CN102298978A - MFM (multilevel flow model)-based indeterminate fault diagnosis method for nuclear power plant for ship - Google Patents

Abstract

The invention provides an MFM (multilevel flow model)-based indeterminate fault diagnosis method for a nuclear power plant. The method is characterized by displaying the faults through alarm and marking the positions of the abnormal parts when the measured values which are detected by a sensor and represent variables of functional parts exceed a normal variation domain; intercepting the effect relationships which possibly cause the faults according to the direction of the flow in the MFM; cutting off a logic loop to form micro fault trees; converting the micro fault trees into micro GO-FLOW models; processing and combining the micro GO-FLOW models to form a GO-FLOW model by taking the intersection of the abnormal functions detected by the sensor as the final signal; and inputting the GO-FLOW model into software and carrying out computation to finally obtain the occurrence probability of the intersection of the abnormal functions caused by the basic faults. The method is suitable to solve complex system problems, is easy in verification, has the advantages of accurate results and high speed, and meets the real-time diagnosis requirement.

Description

Atomic marine plant is based on the indeterminate fauit diagnostic method of multilayer flow model

Technical field

What the present invention relates to is a kind of complication system method for diagnosing faults, specifically at complicated between atomic marine plant fault and sign thereof and have probabilistic logical relation and a kind of method for diagnosing faults based on the multilayer flow model of proposing.

Background technology

Fault diagnosis is meant application testing analysis means and diagnosis theory, system's mechanism that breaks down in service, reason, position and degree etc. is discerned, for further formulating the maintenance program of faulty equipment and taking suitable emergency operation to provide support.System fault diagnosis is identified as the basis with equipment state, and the abnormality of slave unit (being sign) is set out, and according to the cause-effect relationship between system's conservation principle and control law analysis sign, thereby realizes localization of fault, qualitative and determining cause.

Involved in the present invention to as if atomic marine plant, atomic marine plant is in actual motion, the structure of equipment and performance are subjected to noise, wave and factor affecting such as marine corrosion, show different degradation trends: on the one hand, the equipment of different process type is subjected to the influence degree difference of extraneous factor, shows different fault types; On the other hand, because running environment, operator's functipnal capability and the difference of management level, even the equipment of same process also may be different at the aspects such as frequency, the form of expression and performance characteristic that fault takes place.Especially, atomic marine plant belongs to the System in Small Sample Situation problem, lacks complete fault diagnosis knowledge, and is also indeterminate to the failure mechanism of some equipment; Perhaps clear, but because the complicacy of system itself to the failure mechanism of equipment, the influence of miscellaneous equipment and system is difficult to judgement.In addition, because the atomic marine plant system is huge, be subjected to the hull space constraint, can not realize that function is independent, have the shared phenomenon of equipment, in case break down, will cause system's quantity of parameters unusual, cause-effect relationship is difficult to clearly between sign.For example: have uncertainty between the presentation (failure symptom) and the origin cause of formation (fault mode), the corresponding various faults pattern of a kind of failure symptom possibility; Otherwise a certain fault mode also may show the various faults sign.At last, the running environment of equipment, control procedure, operator are intervened, sensor, and expertise etc. all has uncertain factor.Therefore to atomic marine plant carry out key issue that needs to be resolved hurrily of fault diagnosis be familiar with exactly with the handling failure diagnosis in uncertain problem.

Uncertainty can be understood as lacking under the situation of enough information and makes judgement, is the essential characteristic of intelligent problem; Reasoning is human thought process, is from the known fact, and the knowledge that utilization is relevant is progressively released the process of conclusion.The so-called uncertain reasoning is exactly from incomplete initial evidence, and by using uncertain knowledge, the final release has to a certain degree uncertain but the thought process of reasonable or intimate rational conclusion.

Uncertain inference can be qualitatively, quantitative, or qualitative and quantitative combination.Uncertain inference can be divided into symbolic reasoning and numerical value reasoning two big classes.Wherein, symbolic reasoning method information loss in reasoning process is less, but calculated amount is bigger, and typical method comprises approval theoretical (Endorsement Theory) etc.Though the numerical value inference method has certain information loss in reasoning process, be easy to realize typical method such as probability inference etc.In the uncertain inference field, mainly adopt the numerical value inference method at present.

What use the earliest in the probability inference method is uncertain factor method, is proposed in 1975 by Shortliffe and Buchanan, is used for handling the uncertain information of MUCIN system.1976, Duda etc. proposed subjective bayes method, and the design of the expert system PROSPECTOR that is applied to prospect.Subjective bayes method mainly utilizes the Bayesian formula of Bayesian formula and distortion to calculate the probability that hypothesis (given fault) takes place under the given evidence.1985, Per1 provided Bayesian network method, and Bayesian network is a directed acyclic graph, wherein node is a variable; arc is represented the dependence between the correlated variables, and the value of variable is corresponding to evidence and hypothesis, and the confidence level between the variable is represented with conditional probability.In recent years, Chinese scholar Zhang Qin has proposed cause and effect digraph method on the basis of Bayesian network, has solved the problem that Bayesian network can't be handled loop.

Yet, when existing system modeling methods such as utilizing Bayesian network, cause and effect digraph is carried out uncertain knowledge organization, the uncertain inference model itself that relies on expertise and set up exists random, and the model that different experts utilize Same Way to be set up at the same system problem also might be different.In addition, because the complicacy of system itself, institute's established model is huge usually, has caused the computation process complexity, has influenced the real-time of diagnosis.

Fundamental purpose of the present invention is uncertain knowledge representation of manual system and the inference method that proposes a kind of standardization, modularization and stratification for the atomic marine plant fault diagnosis.This method adopts the multilayer flow model, and (Multilevel Flow Models is MFM) as the systematic knowledge expression.MFM is the Morten Lind proposition of Technical University Of Denmark's eighties in last century, describes the purpose of design and the realization means thereof of complication system from aims of systems, function and three angles of physical unit from view of cognition science utilization semiology method.Different with other modeling method, MFM has not only described the process behavior of complication system, and has emphasized the purpose of these behaviors.Target, function and physical unit link into an integrated entity by different relation (comprise the relation reached, realize relation and conditional relationship), and it is abstract at many levels that MFM can be carried out goal systems on " means-purpose ", " partly-integral body " both direction.MFM introduces " stream " notion (Flow), and conserva-tion principle is followed in the propagation of " stream ".Therefore, with the MFM descriptive system have model specification, simple and clear, clear understandable, be easy to revise and advantages such as checking.

Aspect inference method, in conjunction with the successful part of BN, adopt probability that uncertain knowledge is measured, describe between fault mode and the sign, and the response relation between sign and the sign.Because in the multilayer flow model, each function all has corresponding constraint condition, reflects the cause-effect relationship between each function.Therefore, can be according to the annexation between each function in the multilayer flow model.Network progressively launches according to the relation between function and function one by one, finds out the cause-effect relationship between each fault mode and each sign and compares its size, finishes diagnostic reasoning.

Effect of the present invention is mainly reflected in:

1) institute's extracting method modeling standard, simple, be easy to revise, be applicable to the complication system problem that solves.

2) the model logical relation is clear, and the relation between sign and fault meets conservation principle, is easy to checking.

3) adopt method for precisely solving, the gained The reasoning results is accurate.

4) adopt modularization to calculate, speed is fast, satisfy the real-time diagnosis needs.

Summary of the invention

The objective of the invention is provides a kind of indeterminate fauit diagnostic method that can requirement of real time for atomic marine plant.

Before setting forth concrete steps of the present invention, need explain as follows to some nouns and the alphabetical implication that relate in the literary composition:

Target G _i: expression system its intended purposes, realized by drift net by one or more functions usually.

Drift net NETWORK: material, energy or information flow that the one group of same alike result and the function that is mutually related are formed are used to realize target.

Function F _i: the functional unit in the system model is that the actual physics parts are exercised the abstract of its function.The classification of its function and constraint condition are as shown in Figure 2.

State variable S _i: the running status of presentation function unit, S in the reality _iCan be multimode, suppose in the application's book that all functions (except that the Barrier function) are three condition, comprising:

-OK state, presentation function is normal;

-H state, the unusual high state of water level, flow, pressure, temperature etc.;

-L state, the unusual low state of water level, flow, pressure, temperature etc.

The Barrier function is a two condition

-OK state, presentation function is working properly;

-N state, Barrier can not finish predetermined function;

Fault: system can not or partly can not finish the state of its appointed function, can be divided into two kinds in this manual.

● unit failure B _i: be the physical unit failure mode of institute's analytic function, as environment division or all damage, on-off element that operator's maloperation caused open by mistake or mistake is closed etc.

● system failure X _i: refer to the fault that the physical unit of non-institute analytic function lost efficacy and to cause, the corresponding function abnormality can be divided into following 3 types again:

1. upstream failure: the fault that causes owing to the abnormality of upstream function.

2. downstream fault: the fault that causes owing to the abnormality of downstream function.

3. condition fault: owing to the required condition of function is not reached the fault that causes as yet.

Incident: be divided into elementary event and result event.Elementary event is positioned at cause and effect tree bottom, corresponding physics failure mode; Result event is divided into intermediate event and top event, and corresponding to functional status, intermediate event is the reason incident of a certain incident, is again the result event of another incident; Top event is the object event of being analyzed, and launches causality analysis with top event usually.

Influence intensity p _Ij: the possibility (0≤p that result event takes place that causes that represents a certain incident _Ij≤ 1,0 expression is irrelevant fully between the two, and 1 expression is relevant fully between the two).

The cause and effect subtree: a certain state with objective function launches causality analysis as top event, and bottom event only is referred to as the cause and effect subtree by unit failure of this function itself and the tree structure that the system failure constituted directly related with this function.

The cause and effect tree: a certain state with objective function launches causality analysis as top event, and bottom event is referred to as the cause and effect tree by the tree structure that unit failure constituted of each function.

Concrete implementation step of the present invention is as follows:

The first step when system breaks down, is at first measured (being system operational parameters) according to systematic perspective and is determined that whether each target is reached, and determines phase-split network.

In the multilayer flow model, target is to be realized by the form of one or more functions with drift net.Therefore, when certain target is reached, can think and realize that at current time the major function of drift net of this target is normal, therefore can drift net integral body as one not deploy events will not analyze.

Second step, measure according to systematic perspective, determine each function F _iRunning status S _iWith S _iFor top event is launched into the cause and effect subtree respectively.If certain function F _iRunning status S _iFor known, then launch according to known state; If certain function F _iState variable S _iBe the unknown, then need S _iCorresponding all states (H, L and OK) incident is deployment analysis respectively.

In the 3rd step, ask for function F _iState variable S _iCause and effect tree.On the basis of the resulting cause and effect subtree that is target with each functional status variable of the first step, connection also launches intermediate event till elementary event, obtains with S _iCause and effect tree for top event.

Because the non-reflexivity of system, promptly incident can not become the reason of oneself, therefore when launching to form the cause and effect tree, need form cause and effect loop place with the upper strata incident in the chain of causation and block.

The 4th step, above-mentioned cause and effect tree is translated into little GO-FLOW (a kind of probability analysis method) model, the implication of each element representative is as shown in Figure 3 in the GO-FLOW model.

Translation rule is as follows:

1) elementary event is represented by No. 25 operational characters of start signal generator, and No. 25 operational character is used for simulating the mono signal generator in GO-FLOW, and an output signal is only arranged.Can simulate the influence of elementary event in the method propagates.The probability of happening of its elementary event can be represented with the intensity of input signal.

2) influence intensity and represent that by No. 21 operational characters of two-state element No. 21 operational character is used for simulating the component that has only two states in GO-FLOW.Can be used for influencing between the analog representation cause and effect response intensity p that concerns in the method _Ij

3) logical relation " or " represent that with No. 22 operational characters No. 22 operational characters are used for simulating a plurality of signals in GO-FLOW or gate logic relation has a plurality of input signals and an output signal, is used for simulating the logical "or" relation of a plurality of incidents in the method.

4) logical relation " with " represent with No. 30 operational characters, No. 30 operational characters in GO-FLOW, be used for simulating a plurality of signals with the gate logic relation, a plurality of input signals and an output signal are arranged, be used for simulating the logical relation of a plurality of incidents in the method.

5) intermediate event can by elementary event with influence intensity and represent by the combination of logical and logical "or".

The 5th goes on foot, and asks for the state parameter S of all abnormal function states _iCommon factor, obtain evidence E.

In this manual, define the abnormal function state S that all monitor _iCommon factor be called evidence, i.e. E=S ₁S ₂S _n(n is the function number that is in abnormality)

In the 6th step, ask for each unit failure B _iEvidence influenced intensity P _i

Have by the Bayesian formula: $P(B_i/E)=\frac{P_r(B_i\·E)}{P_r\left(E\right)}$

Compare its each unit failure P _iSize, probable value is big more, the possibility of generation is big more.

Description of drawings

Fig. 1 is the outline flowchart of this method;

Fig. 2 is function figure and the constraint condition among the MFM;

Fig. 3 is each operational character functional definition table of GO-FLOW;

Fig. 4 is the nuclear power system sketch;

Fig. 5 is the MFM sketch of steam generator system shown in Figure 2;

Fig. 6-1 is to Fig. 6-the 9th, the cause and effect subtree of each functional status, wherein:

Fig. 6-1: with F1 (H) is the cause and effect subtree that top event is launched;

Fig. 6-2: with F2 (L) is the cause and effect subtree that top event is launched;

Fig. 6-3: with F3 (H) is the cause and effect subtree that top event is launched;

Fig. 6-4: with F4 (OK) is the cause and effect subtree that top event is launched;

Fig. 6-5: with F5 (OK) is the cause and effect subtree that top event is launched;

Fig. 6-6: with F9 (H) is the cause and effect subtree that top event is launched;

Fig. 6-7: with F10 (OK) is the cause and effect subtree that top event is launched;

Fig. 6-8: with F11 (OK) is the cause and effect subtree that top event is launched;

Fig. 6-9: with F12 (OK) is the cause and effect subtree that top event is launched;

Fig. 7-1 is to Fig. 7-the 4th, and the cause and effect of abnormal function state is set, wherein:

Fig. 7-1: with F1 (H) is the cause and effect tree that top event is launched;

Fig. 7-2: with F2 (L) is the cause and effect tree that top event is launched;

Fig. 7-3: with F3 (H) is the cause and effect tree that top event is launched;

Fig. 7-4: with F9 (H) is the cause and effect tree that top event is launched;

Fig. 8-1 is to Fig. 8-the 4th, and the cause and effect tree among Fig. 5 is translated the GO-FLOW model that forms, wherein:

Fig. 8-1: with F1 (H) is the GO-FLOW model that the cause and effect tree of top event is translated into;

Fig. 8-2: with F2 (L) is the GO-FLOW model that the cause and effect tree of top event is translated into;

Fig. 8-3: with F3 (H) is the GO-FLOW model that the cause and effect tree of top event is translated into;

Fig. 8-4: with F9 (H) is the GO-FLOW model that the cause and effect tree of top event is translated into;

Fig. 9 is to be target GO-FLOW model with the evidence;

Figure 10 is the functional semantics and the evaluating table of each function in the MFM model;

Figure 11 is the gained The reasoning results.

Embodiment

Fig. 1 is the outline flowchart of this method, this method is under the prerequisite of a large amount of nuclear power operating experiences, expert by this field estimates the function of nuclear power station operation, set up the multilayer flow model, set statement each functional part variable (temperature, pressure, water level etc.) normal variation territory, " H " of defined function state, " L " and " OK ".And, provide the size of the probable value of mutual response between each function by the true ruuning situation of expert's marking according to equipment.

With reference to Fig. 4 is that nuclear power system comprises reactor S1, voltage stabilizer S2, steam generator S3, steam turbine S4, feed pump S5 and main pump S6.The multilayer flow model that steam generator system is set up as shown in Figure 5, wherein, B1 is steam generator heat-transfer pipe fracture accident (SGTR); B2 is main steam line rupture accident (MSLB); B3 is the main feed pump stall.G0 is a nuclear power station generating target; G1 is for providing the cooling medium target to steam generator; G2 is for providing the chilled water target to steam generator; G3 is the secondary circuit radiological measuring; G4 is that external power is the feed pump power supply.The physical significance of each functional part representative as shown in Figure 8.

The response relation that is provided each functional part by expert's marking is as follows:

B1＝10 ^-5，B2＝10 ^-5，B3＝10 ^-8

B1→F8(N):1.0 B2→F4(L):0.9 B2→F10(L):0.9 B3→F12(L):1.0

F1(H)→F2(H):0.9 F1(H)→F2(OK):0.1

F1(OK)→F2(OK):0.8 F1(OK)→F2(H):0.1 F1(OK)→F2(L):0.1

F1(L)→F2(L):0.9 F1(L)→F2(OK):0.1

F2(H)→F3(H):0.9 F2(H)→F3(OK):0.1 F2(H)→F1(L):0.8 F2(H)→F1(OK):0.2

F2(OK)→F3(OK):0.8 F2(OK)→F3(H):0.1 F2(OK)→F3(L):0.1 F2(OK)→F1(OK):0.8F2(OK)→F1(H):0.1 F2(OK)→F1(L):0.1

F2(L)→F3(L):0.8 F2(L)→F3(OK):0.2 F2(L)→F1(H):0.9 F2(L)→F1(OK):0.1

F3(H)→F4(H):0.9 F3(H)→F4(OK):0.1 F3(H)→F2(L):0.8 F3(H)→F2(OK):0.2

F3(OK)→F4(OK):0.8 F3(OK)→F4(H):0.1 F3(OK)→F4(L):0.1 F3(OK)→F2(OK):0.8F3(OK)→F2(H):0.1 F3(OK)→F2(L):0.1

F3(L)→F4(L):0.9 F3(L)→F4(OK):0.1 F3(L)→F2(H):0.8 F3(OK)→F2(OK):0.2

F4(H)→F5(H):0.8 F4(H)→F5(OK):0.2 F4(H)→F3(L):0.9 F4(H)→F3(OK):0.1

F4(OK)→F5(OK):0.8 F4(OK)→F5(H):0.1 F4(OK)→F5(L):0.1 F4(OK)→F5(OK):0.8F4(OK)→F5(H):0.1 F4(OK)→F5(L):0.1

F4(L)→F5(L):0.9 F4(L)→F5(OK):0.1 F4(L)→F3(H):0.9 F4(L)→F3(OK):0.1

F5(H)→F4(L):0.9 F5(H)→F4(OK):0.1

F5(OK)→F4(OK):0.8 F5(OK)→F4(L):0.1 F5(OK)→F4(H):0.1

F5(L)→F4(H):0.8 F5(L)→F4(OK):0.2

F6(H)→F7(H):0.9 F6(H)→F7(OK):0.1

F6(OK)→F7(OK):0.8 F6(OK)→F7(L):0.1 F6(OK)→F7(H):0.1

F6(L)→F7(L):0.9 F6(L)→F7(OK):0.1

F7(H)→F6(L):0.8 F7(H)→F6(OK):0.2

F7(OK)→F6(OK):0.8 F7(OK)→F6(L):0.1 F7(OK)→F6(H):0.1

F7(L)→F6(H):0.9 F7(L)→F6(OK):0.1

F8(N)→F6(L):0.9 F8(N)→F6(OK):0.1 F8(N)→F9(H):0.9 F8(N)→F9(OK):0.9F8(N)→F3(H):0.9 F8(N)→F3(OK):0.1

F8(OK)→F6(OK):1.0 F8(OK)→F9(OK):1.0 F8(OK2)→F3(OK):1.0

F9(H)→F10(H):0.8 F9(H)→F10(OK):0.2 F9(H)→F12(L):0.9 F9(H)→F12(OK):0.1

F9(OK)→F10(OK):0.8 F9(OK)→F10(H):0.1 F9(OK)→F10(L):0.1 F9(OK)→F12(OK):0.8 F9(OK)→F12(H):0.1 F9(OK)→F12(L):0.1

F9(L)→F10(L):0.9 F9(L)→F10(OK):0.1 F9(L)→F12(H):0.9 F9(L)→F12(OK):0.1

F10(H)→F11(H):0.8 F10(H)→F11(OK):0.2 F10(H)→F9(L):0.9 F10(H)→F9(OK):0.1

F10(OK)→F11(OK):0.8 F10(OK)→F11(H):0.1 F10(OK)→F11(L):0.1 F10(OK)→F9(OK):0.8 F10(OK)→F9(H):0.1 F10(OK)→F9(L):0.1

F10(L)→F11(L):0.9 F10(L)→F11(OK):0.1 F10(L)→F9(H):0.9 F10(L)→F9(OK):0.1

F11(H)→F12(H):0.8 F11(H)→F12(OK):0.2 F11(H)→F10(L):0.9 F11(H)→?F10(OK):0.1

F11(OK)→F12(OK):0.8 F11(OK)→F12(H):0.1 F11(OK)→F12(L):0.1 F11(OK)→F10(OK):0.8 F11(OK)→F10(H):0.1 F11(OK)→F10(L):0.1

F11(L)→F12(L):0.9 F11(L)→F12(OK):0.1 F11(L)→F10(H):0.9 F11(L)→F10(OK):0.1

F12(H)→F9(H):0.8 F12(H)→F9(OK):0.2 F12(H)→F11(L):0.9 F12(H)→F11(OK):0.1

F12(OK)→F9(OK):0.8 F12(OK)→F9(H):0.1 F12(OK)→F9(L):0.1 F12(OK)→F11(OK):0.8 F12(OK)→F11(H):0.1 F12(OK)→F11(L):0.1

F12(L)→F9(L):0.9 F12(L)→F9(OK):0.1 F12(L)→F11(H):0.9 F12(L)→F11(OK):0.1

Suppose that now the SGTR fault takes place total system, the state of each function original paper that detector detects is:

F1(H)，F2(L)，F3(H)，F4(OK)，F5(OK)，F6(OK)，F7(OK)，F8(N)，F9(H)，F10(OK)，F11(OK)，F12(OK)。

Whether the first step is checked each target and is reached, and determines the network of required analysis.

According to the state that detects each function as can be known, G0 does not reach, and G1 reaches, and G2 does not reach, and G3 reaches, and G4 monitors radioactivity.With this, the drift net of required analysis is N0, N2 and function F8.

Second step is according to function F _iState variable S _i, obtain the cause and effect subtree of the state variable of all functions parts, as shown in Figure 6.

In the 3rd step, further abbreviation launches the not deploy events in the cause and effect subtree on the basis of the first step, asks for function F _iState variable S _iCause and effect tree, as shown in Figure 7.

In the 4th step, the cause-and-effect diagram that second step was formed is translated into little GO-FLOW model as shown in Figure 8.

The 5th goes on foot, and asks for the state variable S of all abnormal function states _iCommon factor, obtaining with evidence E is the GO-FLOW model of target, as shown in Figure 9.

Evidence E=F ₁(H) F ₂(L) F ₃(H) F ₉(H)

In the 6th step, ask for each unit failure B _iEvidence influenced intensity P _i

Have by the Bayesian formula: $P({\mathrm{<figure-callout\; id="1"\; label="B"\; filenames="HDA0000061687350000022.png,HDA0000061687350000042.png"\; state="\{\{state\}\}">B</figure-callout>}}_1/E)=\frac{P_r({\mathrm{<figure-callout\; id="1"\; label="B"\; filenames="HDA0000061687350000022.png,HDA0000061687350000042.png"\; state="\{\{state\}\}">B</figure-callout>}}_1\&CenterDot;E)}{P_r\left(E\right)}=0.05824$ $P({\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="HDA0000061687350000091.png,HDA0000061687350000121.png"\; state="\{\{state\}\}">B</figure-callout>}}_2/E)=\frac{P_r({\mathrm{<figure-callout\; id="2"\; label="B"\; filenames="HDA0000061687350000091.png,HDA0000061687350000121.png"\; state="\{\{state\}\}">B</figure-callout>}}_2\&CenterDot;E)}{P_r\left(E\right)}=0.94176$

The gained result as shown in figure 11.

Claims (2)

1. a nuclear power unit is characterized in that based on the indeterminate fauit diagnostic method of multilayer flow model:

The first step, when system broke down, at first measurement was that system operational parameters determines that whether each target is reached, and determines phase-split network according to systematic perspective;

Second step, measure according to systematic perspective, determine each function F _iRunning status S _i, with S _iFor top event is launched into the cause and effect subtree respectively, if certain function F _iRunning status S _iFor known, then launch according to known state; If certain function F _iState variable S _iBe the unknown, then need S _iAll corresponding state events are deployment analysis respectively;

In the 3rd step, ask for function F _iState variable S _iCause and effect tree, on the basis of the resulting cause and effect subtree that is target with each functional status variable of the first step, connect and also launch intermediate event till elementary event, obtain with S _iCause and effect tree for top event;

In the 4th step, above-mentioned cause and effect tree is translated into little GO-FLOW model;

The 5th goes on foot, and asks for the state parameter S of all abnormal function states _iCommon factor, obtain evidence E;

In the 6th step, ask for each unit failure B _iEvidence influenced intensity P _i,

By the Bayesian formula: $P(B_i/E)=\frac{P_r(B_i\&CenterDot;E)}{P_r\left(E\right)}$

Compare its each unit failure P _iSize, probable value is big more, the possibility of generation is big more.

2. nuclear power unit according to claim 1 is characterized in that based on the indeterminate fauit diagnostic method of multilayer flow model: described that cause and effect tree is translated into the translation rule of little GO-FLOW model is as follows:

1) elementary event is represented by No. 25 operational characters of start signal generator, and No. 25 operational character is used for simulating the mono signal generator in GO-FLOW, and an output signal is only arranged;

2) influence intensity and represent that by No. 21 operational characters of two-state element No. 21 operational character is used for simulating the component that has only two states in GO-FLOW;

3) logical relation " or " represent that with No. 22 operational characters No. 22 operational characters are used for simulating a plurality of signals in GO-FLOW or gate logic relation has a plurality of input signals and an output signal;

4) logical relation " with " represent with No. 30 operational characters, No. 30 operational characters in GO-FLOW, be used for simulating a plurality of signals with the gate logic relation, a plurality of input signals and an output signal are arranged;

5) intermediate event by elementary event with influence intensity and represent by the combination of logical and logical "or".

Images