CN103199919A - Multi-parameter-sensed high-accuracy network fault screening and positioning system and method - Google Patents

Abstract

The invention discloses a multi-parameter-sensed high-accuracy network fault screening and positioning method and relates to the technical field of optical communication techniques. According to the method, aiming at the defect that the positioning cost is high in a conventional wavelength-based positioning method, services are introduced to acquire fault warning information, and the positioning cost is saved. The invention further discloses a multi-parameter positioning method aiming at the limitation on performance promotion due to a single parameter in a conventional service-based positioning method, the positioning process is divided into three stages including fault prediction, screening and positioning, and multiple parameters are adopted to respectively finish the compressed-sensing-based fault screening and the further fault positioning. According to the method, the limitation of the single parameter is broken through by the clever union of the multiple parameters, and the positioning accuracy can be effectively improved.

Description

A kind of high-accuracy network fault screening navigation system and method for multi-parameter perception

Technical field

The present invention relates to the optical communication technique field, be specifically related to the periodic line fault location method in the optical wavelength-division multiplex network.

Background technology

Along with the development of wavelength division multiplexing (WDM) technology, simple optical fiber can be supported the information transmission of higher rate, and optical-fiber network institute loaded service amount sharply rises, and can cause mass data to be lost during link failure.Therefore, the survival ability of raising network is most important.The optical-fiber network Study on survivability mainly contains malfunction monitoring, fault location, error protection and four aspects of fault recovery.Wherein, error protection is preserved the professional alternate routing that connected and is also reserved corresponding resource, after fault takes place, business is gone on the alternate routing transmit, and reaches the professional purpose of protection; After fault recovery referred to that fault takes place, the link of fixing a breakdown was finished professional transmission by heavy-route.Fault location is the prerequisite of fault protection and recovery, and fault location can reduce the loss that link failure brings effectively fast and accurately.

After the link occurs fault, its loaded service can produce corresponding sign, and these signs are considered as alarm.By these warning information are handled, can find out location of fault, realize fault location.The mode that alarm produces mainly contains two kinds: the one, set the warning information that special monitoring wavelength initiatively obtains link; The 2nd, according to the professional passive warning information that obtains link that transmits on the link.Difference according to the alarm producing method can be divided into the existing fault localization method two classes: based on the fault location algorithm of monitoring wavelength with based on the fault location algorithm of business.According to the difference of monitoring wavelength design, the fault location algorithm based on the monitoring wavelength can be divided into monitoring circle algorithm, monitoring mark algorithm and monitoring tree algorithm three classes.People such as Hongqing Zeng are at " Fault detection and path performance monitoring in meshed all-optical networks " [IEEE GLOBECOM[C] .Dallas, TX, USA, 2004.2014-2018] a kind of algorithm based on the monitoring circle proposed in the article, this algorithm utilization monitoring wavelength obtains the state information of circle uplink, namely utilize monitoring circle (M-cycle) to collect alarm, can realize quick fault location.The fault location algorithm accurate location that can realize link failure based on the monitoring wavelength, but need special monitoring wavelength when obtaining warning information, caused extra resource overhead, and every link must satisfy the requirement that a plurality of monitoring wavelength cover, the location cost is higher.

Because the business of transmitting in the network can reflect the state of link, utilize the professional warning information that obtains Link State, and realize fault location successively, this type of algorithm is called as the fault location algorithm based on business.Location algorithm based on business can be saved locating resource, has reduced the location expense.People such as Xiaohui Huang are at " Fault management for Internet service:Modeling and algorithms " [IEEE International Conference on Communications (ICC) [C] .Istanbul, Turkey, 2006.854-859] during with link failure in the network the professional sign that produces as alarm, how much represent the size of this fault probability of happening with the alarm number relevant with fault, namely the failure location discriminant parameter is the fault coverage.Be considered as the most possible fault that takes place by the fault with the coverage maximum, determine that it is the root fault, thereby realize fault location.This algorithm has been saved the location expense, but owing to do not have complete corresponding relation between fault coverage and the fault probability of happening, there is certain irrationality in parameter-definition, causes positional accuracy lower., open into for this reason, Liao Jianxin, Zhu Xiaomin has proposed a new discriminant parameter in " based on the heuristic fault location algorithm of the doubtful degree of Bayes " [software journal, 2010,21 (10): 2610-2621]---the doubtful degree of Bayes.The a plurality of relevant sign that this parameter expression is out of order is to the otherness of its feedback information, more reasonably reflected the possibility that fault takes place, greatly improved the accuracy of fault location, but owing to do not consider that sign is to the feedback information otherness of different faults, parameter-definition still has certain limitation, and positioning performance is difficult to further be promoted.As seen, use single parameter to carry out fault location and have the defective that positioning accuracy is low, the cost expense is big.Therefore, demand seeking more rational positional parameter urgently, and realize optical network fault location accurately in conjunction with multi-parameter.

Summary of the invention

Technical problem to be solved by this invention is: at the fault location algorithm expense cost high defective of tradition based on the monitoring wavelength, the link-state information that the present invention perceives business is as sign, and realizes fault location according to this, reduces the location expense; At the defective that has now based on the single discriminant parameter positioning performance of the fault location algorithm difference of business, the present invention has designed screening and location that multi-parameter is carried out fault, with the accuracy of further lifting fault location.

The technical solution adopted for the present invention to solve the technical problems is: propose a kind of high-accuracy network fault screening navigation system of multi-parameter perception, comprise failure predication, fault screening and fault location module.Wherein, prediction module utilizes the fault propagation model to obtain the maximum possible failure collection fast.The screening module is by defined parameters fault-signal intensity, and the employing compression sensing method screens the redundant information in the maximum possible failure collection.Locating module is poor by the defined parameters comentropy, and selecting the bigger fault of probability is final fault, and the realization network failure is accurately located.

At first, prediction module is set up causal bipartite graph fault propagation model between reflection fault and the sign, finds out all faults relevant with the sign that occurs in the network fast according to this model, and adding maximum possible failure collection.The maximum possible failure collection is defined as the set of all contingent faults in the network that the prediction module fast prediction has herein.Its true fault proportion is less, and redundant erroneous judgement fault ratio is higher.

Then, the screening module is screened the element of maximum possible failure collection, keeps the bigger fault of possibility occurrence.The failure definition signal strength signal intensity is represented the possibility that fault takes place, and fault is considered as signal, and the size of fault coverage is the intensity level of fault-signal.Because compression sensing method can effectively reduce the signal dimension under the situation of stick signal important information, therefore, can adopt compression sensing method that the fault-signal after transforming is handled, obtain the lower low-dimensional reconstruction signal of redundancy.Then the signal that reconstructs is reduced to corresponding failure collection, can obtains the littler possible breakdown set of containing element number, realized the screening of failure collection.Possible breakdown element of set prime number order after the screening is more little, and the element of influence location is more few, just more easy realization accurate localization.

At last, the size of in locating module, representing the fault possibility occurrence according to parameter information entropy difference.Failure definition ideally with actual conditions under the difference of comentropy be that the comentropy of fault is poor.The comentropy difference is more little, and actual conditions are more near ideal situation, i.e. the possibility of fault generation is more big.Wherein, ideally the sign set comprises whole signs of fault correspondence, and the sign set is the actual all indications that obtains under the actual conditions.To the element in the set of the possible breakdown after the screening, according to its comentropy difference from small to large, select the failure collection of most possible explanation gained sign as positioning result, realize fault location.

A kind of high-accuracy network fault screening navigation system of multi-parameter perception, comprise failure predication, fault screening and three modules of fault location, set up the bipartite graph model of network as the fault propagation model, prediction module utilizes the fault propagation model to obtain maximum possible failure collection in the networking; The screening module is screened the element of maximum possible failure collection, obtains the failure reconfiguration signal and sets up failure collection H _SlectedThe failure collection H that locating module obtains the screening module according to the comentropy difference _SlectedHandle, obtain final fault location result.

Obtain that the maximum possible failure collection is specially in the networking: set up the bipartite graph model of network, the setting-up time length of window is periodically obtained the sign S set in the network _N, to the sign S set in the single window _NIdentify, to s _i∈ S _N, find out relative all failure collection F (s _i), call formula:

Obtain maximum possible failure collection H _Max

The screening module is with maximum possible failure collection H _MaxBe converted into primary signal X, to failure collection H _Max={ f ₁, f ₂..., f _nIn all elements, according to formula:

Determine the coverage of fault successively, with all C (f _i) to form data sequence be primary signal X, wherein f _iBe set H _MaxIn i element, n is the number of data among the signal X; Obtain the measured value y of signal X according to the perception matrix A according to formula: y=AX, the signal that has minimum nonzero element number in the measured value is reconstruction signal, sets up the failure collection H of reconstruction signal _SlectedThe determining of described comentropy difference specifically comprises: according to formula: $H_1\left(f\right)=-{\mathrm{\&Sigma;}}_{s_i\&Element;S\left(f\right)\&cap;S_O}p\left(f\right|s_i)\log p\left(f\right|s_i)$ Calculate desirable comentropy H ₁(f), according to formula: $H_2\left(f\right)=-{\mathrm{\&Sigma;}}_{s_i\&Element;S\left(f\right)\&cap;S_N}p\left(f\right|s_i)\log p\left(f\right|s_i)$ Calculate the sign S set _NThe comentropy H that fault f is provided ₂(f); According to formula: △ H (f)=H ₁(f)-H ₂(f) the comentropy difference △ H (f) of calculating fault f; Wherein, S (f) represents the set of the sign that all are relevant with fault f, S _OThe set of all signs that can be observed in the expression network.Described perception matrix A is diagonal matrix, and concrete construction method is, according to formula α _SI=μ Max{x ₁, x ₂..., x _nCalculated threshold α _SI, wherein, μ (0≤μ≤1) is scale factor, according to formula: $a_i=\left\{\begin{array}{cc}1& x_i\&GreaterEqual;{\mathrm{\&alpha;}}_{\mathrm{SI}}\\ 0& x_i<{\mathrm{\&alpha;}}_{\mathrm{SI}}\end{array}\right.$ Determine the diagonal element a of perception matrix A _i, A=diag (a ₁, a _a..., a _n).Locating module takes out HS successively _LectedIn fault f _i(f _i∈ H _Slected), if

Then think fault f _iTake place, according to formula $S_{\mathrm{explained}}=S\left(f_i\right)\&cup;S_{\mathrm{explained}},$ $H^{\&prime;}=H\&cup;\left\{f_i\right\}$ Upgrade S set _ExplainedAnd H; To H _SlectedAfter middle fault was upgraded one by one, obtaining the result was final fault location result.

The present invention also proposes a kind of high-accuracy network fault screening localization method of multi-parameter perception, it is characterized in that set up the bipartite graph model of network as the fault propagation model, prediction module utilizes the fault propagation model to obtain maximum possible failure collection in the networking; The screening module is screened the element of maximum possible failure collection, obtains the failure reconfiguration signal and sets up failure collection H _SlectedThe failure collection H that locating module obtains the screening module according to the comentropy difference _SlectedHandle, obtain final fault location result.

The present invention adopts business to obtain the sign information of link, compares the location algorithm based on the monitoring wavelength, in the sign information access process, has saved resource overhead greatly, has reduced the location cost; Before fault location, utilize the parametic fault coverage that the maximum possible failure collection is converted into signal, and introduce compression sensing method signal is handled, because the failure collection of the signal correspondence after the reconstruct has less element number, realized the preliminary screening of possible breakdown.Further realized fault location according to parametic fault comentropy difference.Utilize the advantage of different parameters, break the limitation of single parameter, realize the more fault location of high-precision low cost.

Description of drawings

Fig. 1 fault location flow chart of the present invention;

Fig. 2 fault propagation model---bipartite graph model example figure.

Embodiment

Below at accompanying drawing and instantiation the present invention is further described in detail.

The present invention proposes a kind of high-accuracy network fault screening navigation system of multi-parameter perception, comprises failure predication, fault screening and fault location module.Wherein, prediction module utilizes the fault propagation model to obtain the maximum possible failure collection fast.The screening module is by defined parameters fault-signal intensity, and the employing compression sensing method screens the redundant information in the maximum possible failure collection.Locating module is poor by the defined parameters comentropy, and selecting the bigger fault of probability is final fault, and the realization network failure is accurately located.

Fig. 1 is fault location flow chart of the present invention.Below will introduce respectively according to flow chart shown in Figure 1.

One, failure predication module

Set up the fault propagation model, prediction module utilizes the fault propagation model to obtain maximum possible failure collection in the networking.

Adopt the bipartite graph model to set up the fault propagation model, the causality in the expression network between all faults and all indications, Fig. 2 is an example of bipartite graph model.In the failure predication module, set up the bipartite graph model of network, the setting-up time length of window is periodically obtained the sign S set in the network _N.To the sign S set in the single window _N, if S _NBe empty set, think that then network does not break down in this window; If S _NBe not empty set, to s _i∈ S _N, form set F (s according to finding out relevant all faults with it _i).Call formula:

$H_{\mathrm{Max}}={\&cup;}_{i=1}^{\left|S_N\right|}F\left(s_i\right)---\left(1\right)$

Ask for all F (s _i) the union of sets collection, be maximum possible failure collection H _Max

Two, fault screening module

The screening module is screened the element of maximum possible failure collection, keeps the bigger fault of possibility occurrence.Concrete steps are as follows:

Step 1: with the maximum possible failure collection H that asks for _MaxBe converted into primary signal X, H _Max={ f ₁, f ₂..., f _n, X=(x ₁, x ₂..., x _n).

According to formula: $C\left(f\right)=\left|s\right|s\&Element;d\left(f\right)\&cap;S_N\left\}\right|---\left(2\right)$

Determine the coverage of fault f.Wherein, the set of all indications of d (f) expression fault f is to f ∈ H _Max, ask for coverage C (f) successively, all C (f) are formed data sequence, this sequence is primary signal X, x _i=C (f _i).

Step 2: utilize compression sensing method processing signals X.Mainly comprise: set up the perception matrix, obtain signal measured value and reconstruction signal.

At first, set up the perception matrix A, A is diagonal matrix, A=diag (a ₁, a ₂..., a _n), wherein n is the number of data among the signal X.

Set scale factor μ (0≤μ≤1), obtain threshold alpha according to formula (3) _SIObtained the diagonal element a of perception matrix A by formula (4) _iValue, so far finished the perception matrix design.

α _SI=μ·Max{x ₁,x ₂,…,x _n} (3)

$a_i=\left\{\begin{array}{cc}1& x_i\&GreaterEqual;{\mathrm{\&alpha;}}_{\mathrm{SI}}\\ 0& x_i<{\mathrm{\&alpha;}}_{\mathrm{SI}}\end{array}\right.---\left(4\right)$

Then, obtain the measured value y of signal X according to the perception matrix A.Measured value y can be calculated by formula (5).

y=AX (5)

At last, reconstruct signal X ＇ by measured value y.As shown in Equation (6), in all signals that satisfy formula (5), the signal with minimum nonzero element number is the reconstruction signal of asking, and the signal X ＇ nonzero element number that namely reconstructs is minimum.

min||X＇|| ₀ s.t. y=AX (6)

Step 3: reconstruction signal is reduced to failure collection, the failure collection H after obtaining screening _Slected

To x _iIf ∈ X ＇ is x _i≠ 0, then with H _MaxMiddle corresponding f _iAdd set H _SlectedBecause the nonzero element number is minimum in the reconstruction signal, therefore, H _SlectedElement number be not more than H _MaxElement number, realized the screening of fault.

Three, fault location module

In the fault location module, locating module is represented the size of fault possibility occurrence according to parameter information entropy difference, the failure collection H that the screening module is obtained _SlectedHandle, obtain final positioning result H.Below be key step:

Step 1, parameter information entropy difference △ H (f) expresses the possibility size that fault f takes place.For analysis of failure comentropy difference △ H (f), define following two parameters: desirable comentropy H ₁(f) and actual information entropy H ₂(f).When sign takes place in the network, will provide certain amount of information to relative fault.Comentropy refers to the average information that a plurality of signs relevant with certain fault provide this fault.

The desirable comentropy H of fault f ₁(f) refer in ideal conditions, namely under the situation that all indications relevant with fault f all takes place, the comentropy of fault f.Can be according to formula:

$H_1\left(f\right)=-{\mathrm{\&Sigma;}}_{s_i\&Element;S\left(f\right)\&cap;S_O}p\left(f\right|s_i)\log p\left(f\right|s_i)---\left(7\right)$

Calculate desirable comentropy H ₁(f).

Wherein, S (f) represents the set of the sign that all are relevant with fault f, S _OThe set of all signs that can be observed in the expression network, p (f|s _i) expression sign s _iCan explain the probability of fault f, can be that formula (8) draws by Bayesian formula.

$p\left(f\right|s_i)=\frac{p\left(f\right)p\left(s_i\right|f)}{{\mathrm{\&Sigma;}}_{f_j\&Element;F}p\left(f_j\right)p\left(s_i\right|f_j)}---\left(8\right)$

Wherein, the probability that p (f) expression fault f takes place, p (s _i| f) expression fault f causes sign s _iThe probability that takes place, p (f _j) expression fault f _jThe probability that takes place, p (s _i| f _j) expression fault f _jCause sign s _iThe probability that takes place.

The actual information entropy H of fault f ₂(f) be under the situation of reality, collect the sign S set _NThe comentropy that fault f is provided.According to formula:

$H_2\left(f\right)=-{\mathrm{\&Sigma;}}_{s_i\&Element;S\left(f\right)\&cap;S_N}p\left(f\right|s_i)\log p\left(f\right|s_i)---\left(9\right)$

Calculate H ₂(f).Wherein, p (f|s _i) expression sign s _iCan explain the probability of fault f, can be drawn by formula (8).

The comentropy difference △ H (f) of fault f can be drawn by formula (10), and the comentropy difference △ H (f) of fault f is more little, shows the sign of the actual initiation of fault f with ideally more approaching, can think that the possibility of fault f generation is more big.

△H(f)=H ₁(f)-H ₂(f) (10)

Step 2: to f ∈ H _Slected, ask for comentropy difference △ H (f) one by one, and with H _SlectedMiddle fault is arranged from small to large according to the entropy difference.

Step 3: take out H successively _SlectedIn fault f _i(f _i∈ H _Slected), failure judgement f _iWhether can explain new sign.If S _ExplainedBe the sign set that failure collection H can explain, S (f _i) be and fault f _iRelevant sign set.If f _iCan explain new sign, namely

Then think fault f _iTake place, upgrade S set _ExplainedAnd H, S _Explained=S (f _i) ∪ S _Explained, H=H ∪ { f _i; If

Then think fault f _iSign is explained not contribution, i.e. fault f _iDo not take place.

Step 4: to H _SlectedAfter middle fault is handled one by one, can obtain fault location H as a result.Because fault can be explained all indications in the network among the H, and its comentropy difference △ H (f) value is less to be that the possibility of its generation is bigger, so H is the failure collection that most probable takes place in the network that set is inferred according to sign, namely H is the fault location result of gained.

The present invention utilizes multi-parameter respectively fault to be carried out Preliminary screening and further location, through failure predication, fault screening and fault location, has realized fault location.Improved the fault location accuracy.

Claims (11)

1. the high-accuracy network fault of a multi-parameter perception is screened navigation system, comprise failure predication, fault screening and three modules of fault location, it is characterized in that, set up the bipartite graph model of network as the fault propagation model, prediction module utilizes the fault propagation model to obtain maximum possible failure collection in the networking; The screening module is screened the element of maximum possible failure collection, obtains the failure reconfiguration signal and sets up failure collection H _SlectedThe failure collection H that locating module obtains the screening module according to the comentropy difference _SlectedHandle, obtain final fault location result.

2. fault according to claim 1 screening navigation system is characterized in that, obtains that the maximum possible failure collection is specially in the networking: set up the bipartite graph model of network, the setting-up time length of window is periodically obtained the sign S set in the network _N, to the sign S set in the single window _NIdentify, to s _i∈ S _N, find out relative all failure collection F (s _i), call formula

Obtain maximum possible failure collection H _Max

3. fault screening navigation system according to claim 1 is characterized in that the screening module is with maximum possible failure collection H _MaxBe converted into primary signal X, to failure collection H _Max={ f ₁, f ₂..., f _nIn all elements, according to formula C (f _i)=s|s ∈ d (f _i) ∩ S _NDetermine the coverage of fault successively, with all C (f _i) to form data sequence be primary signal X, wherein f _iBe set H _MaxIn i element, n is the number of data among the signal X; Obtain the measured value y of signal X according to the perception matrix A according to formula y=AX, the signal that has the minimum non-zero element number in the measured value is reconstruction signal, sets up the failure collection H of reconstruction signal _Slected

4. fault screening navigation system according to claim 1 is characterized in that the definite of described comentropy difference specifically comprises: according to formula $H_1\left(f\right)=-{\mathrm{\&Sigma;}}_{s_i\&Element;S\left(f\right)\&cap;S_O}p\left(f\right|s_i)\log P\left(f\right|s_i)$ Calculate desirable comentropy H ₁(f), according to formula $H_2\left(f\right)=-{\mathrm{\&Sigma;}}_{s_i\&Element;S\left(f\right)\&cap;S_N}p\left(f\right|s_i)\log p\left(f\right|s_i)$ Calculate the sign S set _NThe comentropy H that fault f is provided ₂(f); According to formula △ H (f)=H ₁(f)-H ₂(f) the comentropy difference △ H (f) of calculating fault f; Wherein, S (f) represents the set of the sign that all are relevant with fault f, S _OThe set of all signs that can be observed in the expression network, S _NRepresent the set of all signs of collecting in the network in the time window, p (f|s _i) expression sign s _iCan explain the probability of fault f.

5. fault screening navigation system according to claim 3 is characterized in that described perception matrix A is diagonal matrix, and concrete construction method is, according to formula α _SI=μ Max{x ₁, x ₂,, x _nCalculated threshold α _SI, wherein, μ (0≤μ≤1) is scale factor, according to formula $a_i=\left\{\begin{array}{cc}1& x_i\&GreaterEqual;{\mathrm{\&alpha;}}_{\mathrm{SI}}\\ 0& x_i<{\mathrm{\&alpha;}}_{\mathrm{SI}}\end{array}\right.$ Determine the diagonal element a of perception matrix A _i, A=diag (a ₁, a ₂..., a _n).

6. the high-accuracy network fault of multi-parameter perception screening localization method is characterized in that set up the bipartite graph model of network as the fault propagation model, prediction module utilizes the fault propagation model to obtain maximum possible failure collection in the networking; The screening module is screened the element of maximum possible failure collection, obtains the failure reconfiguration signal and sets up failure collection H _SlectedThe failure collection H that locating module obtains the screening module according to the comentropy difference _SlectedHandle, obtain final fault location result.

7. fault according to claim 6 screening localization method is characterized in that, obtains that the maximum possible failure collection is specially in the networking: set up the bipartite graph model of network, the setting-up time length of window is periodically obtained the sign S set in the network _N, to the sign S set in the single window _NIdentify, to s _i∈ S _N, find out relative all failure collection F (s _i), call formula:

Obtain maximum possible failure collection H _Max

8. fault screening localization method according to claim 6 is characterized in that the screening module is with maximum possible failure collection H _MaxBe converted into primary signal X, to failure collection H _Max={ f ₁, f ₂..., f _nIn all elements, according to formula: C (f _i)=s|s ∈ d (f _i) ∩ S _NDetermine the coverage of fault successively, with all C (f _i) to form data sequence be primary signal X, wherein f _iBe set H _MaxIn i element, n is the number of data among the signal X; Obtain the measured value y of signal X according to the perception matrix A according to formula: y=AX, the signal that has minimum nonzero element number in the measured value is reconstruction signal, sets up the failure collection H of reconstruction signal _Slected

9. fault according to claim 6 screening localization method is characterized in that, the determining of described comentropy difference specifically comprises: according to formula: $H_1\left(f\right)=-{\mathrm{\&Sigma;}}_{s_i\&Element;S\left(f\right)\&cap;S_O}p\left(f\right|s_i)\log p\left(f\right|s_i)$ Calculate desirable comentropy H ₁(f), according to formula: $H_2\left(f\right)=-{\mathrm{\&Sigma;}}_{s_i\&Element;S\left(f\right)\&cap;S_N}p\left(f\right|s_i)\log p\left(f\right|s_i)$ Calculate the sign S set _NThe comentropy H that fault f is provided ₂(f); According to formula: △ H (f)=H ₁(f)-H ₂(f) the comentropy difference △ H (f) of calculating fault f; Wherein, S (f) represents the set of the sign that all are relevant with fault f, S _OThe set of all signs that can be observed in the expression network, S _NRepresent the set of all signs of collecting in the network in the time window, p (f|s _i) expression sign s _iCan explain the probability of fault f.

10. fault screening localization method according to claim 1 is characterized in that, takes out H successively _SlectedIn fault f _i(f _i∈ H _Slected), if

Then think fault f _iTake place, according to formula S _Explained=S (f _i) ∪ S _Explained, H'=H ∪ { f _iThe renewal S set _ExplainedAnd H; To H _SlectedAfter middle fault was upgraded one by one, obtaining the result was final fault location result, wherein, and S (f _i) be and fault f _iRelevant sign set, S _ExplainedFor the sign that failure collection H explains is gathered.

11. fault screening localization method according to claim 8 is characterized in that described perception matrix A is diagonal matrix, concrete construction method is, according to formula α _SI=μ Max{x ₁, x ₂..., x _nCalculated threshold α _SI,Wherein, μ (0≤μ≤1) is scale factor, according to formula: $a_i=\left\{\begin{array}{cc}1& x_i\&GreaterEqual;{\mathrm{\&alpha;}}_{\mathrm{SI}}\\ 0& x_i<{\mathrm{\&alpha;}}_{\mathrm{SI}}\end{array}\right.$ Determine the diagonal element a of perception matrix A _i, A=diag (a ₁, a ₂..., a _n).

Images