DUCG中还定义了如下变量:
逻辑门变量,可用G i表示,至少有两个输入变量和一个输出变量。G ij表示G i的j状态。G i用以表达输入变量各种所关注状态的逻辑组合,这些逻辑组合用逻辑门说明表LGS i说明。例如G 1用LGS 1说明:G 1,1=B 3,1X 1,1,G 11,2=B 3,1X 1,2,G 1,0=其它情况(剩余状态),等等。G ijG ij’=0(空集,其中j≠j’),也就是说,G的不同状态互斥。如上述专利文献附图所示,G i和G ij可分别用图形符号
Figure PCTCN2018085539-appb-000004
Figure PCTCN2018085539-appb-000005
表示。
缺省原因变量,可用D表示,是对应的X变量的原因变量之一。例如D 4是X 4的缺省原因变量。Pr{D i}≡1。如图1所示,D i可用图形
Figure PCTCN2018085539-appb-000006
表示。
B、X、BX、D和G变量可称为节点,变量本身及其各种状态的物理意义可以根据所描述对象定义。其中B、X、BX、D和G为直接原因变量时,称为父变量,可用V统一表达,V∈{B,X,BX,D,G},下标不变。例如V 2=X 2,V 3,2=B 3,2,等等。结果变量只能是X或BX变量。变量的一个状态就是一个事件。例如X yg、B kj、BX kj、G ij、H kj或V ij等都是事件。
DUCG是由上述变量以及变量之间确定或不确定的因果关系构成的,通常用图形符号来表示。一个DUCG示例如图1所示,其中B变量或事件用矩形表示;X变量或事件用圆形表示;BX变量或事件用双圆表示;G变量或事件用逻辑门表示,并用有向弧
Figure PCTCN2018085539-appb-000007
将输入变量与G i相连;D变量用五边形表示。
作用变量F n;i是表达父变量V i与子变量X n或BX n之间的因果关系矩阵,可用有向弧
Figure PCTCN2018085539-appb-000008
或其它图形符号表示,从原因指向结果,矩阵中的元素称为作用事件F nk;ij,表达父事件V ij与子事件X nk或BX nk之间的因果关系,F nk;i表达父变量V i与子事件X nk之间的因果关系,F n;ij表达父事件V ij与子变量X n或BX n之间的因果关系。F nk;ij≡(r n;i/r n)A nk;ij。其中r n;i>0为父变量V i与子变量X n或BX n之间的因果关系强度,r n≡∑ ir n;i,A nk;ij为V ij导致X nk或BX nk发生这一随机事件,a nk;ij≡Pr{A nk;ij},满足
Figure PCTCN2018085539-appb-000009
f nk;ij≡Pr{F nk;ij}≡(r n;i/r n)a nk;ij。f nk;ij是V ij对X nk的权重概率的贡献值,满足
Figure PCTCN2018085539-appb-000010
F nk;ij、f nk;ij、A nk;ij、a nk;ij分别是矩阵F n;i、f n;i、A n;i、a n;i的元素。v ij=Pr{V ij},v∈{b,x,bx,d,g},V ij和v ij分别是事件向量V i和参数向量v i的元素。当原因变量为D i时,F nk;ij≡F nk;iD,此 时j=D。其它因果关系变量的表达类似。
F nk;ij可以是条件作用事件,用虚线有向弧
Figure PCTCN2018085539-appb-000011
表示。条件作用事件表达其原因事件V ij与结果事件X nk或BX nk之间是条件作用关系,即根据条件事件Z nk;ij是否满足来判定F nk;ij是否成立。例如Z nk;ij=X 1,2,当X 1,2为真时,Z nk;ij满足、F nk;ij成立;当X 1,2非真时,Z nk;ij不满足、F nk;ij不成立。当同一对父子变量之间的条件因果作用关系成立的条件事件相同时,该条件事件统一用Z n;i来表示,例如Z n;i=X 1,2,当X 1,2为真时,Z n;i满足、F n;i成立、
Figure PCTCN2018085539-appb-000012
成为
Figure PCTCN2018085539-appb-000013
当X 1,2非真时,Z n;i不满足、F n;i不成立、
Figure PCTCN2018085539-appb-000014
被删除。
为方便起见,全集记为1,空集记为0。上述变量及其状态也可以选择用其它图形或符号表示。
当收到证据E后,可采用下述规则对DUCG进行化简:
规则1:如果E显示Z nk;ij或Z n;i不满足,将F nk;ij或F n;i从DUCG中删除,当E显示Z nk;ij或Z n;i已经满足,虚线的F nk;ij或F n;i成为实线的F nk;ij或F n;i
规则2:如果E显示V ij(V∈{B,X})为真,但V ij却不是X n或BX n的父事件,将F n;i从DUCG中删除。
规则3:如果E显示X nk为真,但X nk不可能被V i(V∈{B,X,BX,G,D})的任何状态引起,将F n;i从DUCG中删除,除非X nk是待求变量的下游变量,且它们之间没有已知证据阻断。
规则4:如果E显示{B,X}类型节点状态未知且无输出有向弧,将该节点及其输入有向弧从DUCG中删除。
规则5:如果E显示X n0为真,且X n0与异常证据E’无任何连通关系,将X n0从DUCG中删除,除非X n0是待求B或BX变量的下游(结果)变量,且中间没有任何状态已知变量阻断。
规则6:如果E显示一组状态未知节点除非通过X n0,否则不与X nk(k≠0)相连,将这组状态未知节点及与之相连的有向弧和D节点从DUCG中删除。
规则7:因任何原因导致G i没有输出,将G i及其输入有向弧
Figure PCTCN2018085539-appb-000015
从DUCG中删除;当G i没有输入,将G i及其输出有向弧从DUCG中删除。
规则8:当有向弧没有父节点或没有子节点,将这条有向弧从DUCG中删除。
规则9:当存在一组节点和有向弧与E中涉及的节点无连通关系,将这组节点和有向弧从DUCG中删除。
规则10:如果E显示异常状态X nk为真,但X nk因任何原因没有输入,对X nk增加一个虚拟事件D n作为其输入,在从D n到X nk的有向弧中,a nk;nD=1且a nk’;nD=0,k’≠k,r n;D可以为任何值。D n可用符号
Figure PCTCN2018085539-appb-000016
表示。
规则11:如果E显示存在一组状态正常的X类型事件X n0∈S I(下标0代表正常状态),它们仅与状态未知变量相连,这些状态未知变量被X n0∈S I阻断而不与其它状态已知变量相连,且不包含已确定的假设事件H kj,则将这组状态未知变量和X n0∈S I删除,但当X n是被关注的假设事件H kj的结果变量且中间无已知证据事件阻断时,X n0不被删除。
规则12:上述规则可以按照任何顺序单独使用、联合使用、重复使用。
根据化简后的DUCG图,可进一步假定其中某个B变量成立时,其他B变量均不成立,从而将化简图分解为各个子化简图,并按照上述化简规则进一步化简。有些子图在化简后,B或BX变量没有了下游异常证据,则该子图被删掉。 剩下的子图中的B或BX变量的异常状态构成系统异常的可能原因事件空间S H。通常由B kj或BX kj(k≠0)构成。对H kj∈S H,求H kj=B kj或H kj=BX kj(k≠0)的验后概率:
Figure PCTCN2018085539-appb-000017
ζ k=Pr{E|子图k}。进而可以求出归一化排序概率:
Figure PCTCN2018085539-appb-000018
根据
Figure PCTCN2018085539-appb-000019
Figure PCTCN2018085539-appb-000020
人们即可知道系统异常的可能原因和排序,从而采取正确的应对措施,使系统尽快恢复正常。
为了解决有效收集证据的问题,专利文献《一种基于动态不确定因果图的启发式检测系统异常原因的方法》(专利号:ZL 2016 1 0282052.1)提出了推荐检测状态未知的X变量状态的方法,从而使上述证据E更加丰富有效,以便更加准确地进行推理判断。在该专利文献权利要求5中,在计算待检测变量X i的概率重要度ρ i时,采用了多种计算公式,例如:
Figure PCTCN2018085539-appb-000021
在这些计算公式中,ω k均与下标j无关,即与根原因变量的异常状态无关。但实际上,根原因变量的不同异常状态的受关注程度应当有所不同。
上述技术方案还存在如下限制:(1)G ijG ij’=0(j≠j’),这对表达逻辑组合提出了严格的要求。当人们只是考虑各种影响因素X nk的组合对B i各状态的概率分布的影响时,希望组合方式更加灵活、不受或少受上述限制;(2)
Figure PCTCN2018085539-appb-000022
其中包含a nk;ij≤1。但有时a参数的意义为放大或缩小事件发生概率的倍率,因而a nk;ij
Figure PCTCN2018085539-appb-000023
均可能大于1;(3)逻辑门G只考虑了其输入变量的状态组合。但有时人们需要表达逻辑门的输出变量的状态组合;(4)实际应用中存在某些X类型变量,称为特异性变量,一旦其异常状态被观测到,其对应的B或BX类型的根原因就可以确定,不需要进行复杂的概率推理;(5)DUCG根据E来推理,但有时X变量状态的异常并不是由于当前的B或BX引起的,而是由于其他未知原因引起的。对不同X的不同状态,人们关注其是否能被解释的程度是不同的。因此需要定义X nk(k≠0)的关注度及其参与计算的方法,以便达到在其他条件相同的情况下,不可解释的状态异常X类型证据越多,相对应的B或BX是当前系统异常原因的可能性就越小。
为此,本发明提出了扩展的技术方案来解决上述问题。
本发明既有技术参考文献:
[1]中国发明专利:“一种处理不确定因果关系类信息的智能系统的构造方法”,专利号:ZL 2006 8 0055266.X;授权日期:2010年4月14日;权利人:张湛;发明人:张勤、张湛。
[2]美国发明专利:“Method for constructing an intelligent system processing uncertain causal relationship information”;专利号:US 8255353 B2;授权日期:2012年8月28日;权利人:张湛;发明人:张勤、张湛。
[3]中国发明专利:“一种构造立体DUCG智能系统用于动态故障诊断的方法”;专利号:ZL 2013 1 0718596.4;授权日期:2015年4月15日;权利人:张湛、北京清能创新科技有限公司;发明人:张勤、董春玲。
[4]中国发明专利:“一种基于动态不确定因果图的启发式检测系统异常原因的方法”;专利号:ZL 2016 1 0282052.1;授权日期:2015年2月;权利人:北京清睿智能科技有限公司;发明人:张勤。
[5]Q.Zhang.“Dynamic uncertain causality graph for knowledge representation and reasoning:discrete DAG cases”,Journal of Computer Science and Technology,vol.27,no.1,pp.1-23,2012.
[6]Q.Zhang,C.Dong,Y.Cui and Z.Yang.“Dynamic uncertain causality graph for knowledge representation and probabilistic reasoning:statistics base,matrix and fault diagnosis”,IEEE Trans.Neural Networks and Learning Systems,vol.25,no.4,pp.645-663,2014.
[7]Q.Zhang.“Dynamic uncertain causality graph for knowledge representation and probabilistic reasoning:directed cyclic graph and joint probability distribution”,IEEE Trans.Neural Networks and Learning Systems,vol.26,no.7,pp.1503-1517,2015.
[8]Q.Zhang.“Dynamic uncertain causality graph for knowledge representation and probabilistic reasoning:continuous variable,uncertain evidence and failure forecast”,IEEE Trans.Systems,Man and Cybernetics,vol.45,no.7,pp.990-1003,2015.
[9]Q.Zhang and S.Geng.“Dynamic uncertain causality graph applied to dynamic fault diagnosis of large and complex systems”,IEEE Trans.Reliability,vol.64,no.3,pp 910-927,2015
[10]Q.Zhang and Z.Zhang.“Dynamic uncertain causality graph applied to dynamic fault diagnoses and predictions with negative feedbacks”,IEEE Trans.Reliability,vol.65,no.2,pp 1030-1044,2016.
[11]Q.Zhang&Q.Yao.Dynamic Uncertain Causality Graph for Knowledge Representation and Reasoning:Utilization of Statistical Data and Domain Knowledge in Complex Cases.IEEE Trans.Neural Networks and Learning Systems,DOI:10.1109/TNNLS.2017.2673243,2017.
发明内容
本发明公开了一种技术方案,进一步扩展了已授权中国发明专利ZL 2006 8 0055266.X、ZL 2016 1 0282052.1、美国发明专利US 8255353 B2以及上述文献所公开的DUCG技术方案。
本发明的技术方案如下:
1、一种通过利用一种计算机可读存储介质,其特征在于:所述存储介质上存储有计算机程序,所述计算机程序被执行时,能够执行如下处理不确定因果关系类信息的智能系统的构造和推理方法,该方法在既有DUCG技术方案的基础上,增加了表达和推断对象系统异常原因B k的方法,增加的内容包括:(1)用一种新的逻辑门SG k和新的作用变量SA k;k来表达证据X yg及其组合对B k各状态发生概率的直接影响,影响后的B k用BX k来表示,X y和B k是SG k的输入,事件矩阵SA k;k是SG k的输出,指向BX k,SA k;k的成员事件为SA kj;kn;(2)用反向逻辑门RG i来表达原因变量各状态与一个以上结果变量的状态组合之间的逻辑关系,并根据结果变量的有意义的状态证据组合确定反向逻辑门的状态,并据其进行DUCG推理;(3)用变量SX y来表达与特定的B变量的某异常状态对应的特异性X变量,当观测到SX yg(g≠0)时,不需要对SX yg进行推理计算即判定其对应的B变量的对应异常状态发生;(4)用X yg或SX yg的关注度ε yg(g≠0)来表达X yg或SX yg不能被推理结果H kj解释而导致H kj成立的可能性下降的程度,并用ε yg参与H kj的状态概率计算,使得参与计算的ε yg越多、值越大,H kj成立的可能性越小;(5)用B k的异常状态B kj的危险度μ kj来表达B kj对系统的危害程度,使得μ kj越大,检测对确定B k的状态有帮助的X类型变量状态的需求越大。
2、如1(1)所述,其特征还在于:1)当B k=B kj时,BX k=BX kj,反之亦然;2)用某种图形符号表达SG k,用一种有向弧表达从B k或X y指向SG k的输入关系;3)用另一种有向弧表达SA k;k,从SG k指向BX k;4)sa kj;kn≡Pr{SA kj;kn}代表将Pr{B kj}放大或缩小为Pr{BX kj}的缩放倍率,不受Pr{SA kj;kn}≤1的限制;5)SA k;k可以是条件事件矩阵,用与3)的有向弧有区别的有向弧表达,从SG k指向BX k,其条件事件用Z k;k表示,是一个可观测事件,当Z k;k成立时,SA k;k被删掉,否则将其保留成常规SA k;k;6)在SG k的逻辑门说明表LGS k中,用事件组合式n来表达SG kn的X类型输入事件的标号为n≠1的事件组合;7)当n=1时,SG k1的事件组合式为其它标号事件组合式的剩余状态,剩余状态也可以用n≠1的其它标号来标识;8)对n按优先顺序排序;9)根据现场收集到的X类型证据依n的排序与事件组合式n进行匹配,一旦事件组合式n被匹配上,即确定SG k=SG kn;10)当匹配上的n是例如0的特殊标号时,B k不成立,此时B k和SG k及其输入和输出有向弧可以被删除;11)不在事件组合式n中的SG kn输入端的状态未知或正常的X变量的指向SG kn的有向弧可以被删除;12)当匹配上的n不是上述特殊标号时,把Pr{B kj|E}替换为Pr{BX kj|E},BX kj=SA kj;knB kj,从而Pr{B kj|E}=sa kj;knb kj,其中E为收集到的证 据。
3、如1(2)所述,其特征还在于:1)反向逻辑门RG i可用某种图形符号表示,有至少一个输入变量,用从输入变量指向RG i的F类型有向弧连接,有不少于两个输出变量,用从RG i指向输出变量的有向弧连接;2)RG in为RG i的标号为n的状态,代表标号为n的输出变量的状态组合,用事件组合式n表达;3)推理中,RG in作为一个X类型事件进行DUCG逻辑展开;4)当n是例如0的特殊标号时,表示有意义的输出变量状态组合不存在,将RG i0及其输入和输出有向弧删掉;5)对n按照优先程度排序,当收到证据E后,按照n的优先顺序进行RG in的事件组合式匹配,一旦匹配上,即确定RG i=RG in;6)RG i输出的F类型有向弧中的a参数可以根据RG i的LGS i自动生成,生成的规则为:考察RG in的事件组合式中是否有X yg,若有,则a yg;in=1,即A yg;in=1,否则a yg;in=0或“-”,即A yg;in=0。
4、如1(4)所述,其特征还在于:1)ε yg仅在H kj在其所在的DUCG子图k中不能作为原因解释证据X yg或SX yg时才参与H kj的状态概率
Figure PCTCN2018085539-appb-000024
的计算;2)ε yg参与计算的方式为:在对H kj所在标号为k的子图的权重系数
Figure PCTCN2018085539-appb-000025
的计算中,在计算ζ k时,
Figure PCTCN2018085539-appb-000026
其中S 1为在子图k中能够被H kj解释的证据标识的集合,S 2为在子图k中不能被H kj解释的X yg或SX yg类型的证据标识的集合。
5、如1(5)所述,其特征还在于:1)在计算待检测变量X i的概率重要度ρ i时,用μ kj替代ω k,2)在计算概率重要度ρ i时,将μ kj放到对ρ i的计算公式的下标j的内层,包括但不限于将
Figure PCTCN2018085539-appb-000027
改为
Figure PCTCN2018085539-appb-000028
Figure PCTCN2018085539-appb-000029
其中J k为B k的异常状态数,S iG(y)为X ig在y时刻的全部或部分相关状态下标集合。
附图说明
图1 DUCG示例;
图2 收到证据E=X 6,2X 7,1X 14,1后的简化表达的图1;
图3 例1中的新型DUCG图;
图4 E=X 1,1X 2,1时图3的情况;
图5 例2中收到证据后的情况;
图6 按照权利要求2-9)化简图5的结果;
图7 收到证据E=1后的图3;
图8 图7的进一步化简结果;
图9 图8的进一步化简结果;
图10 图9的进一步化简结果;
图11 SA k;k为条件作用变量的情况;
图12 SA k;k被删除的情况;
图13 反向逻辑门的示例;
图14 收到E=X 1,1X 2,1X 4,1X 5,1后的图13;
图15 反向逻辑门的另一个示例;
图16 收到E=X 1,1X 2,1X 4,1X 5,1后的图15;
图17 收到E=X 2,1X 4,1X 5,1SX 8,1后的DUCG子图;
图18 基于E=X 1,1X 2,1X 4,1X 5,1X 6,2X 7,1化简后的DUCG子图。
具体实施方式
例1
如图3所示,SG k用图形符号
Figure PCTCN2018085539-appb-000030
表达,SG kj用图形符号
Figure PCTCN2018085539-appb-000031
表达,其输入用有向弧用
Figure PCTCN2018085539-appb-000032
表达,SA k;k用有向弧
Figure PCTCN2018085539-appb-000033
表达。在B kj和BX kj中,j∈{0,1,2}。根据权利要求1(1)和2,X 1、X 2和X 3与B k共同构成了SG k的输入变量,BX k为SG k的输出变量,由SA k;k连接,图3中内嵌了逻辑表达式LGS k(表1)、SA k;k的参数和B k的参数。
表1 图3中的LGS k
Figure PCTCN2018085539-appb-000034
其中n=0为特殊标号;
图3中B k的参数为:b k=(- 0.01 0.002) T,SA k;k的参数sa k;k为:
Figure PCTCN2018085539-appb-000035
其中“-”表示不关心(与0等价);sa kj;kn的意义为:在基于证据E确定双线逻辑门SG k输入端的X类型变量状态组合为LGS k中的事件组合式n的情况下,Pr{B kj}=b kj被缩放为Pr{BX kj}的倍率,也是就是说,对每个E,sa k;k中只有一列 参数参与运算。例如当E确定n=2时,只有sa k;k2=(- 10 2) T参与运算。
设H kj=B kj、E=X 1,1X 2,1、n的排序为0、3、2、1。根据权利要求2-9),当收到E后,按照n的排序进行事件组合式匹配,结果SG k2=X 1,1X 2,1被匹配上,所以SG k2为真。由于X 3不在E中,按照权利要求2-11),X 3的输入和输出被删掉。最后,图3化简为图4(其中状态的颜色可以自由定义)。
基于图4,根据权利要求1(1)和2-12),我们有
Figure PCTCN2018085539-appb-000036
当j=0时,Pr{H k0|E}=sa k0;k2b k0=“-”ד-”=“-”;
当j=1时,Pr{H k1|E}=sa k1;k2b k1=10×0.01=0.1;
当j=2时,Pr{H k2|E}=sa k2;k2b k2=2×0.002=0.004。
采用DUCG中定义的“*”算符,以上可简写为:
Figure PCTCN2018085539-appb-000037
其中算符“*”在DUCG中的定义是将两个行数相同矩阵的相同行的事件或数据进行逻辑与或相乘的运算(详见参考文献[6])。
例2
除E=X 1,0外,其它与例1相同。按照n的排序和表1,SG k0=X 1,0∪X 2,0X 3,0被匹配上。于是图3成为图5。
按照权利要求2-10),B k和SG k0及其输入和输出有向弧均被删除,成为图6。设X 1,0为X 1的正常状态,按照DUCG的化简规则,图6中的所有变量均被删掉,即图6被删掉,B k也就被删掉了,即B k的异常状态不存在,不再需要进行后续计算。
例3
除E=1(全集)外,其它与例1相同。E=1表示X 1、X 2和X 3的状态均未知,在表1中仅匹配剩余状态。因此SG k1成立,图3成为图7。根据权利要求2-11),图7被化简为图8。根据DUCG的化简规则,图8进一步化简为图9。
由于SG k=SG k1,从表1可知,sa k0;k1=“-”,sa k1;k1=sa k2;k1=1。根据权利要求2-12),与例1的计算类似,我们有
Pr{H k0|E}=sa k0;k2b k0=“-”ד-”=“-”;
Pr{H k1|E}=sa k1;k2b k1=1×0.01=0.01;
Pr{H k2|E}=sa k2;k2b k2=1×0.002=0.002。
也就是说,BX k的状态概率分布与B k完全相同。在这种情况下,B k完全等同于BX k,因此可以用B k替代BX k。也就是图9可以被进一步化简为图10。
例4
将图3改为图11,Z k;k=X 1,0∪X 2,0X 3,0,E=X 1,0,其余不变,其中用
Figure PCTCN2018085539-appb-000038
表达权利要求2-5)所述有向弧。
根据权利要求2-5),E使得Z k;k成立,SA k;k被删掉,图11变成图12。再应用DUCG的化简规则,整个图12(包括B k)也被删掉,不再需要后续计算。本例与例2等价,只是采用了不同的表达方式,所以结果相同。
例5
权利要求1和3所述之反向逻辑门的情况如图13所示,其中反向逻辑门RG i用图形符号
Figure PCTCN2018085539-appb-000039
表示,RG in用图形符号
Figure PCTCN2018085539-appb-000040
表示,BX k是反向逻辑门RG i的输入,X 4和X 5是RG i=(RG i0 RG i1 RG i2) T的输出,LGS i如表2所示:
表2 图13中的LGS i
Figure PCTCN2018085539-appb-000041
此外,
Figure PCTCN2018085539-appb-000042
对n的排序为3、2、1、0,其它与例1相同。
根据权利要求3-6)和表2,生成
Figure PCTCN2018085539-appb-000043
设E=X 1,1X 2,1X 4,1X 5,1,根据LGS i,RG i=RG i3,图13成为图14。基于图14,根据DUCG的展开方法,从下游往上游逐步展开:
Figure PCTCN2018085539-appb-000044
在只有一个输入的情况下,r参数不起作用
由于a 4,1;i3=a 5,1;i3=1,即A 4,1;i3=A 5,1;i3=1,上式为
Figure PCTCN2018085539-appb-000045
此处用到了权利要求2-12)
其中为方便起见,采用了DUCG中的*算符,定义已如例1所述。
进而,
X 1,1=F 1,1;1DD 1=A 1,1;1DD 1
X 2,1=F 2,1;2DD 2=A 2,1;2DD 2
E=X 1,1X 2,1X 4,1X 5,1
=A 1,1;1DD 1A 2,1;2DD 2A i3;k(SA k;k2*B k)
设H kj=B kj,我们有
Figure PCTCN2018085539-appb-000046
当j=0时,
Figure PCTCN2018085539-appb-000047
当j=1时,
Figure PCTCN2018085539-appb-000048
当j=2时,
Figure PCTCN2018085539-appb-000049
例6
如图15所示,与图13比较,RG i的输出有向弧改为
Figure PCTCN2018085539-appb-000050
a 4;i和a 5;i去掉,其它仍如例5所示,即仍有E=X 1,1X 2,1X 4,1X 5,1,根据LGS i,RG i=RG i3,图15成为图16。
根据权利要求3-3),RG i3被作为证据纳入E中进行展开,即E=X 1,1X 2,1X 4,1X 5,1RG i3。其中X 4,1和X 5,1的上游由于没有F类型有向弧,展开终止,对E=X 1,1X 2,1X 4,1X 5,1RG i3的展开等价于对E=X 1,1X 2,1RG i3的展开。其实在例5中也有X 4,1X 5,1=RG i3,所以最后的计算结果与例5完全相同。
例7
如图17所示,其中E=X 1,1X 2,1X 3,1X 4,1X 5,1SX 8,1。与图13比较,缺少了证据X 1,1,增加了特异性证据SX 8,1。设SX 8,1的对应B k的状态为B k2。根据权利要1和4,可知Pr{B k2|E}=1。
例8
如图18所示。
图18与图13的唯一区别在于图18多了两个不能被B或BX变量解释的证据X 6,2和X 7,1。设a 1,1;1D=0.4、a 2,1;2D=0.5、关注度ε按照百分制打分,X 6,2的关注度ε 6,2=20,X 7,1的关注度ε 7,1=10,取“ε yg越大、值越小的算式”=1/ε yg。根据权利要求5,S 1={X 1,1,X 2,1,X 4,1,X 5,1},S 2={X 6,2,X 7,1}。于是有
Figure PCTCN2018085539-appb-000051

Claims (5)

  1. 一种通过利用一种计算机可读存储介质,其特征在于:所述存储介质上存储有计算机程序,所述计算机程序被执行时,能够执行如下处理不确定因果关系类信息的智能系统的构造和推理方法,该方法在既有DUCG技术方案的基础上,增加了表达和推断对象系统异常原因B k的方法,增加的内容包括:(1)用一种新的逻辑门SG k和新的作用变量SA k;k来表达证据X yg及其组合对B k各状态发生概率的直接影响,影响后的B k用BX k来表示,X y和B k是SG k的输入,事件矩阵SA k;k是SG k的输出,指向BX k,SA k;k的成员事件为SA kj;kn;(2)用反向逻辑门RG i来表达原因变量各状态与一个以上结果变量的状态组合之间的逻辑关系,并根据结果变量的有意义的状态证据组合确定反向逻辑门的状态,并据其进行DUCG推理;(3)用变量SX y来表达与特定的B变量的某异常状态对应的特异性X变量,当观测到SX yg(g≠0)时,不需要对SX yg进行推理计算即判定其对应的B变量的对应异常状态发生;(4)用X yg或SX yg的关注度ε yg(g≠0)来表达X yg或SX yg不能被推理结果H kj解释而导致H kj成立的可能性下降的程度,并用ε yg参与H kj的状态概率计算,使得参与计算的ε yg越多、值越大,H kj成立的可能性越小;(5)用B k的异常状态B kj的危险度μ kj来表达B kj对系统的危害程度,使得μ kj越大,检测对确定B k的状态有帮助的X类型变量状态的需求越大。
  2. 如1(1)所述,其特征还在于:1)当B k=B kj时,BX k=BX kj,反之亦然;2)用某种图形符号表达SG k,用一种有向弧表达从B k或X y指向SG k的输入关系;3)用另一种有向弧表达SA k;k,从SG k指向BX k;4)sa kj;kn≡Pr{SA kj;kn}代表将Pr{B kj}放大或缩小为Pr{BX kj}的缩放倍率,不受Pr{SA kj;kn}≤1的限制;5)SA k;k可以是条件事件矩阵,用与3)的有向弧有区别的有向弧表达,从SG k指向BX k,其条件事件用Z k;k表示,是一个可观测事件,当Z k;k成立时,SA k;k被删掉,否则将其保留成常规SA k;k;6)在SG k的逻辑门说明表LGS k中,用事件组合式n来表达SG kn的X类型输入事件的标号为n≠1的事件组合;7)当n=1时,SG k1的事件组合式为其它标号事件组合式的剩余状态,剩余状态也可以用n≠1的其它标号来标识;8)对n按优先顺序排序;9)根据现场收集到的X类型证据依n的排序与事件组合式n进行匹配,一旦事件组合式n被匹配上,即确定SG k=SG kn;10)当匹配上的n是例如0的特殊标号时,B k不成立,此时B k和SG k及其输入和输出有向弧可以被删除;11)不在事件组合式n中的SG kn输入端的状态未知或正常的X变量的指向SG kn的有向弧可以被删除;12)当匹配上的n不是上述特殊标号时,把Pr{B kj|E}替换为Pr{BX kj|E},BX kj=SA kj;knB kj,从而Pr{B kj|E}=sa kj;knb kj,其中E为收集到的证据。
  3. 如1(2)所述,其特征还在于:1)反向逻辑门RG i可用某种图形符号表示,有至少一个输入变量,用从输入变量指向RG i的F类型有向弧连接,有不少 于两个输出变量,用从RG i指向输出变量的有向弧连接;2)RG in为RG i的标号为n的状态,代表标号为n的输出变量的状态组合,用事件组合式n表达;3)推理中,RG in作为一个X类型事件进行DUCG逻辑展开;4)当n是例如0的特殊标号时,表示有意义的输出变量状态组合不存在,将RG i0及其输入和输出有向弧删掉;5)对n按照优先程度排序,当收到证据E后,按照n的优先顺序进行RG in的事件组合式匹配,一旦匹配上,即确定RG i=RG in;6)RG i输出的F类型有向弧中的a参数可以根据RG i的LGS i自动生成,生成的规则为:考察RG in的事件组合式中是否有X yg,若有,则a yg;in=1,即A yg;in=1,否则a yg;in=0或“-”,即A yg;in=0。
  4. 如1(4)所述,其特征还在于:1)ε yg仅在H kj在其所在的DUCG子图k中不能作为原因解释证据X yg或SX yg时才参与H kj的状态概率
    Figure PCTCN2018085539-appb-100001
    的计算;2)ε yg参与计算的方式为:在对H kj所在标号为k的子图的权重系数
    Figure PCTCN2018085539-appb-100002
    的计算中,在计算ζ k时,
    Figure PCTCN2018085539-appb-100003
    yg越大、值越小的算式),其中S 1为在子图k中能够被H kj解释的证据标识的集合,S 2为在子图k中不能被H kj解释的X yg或SX yg类型的证据标识的集合。
  5. 如1(5)所述,其特征还在于:1)在计算待检测变量X i的概率重要度ρ i时,用μ kj替代ω k,2)在计算概率重要度ρ i时,将μ kj放到对ρ i的计算公式的下标j的内层,包括但不限于将
    Figure PCTCN2018085539-appb-100004
    改为
    Figure PCTCN2018085539-appb-100005
    Figure PCTCN2018085539-appb-100006
    其中J k为B k的异常状态数,S iG(y)为X ig在y时刻的全部或部分相关状态小标集合。
PCT/CN2018/085539 2017-10-17 2018-05-04 一种扩展的处理不确定因果关系类信息的智能系统的构造方法 WO2019076026A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/841,065 US20200234169A1 (en) 2017-10-17 2020-04-06 Constructing intelligent system processing uncertain causal relationship type information

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710967729.XA CN107944562B (zh) 2017-10-17 2017-10-17 一种扩展的处理不确定因果关系类信息的智能系统的构造方法
CN201710967729.X 2017-10-17

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/841,065 Continuation US20200234169A1 (en) 2017-10-17 2020-04-06 Constructing intelligent system processing uncertain causal relationship type information

Publications (1)

Publication Number Publication Date
WO2019076026A1 true WO2019076026A1 (zh) 2019-04-25

Family

ID=61936171

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/085539 WO2019076026A1 (zh) 2017-10-17 2018-05-04 一种扩展的处理不确定因果关系类信息的智能系统的构造方法

Country Status (3)

Country Link
US (1) US20200234169A1 (zh)
CN (1) CN107944562B (zh)
WO (1) WO2019076026A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112381110A (zh) * 2020-10-10 2021-02-19 神华北电胜利能源有限公司 基于模糊规则的煤矿生产多维度数据异常概率预测方法

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107944562B (zh) * 2017-10-17 2019-07-05 北京清睿智能科技有限公司 一种扩展的处理不确定因果关系类信息的智能系统的构造方法
CN108629417B (zh) * 2018-05-14 2019-12-03 北京予同智能科技有限公司 一种对ducg进行分层递归推理的高效方法
CN117493980B (zh) * 2023-12-29 2024-03-19 合肥工业大学 一种集成特征提取和排序因果发现的轴承故障诊断方法

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101127100A (zh) * 2006-08-18 2008-02-20 张湛 一种处理不确定因果关系类信息的智能系统的构造方法
CN107944562A (zh) * 2017-10-17 2018-04-20 北京清睿智能科技有限公司 一种扩展的处理不确定因果关系类信息的智能系统的构造方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030121027A1 (en) * 2000-06-23 2003-06-26 Hines Kenneth J. Behavioral abstractions for debugging coordination-centric software designs
DE10132014A1 (de) * 2001-07-03 2003-01-23 Siemens Ag Verfahren zum Erstellen eines Kausalen Netzes auf Grundlage einer Wissensakquisition
CN101484891B (zh) * 2006-05-16 2010-04-14 张湛 一种处理不确定因果关系类信息的智能系统的构造方法
CN101814114B (zh) * 2010-01-07 2012-03-21 北京航空航天大学 一种利用模型转换进行故障诊断的方法
CN103745261B (zh) * 2013-12-24 2015-04-15 张湛 一种构造立体ducg智能系统用于动态故障诊断的方法
CN104063586B (zh) * 2014-06-11 2017-03-01 西北工业大学 基于多态故障树的贝叶斯网络故障预测方法
CN105956665B (zh) * 2016-04-29 2017-06-06 北京清睿智能科技有限公司 一种基于动态不确定因果图的启发式检测系统异常原因的方法

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101127100A (zh) * 2006-08-18 2008-02-20 张湛 一种处理不确定因果关系类信息的智能系统的构造方法
CN107944562A (zh) * 2017-10-17 2018-04-20 北京清睿智能科技有限公司 一种扩展的处理不确定因果关系类信息的智能系统的构造方法

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Z HANG, QIN ET AL.: "Dynamic Uncertain Causality Graph Applied to Dynamic Fault Diagnoses and Predictions With Negative Feedbacks", IEEE TRANSACTIONS ON RELIABILITY, vol. 65, no. 2, 30 June 2016 (2016-06-30), pages 1030 - 1044, XP011612443, DOI: doi:10.1109/TR.2015.2503759 *
ZHANG, QIN ET AL.: "Dynamic Uncertain Causality Graph for Knowledge Representation and Reasoning: Continuous Variable, Uncertain Evidence, and Failure Forecast", IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, vol. 45, no. 7, 31 July 2015 (2015-07-31), pages 990 - 1003, XP011584446, DOI: doi:10.1109/TSMC.2015.2392711 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112381110A (zh) * 2020-10-10 2021-02-19 神华北电胜利能源有限公司 基于模糊规则的煤矿生产多维度数据异常概率预测方法

Also Published As

Publication number Publication date
CN107944562B (zh) 2019-07-05
US20200234169A1 (en) 2020-07-23
CN107944562A (zh) 2018-04-20

Similar Documents

Publication Publication Date Title
Luo et al. Time series prediction of COVID-19 transmission in America using LSTM and XGBoost algorithms
Leslie Tackling COVID-19 through responsible AI innovation: Five steps in the right direction
De Ron et al. Psychological networks in clinical populations: Investigating the consequences of Berkson's bias
WO2019076026A1 (zh) 一种扩展的处理不确定因果关系类信息的智能系统的构造方法
Chiolero et al. Glossary for public health surveillance in the age of data science
WO2017185449A1 (zh) 一种基于动态不确定因果图的启发式检测系统异常原因的方法
Probert et al. Real-time decision-making during emergency disease outbreaks
US7720787B2 (en) Intelligence analysis method and system using subjective logic
Szwed et al. A new lightweight method for security risk assessment based on fuzzy cognitive maps
Liu et al. Explainable ai for medical image analysis in medical cyber-physical systems: Enhancing transparency and trustworthiness of iomt
Fenton et al. A privacy-preserving Bayesian network model for personalised COVID19 risk assessment and contact tracing
de Wolff et al. Evaluation of pool-based testing approaches to enable population-wide screening for COVID-19
Hatt et al. Sequential deconfounding for causal inference with unobserved confounders
Bhatia et al. Artificial intelligence-inspired comprehensive framework for Covid-19 outbreak control
Poole et al. A holistic approach for suppression of COVID-19 spread in workplaces and universities
Makar et al. Learning the probability of activation in the presence of latent spreaders
Schuetz et al. It's not just about accuracy: An investigation of the human factors in users' reliance on anti-phishing tools
Uddin et al. Impact of physician community structure on healthcare outcomes
Wang et al. Toward combatting COVID-19: A risk assessment system
Aylward et al. An introduction to artificial intelligence in developmental and behavioral pediatrics
Carney et al. Informatics metrics and measures for a smart public health systems approach: information science perspective
Vardavas et al. Modeling infectious behaviors: the need to account for behavioral adaptation in COVID-19 models
Priyadarshini et al. Novel framework based on ensemble classification and secure feature extraction for COVID-19 critical health prediction
Winston et al. Exploring Socioeconomic Status as a Global Determinant of COVID-19 Prevalence, Using Exploratory Data Analytic and Supervised Machine Learning Techniques: Algorithm Development and Validation Study
McBride et al. The efficacy of tables versus graphs in disrupting dark networks: An experimental study

Legal Events

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

Ref document number: 18867754

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18867754

Country of ref document: EP

Kind code of ref document: A1