CN102902623B - Implementation method for test optimization of complex system - Google Patents

Abstract

The invention discloses an implementation method for the test optimization of a complex system in the technical field of test diagnosis in equipment production and maintenance process. The method comprises the steps of: initializing a system to be tested; preprocessing a fault diagnosis matrix labels.MC of the system to be tested; judging whether a set, namely labels-set, is an empty set, if the set, namely labels-set, is not the empty set, selecting a node as a current node from the set, namely labels-set, and preprocessing the current node; calculating the upper limit new-U of the current node; judging whether the current node is a leaf node, if so, processing other nodes in the set labels-set; otherwise, dividing the current node into two child nodes, and adding the two divided child nodes into the set, namely labels-set, and deleting the current node from the set labels-set, and then, processing other nodes in the set labels-set. The implementation method ensures that the optimal test points can be searched, and the high speed is satisfied.

Description

A kind of complication system tests preferred implementation method

Technical field

The invention belongs to the testing and diagnosing technical field in equipment production and maintenance process, particularly relate to a kind of complication system and test preferred implementation method.

Background technology

Along with the development of science and technology, modern equipment or constituent system components get more and more, and function becomes increasingly complex.In system, physics measuring point is numerous, and test approaches is various.And want to carry out fault diagnosis to system complicated like this or equipment, must first test, test is the important means obtaining diagnosis information needed.There is no complete test, sufficient quantity of information can not be obtained, be then difficult to the accuracy ensureing fault diagnosis.But gather in the face of so huge feasible test, how to select the test of minimal number, to meet the requirement of fault diagnosis, be diagnose before the major issue that must solve.If effective test selection can be completed before the diagnosis, choose the blindness of test when both can overcome diagnosis, test number can be reduced again, thus save testing cost, higher diagnostic accuracy can also be ensured.

Test method for optimizing is a lot, and current typical method has:

1) based on information-theoretical method: the basic ideas of the method determine the evaluation index of a test point, such as information entropy E (j), measuring point quantity of information ICN, information content I (y) etc., the mode of then single step forward direction optimizing on this basis selects local optimum test point gradually.Wherein most widely used when number information entropy.

Based on the test point optimization algorithm of information entropy by document 1 (Starzyk, J.A.; Dong Liu; Zhi-Hong Liu; Et.al., Entropy-based optimum test points selection for analog fault dictionary techniques [J] Instrumentation and Measurement, IEEE Transactions on, Volume 53, Issue 3, June 2004Page (s): 754 – 761) propose.Its dominant ideas evaluate the segregate possibility of each fault according to each cardinality of a fuzzy set.Suppose for certain test point n _jthere is k the fuzzy set do not overlapped each other, and be and test point n _jwith the fuzzy set that integer coding i is associated the number of middle fault.Fault is by from ambiguity group the probability of keeping apart is roughly wherein f is all number of defects listed in dictionary.Suppose that test point is separate, and the possibility that each fault occurs is impartial, then for certain specific test point n _j, information entropy E (j) is defined as

$E\left(j\right)=\underset{i=1}{\overset{f}{\mathrm{\Σ}}}{F}_{\mathrm{ij}}\log F_{\mathrm{ij}}$

Because in given dictionary, out of order number f be fixing, in above formula, certain fc-specific test FC point n _jinformation content I (j) can maximize along with minimizing of entropy E (j).If have minimum value E (j) test point n _jbe added to and expect test point set N _optin, then this act will ensure that N _optthe maximum increasing degree of middle information.Therefore comprise (inclusive) strategy and ensure that each stage selected in test point can obtain maximum fault diagnosis.

Document 2 (Pinjala, K.K.; Kim, B.C.; An approach for selection of test points for analog fault diagnosis [C], Defect and Fault To lerance in VLSI Systems, the ICN of in definition, it be in essence test point n 2003.Proceedings.18th IEEE International Symposium on 3-5Nov.2003 Page (s): 287-294) _jthe number of faults q that can isolate _j.Carrying out, in test point selection course, getting max _j(q _j), namely this measuring point of the larger explanation of I CN is better.

Document 3(Chenglin Yang, et al; Application of Heuristic Graph Search to Test-Point Selection for Analog Fault Dictionary Techniques. [J], IEEETRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT.2011) quantity of information I (y) that fault set y comprises is proposed, it is a distortion of ICN index in essence.

But have some limitations based on information-theoretical test point method for optimizing: on the one hand, based on the evaluation index of the various measuring points that information theory defines, be all only describe measuring point based on the thought of probability, not exclusively conform to the true isolated fault ability of this measuring point, often there is deviation.In addition, these class methods are all the thought optimized forward based on single step, and each selection optimum test point of what is called is added in optimum test point set, can only ensure local optimum, can not ensure global optimization.Therefore, the choosing result and often can not obtain optimum solution of this class methods, and usually only obtain suboptimal solution or near-optimum solution.

2) based on graph search algorithm

Document 3 and document 4(poplar such as to grow into forest at minimum test point set preferred new algorithm [J] Chinese journal of scientific instrument the 29th volume the 12nd phase based on heuristic graph searching, 2008,12.2497-2503 pages) measuring point selection course is grouped into the expansion process of node of graph.In a digraph, the problem that root nodes stand is to be solved, then select each measuring point to expand successively, expansion process is based on A* algorithm realization.Basic thought is by setting suitable heuristic function, the cost value of each expanded search node of comprehensive assessment, by the size of more each expanding node cost value, selects most promising point to be expanded, until find destination node.If there is certain node y, wherein not non-isolated fault, then show that the solution of initial problem obtains.The measuring point that path from node y to root node S is used constitutes minimum measuring point collection, i.e. optimum solution.

The expansion process of node of graph still follows rule 1 Sum fanction 2 in document 3.Those measuring points of the measuring point used when only having sequence number to be greater than from his father's point spread to present node of rule 1 regulation number just can do available measuring point, guarantee the generation avoiding redundant node in node of graph expansion process, thus improve search efficiency.Rule 2 regulation is when the available measuring point expanding certain node of graph is less than the lower bound needing measure-point amount, and fault necessarily can not be completely isolated, and path can not obtain optimum solution thus.Rule 2 further reduces the expanding node of node of graph.

A* algorithm is a kind of intelligent search algorithm, can ensure the convergence of globally optimal solution in theory.Change sentence to say, utilize A* algorithm one to find an optimal path surely in theory, namely one find minimum measuring point collection surely.Minimum measuring point collection is made up of the measuring point on from root node to destination node.

But this kind of method Problems existing is when heuristic function f (x) is equal, A* algorithm has been absorbed in dies for the sake of honour a little.Now, adopt information entropy to carry out distinguishing then with certain subjectivity, test point can be caused in practical operation to select inaccurate, and introduce multiple attribute decision making (MADM) to evaluate and distinguish each node, then algorithm complex can be caused to increase.Many attributes are also evaluate from information theory view in addition, equally also can cause the inaccurate problem of the selection of evaluation.

3) based on intelligent algorithms such as GA

GA based on single object optimization realizes the optimum option of test point, be by people such as Golonek in 2007 at document 5(Golonek, T.; Rutkowski, J.; Genetic-Algorithm-Based Method for Optimal Analog Test Points Selection, [J], Circuits and Systems II:Express Briefs, IEEE Transactions on Volume 54, Issue 2, Feb.2007Page (s): 117 – 121) middle proposition.Author on this basis, at document 6(international journal of modeling, identification and control) propose multiple goal GA optimized algorithm to realize the optimum option of test point.The basic thought of these class methods be by test point according to monoploid principle to determine chromosome, then stochastic generation initial population, repeat to select for this population, intersect, the basic biological evolution operation such as variation, after finally reaching certain evolution iterations, select to determine optimum test point set.

But these class methods also exist following deficiency: a) determine to have employed haploid principle due to chromosomal, thus when system scale increases, chromosome length is also thereupon elongated.Usually, after total test point number is more than 30, chromosomal length will directly cause algorithm travelling speed very slow, thus affects efficiency of algorithm.When being applied to scale complex system, efficiency of algorithm is very low, thus causes its practicality poor; B) in addition, due to the random nature of intelligent algorithm, can not ensure can find optimum solution, what many times find is suboptimal solution at every turn.

The common limitation of above-mentioned three class methods can not ensure necessarily to find Minimum test set to close.On this basis, the present invention proposes based on Branch-and-Bound Algorithm, can guarantee to find a Minimum test set to close, and algorithm speed is very fast.

Summary of the invention

The object of the invention is to, propose a kind of complication system and test preferred implementation method, for solving existing complication system test method for optimizing Problems existing.

To achieve these goals, the technical scheme that the present invention proposes is that a kind of complication system tests preferred implementation method, it is characterized in that described method comprises:

Step 1: treat examining system and carry out initialization;

Setting present node correspondence treats that the fault diagnosis matrix of examining system is labels.MC, the test point set selected at present node place is labels.T1, present node place determines that the test point set do not selected is labels.T2, and pending node set is labels_set;

Step 2: treat examining system fault diagnosis matrix labels.MC and carry out pre-service;

If more than two row in system fault diagnosis matrix to be measured and two row identical, then the fault corresponding to it is claimed to be of equal value; In fault of equal value, only need retain a representing fault, all the other faults are then defined as redundant fault, delete redundant fault;

If two row and two in system fault diagnosis matrix to be measured arrange above identical, then claim the test corresponding to it to be of equal value, in equivalence test, only need retain a representative and test, all the other tests are then defined as redundancy testing, delete redundancy testing;

Step 3: judge whether set is null set for labels_set, if set labels_set is null set, then terminates; Otherwise, perform step 4;

Step 4: select a node as present node from set labels_set, pre-service is carried out to present node, comprising:

41) by present node institute essential test point and must delete test point add to respectively set labels.T1 and set labels.T2 in;

42) try to achieve make set labels.T1 still can not independent isolating diagnosis fault ambiguity group, then for each ambiguity group, if present node remains selectable test point, namely present node correspondence is treated in each row of fault diagnosis matrix labels.MC of examining system, there are row, make in this ambiguity group two faults only have these row to be difference, then the test point corresponding to these row necessarily should be selected in set labels.T1;

43) if present node correspondence treat the fault diagnosis matrix labels.MC of examining system exist each element be entirely 0 or be entirely 1 row, then by these row add to set labels.T2 in, namely these row to diagnosis without any meaning, deleted;

Step 5: calculate the upper limit new_U of present node, specifically finds out the test that in the test that present node comprises, diagnosis capability is the strongest and joins in set labels.T1; The test that described diagnosis capability is the strongest refers to that this test point adds the test in labels.T1, its can exclusive diagnosis isolation the number of defects maximum;

Step 6: judge whether present node is leaf node, if present node is leaf node, then returns step 3; Otherwise, perform step 7;

Step 7: present node is split into two child nodes, comprises:

71) from the remaining test point of present node, an optional test point, is added in set labels.T1, and in current sub-system labels.MC, is deleted the row corresponding to this test point, obtained a split vertexes;

72) this optional test point is joined in set labels.T2, and in matrix labels.MC, delete the row corresponding to this test point, thus obtain another node after dividing;

73) add in set labels.set dividing two nodes obtained, and therefrom delete present node;

74) step 3 is returned.

Described judge present node be whether leaf node specifically, when meeting following condition for the moment, present node is leaf node;

1) if set labels.T1 and set labels.T2 comprises all test points;

2) if the lower limit of present node exceedes the upper limit of current record;

3) if set labels.T1 is efficient solution;

4) if the set labels.T1 ambiguity group of establishing is not diagnosable;

5) if the lower limit of present node is greater than the row rest_tp of matrix labels.MC.

The computing method of the lower limit of described present node are, first with the number of defects that the independent failure total number deduction set labels.T1 in primal system can diagnose, namely obtain the fault F that present node waits to diagnose _m; Again in matrix labels.MC, in the row matrix corresponding to the fault set waiting to diagnose, calculate the number of the different elements that each test point comprises, wherein maximum one is decided to be N _m; Last according to formula calculate the lower limit of present node.

Described set labels.T1 is the decision method of efficient solution: if set labels.T1 all covers the fault of isolation diagnostic, then gathering labels.T1 is efficient solution; Otherwise set labels.T1 is not efficient solution.

The not diagnosable decision method of ambiguity group that described set labels.T1 establishes is: first determine that present node still can not by the ambiguity group of the independent completely isolated fault opened according to set labels.T1, then for matrix labels.MC, check whether each ambiguity group exists not diagnosable situation successively, the row vector that whether there are at least two row during the matrix labels.MC namely corresponding to ambiguity group fault is capable is identical; If existed, then the ambiguity group that set labels.T1 establishes is not diagnosable, otherwise the ambiguity group that set labels.T1 establishes is diagnosable.

The present invention considers that measuring point selects the relation between fault diagnosis, devises the computing method of upper and lower bound, thus has both ensured that branch and bound method had enough search capabilities, have and can delete invalid node in time, thus avoid excessive search; The method is guaranteed to search optimum test point, and has satisfied rapidity, can be applied to large-scale complicated system.

Accompanying drawing explanation

Fig. 1 is the general type of fault diagnosis matrix;

Fig. 2 is that complication system tests preferred implementation method process flow diagram;

Fig. 3 is leapfrog Filter Principle figure;

Fig. 4 is the fault diagnosis Matrix List of leapfrog wave filter;

Fig. 5 is the application process of node 0 place rule 2;

Fig. 6 is the process asking upper limit measuring point collection;

Fig. 7 is the search node figure that wave filter is complete;

Fig. 8 is the subsystem fault diagnostic matrix labels.MC schematic diagram that each node searched for by wave filter.

Embodiment

Below in conjunction with accompanying drawing, preferred embodiment is elaborated.It is emphasized that following explanation is only exemplary, instead of in order to limit the scope of the invention and apply.

Method provided by the invention realizes for fault diagnosis table.As shown in Figure 1, the corresponding different faults of row of matrix, row correspond to different test to the concrete form of so-called fault diagnosis table.In matrix, element is the fault that its test arranging correspondence of 1 expression can detect corresponding to its row, and the test for its row correspondence of 0 expression can not detect the fault corresponding to its row.

Fig. 2 is that complication system tests preferred implementation method process flow diagram.In Fig. 2, complication system is tested preferred implementation method and is comprised:

Step 1: treat examining system and carry out initialization.

Original problem definition to be solved is become start node labels={C, Φ, Φ }, wherein C is original system fault diagnosis matrix to be solved.It should be noted that, represented by each nodal information labels in branch-and-bound search procedure in this invention, it comprises three partial information labels={MC, T ₁, T ₂}: setting present node correspondence treats that the fault diagnosis matrix of examining system is labels.MC, the test point set selected at present node place is labels.T1, present node place determines that the test point set do not selected is labels.T2, and pending node set is labels_set.Setting initial ranging higher limit is U, and its initial value is infinitely great.Setting solution deposits the efficient solution that current search arrives.

Step 2: treat examining system fault diagnosis matrix labels.MC and carry out pre-service.

If more than two row in system fault diagnosis matrix to be measured and two row identical, then the fault corresponding to it is claimed to be of equal value; In fault of equal value, only need retain a representing fault, all the other faults are then defined as redundant fault, delete redundant fault.

If two row and two in system fault diagnosis matrix to be measured arrange above identical, then claim the test corresponding to it to be of equal value, in equivalence test, only need retain a representative and test, all the other tests are then defined as redundancy testing, delete redundancy testing.

Step 3: judge whether set is null set for labels_set, if set labels_set is null set, then terminates; Otherwise, perform step 4.

Step 4: select a node as present node from set labels_set, pre-service is carried out to present node, comprising:

41) by present node institute essential test point and must delete test point add to respectively set labels.T1 and set labels.T2 in.

42) try to achieve make set labels.T1 still can not independent isolating diagnosis fault ambiguity group, then for each ambiguity group, if present node remains selectable test point, namely present node correspondence is treated in each row of fault diagnosis matrix labels.MC of examining system, there are row, make in this ambiguity group two faults only have these row to be difference, then the test point corresponding to these row necessarily should be selected in set labels.T1.

43) if present node correspondence treat the fault diagnosis matrix labels.MC of examining system exist each element be entirely 0 or be entirely 1 row, then by these row add to set labels.T2 in, namely these row to diagnosis without any meaning, deleted.

Step 5: calculate the upper limit new_U of present node, specifically finds out the test that in the test that present node comprises, diagnosis capability is the strongest and joins in set labels.T1; The test that described diagnosis capability is the strongest refers to that this test point adds the test in labels.T1, its can exclusive diagnosis isolation the number of defects maximum.

Step 6: judge whether present node is leaf node, if present node is leaf node, then returns step 3; Otherwise, perform step 7.

Judge present node be whether leaf node specifically, when meeting following condition for the moment, present node is leaf node;

1) if set labels.T1 and set labels.T2 comprises all test points;

2) if the lower limit of present node exceedes the upper limit of current record;

3) if set labels.T1 is efficient solution;

4) if the set labels.T1 ambiguity group of establishing is not diagnosable;

5) if the lower limit of present node is greater than the row rest_tp of matrix labels.MC.

Namely be formulated, then have

$\mathrm{test}\_\mathrm{leaf}[\mathrm{\λ},U]\left\{\begin{array}{ccc}\mathrm{true}& \mathrm{if}& T_1\∪T_2=[\mathrm{1,2},...,n]\\ \mathrm{true}& \mathrm{if}& \mathrm{Lower}\_\mathrm{Bound}\left[\mathrm{\λ}\right]\≥U\_\min \\ \mathrm{true}& \mathrm{if}& \mathrm{test}\_\mathrm{solution}\left[\mathrm{\λ}\right]=\mathrm{true}\\ \mathrm{true}& \mathrm{if}& \mathrm{Is}\_\mathrm{diag}\_\mathrm{ambiguity}\left[\mathrm{\λ}\right]=\mathrm{true}\\ \mathrm{ture}& \mathrm{if}& \mathrm{Lower}\_\mathrm{Bound}\left[\mathrm{\λ}\right]>\mathrm{rest}\_\mathrm{tp}\\ \mathrm{false}& \mathrm{otherwise}& \end{array}\right.---\left(1\right)$

The computing method of the lower limit of present node are, first with the number of defects that the independent failure total number deduction set labels.T1 in primal system can diagnose, namely obtain the fault F that present node waits to diagnose _m; Again in matrix labels.MC, in the row matrix corresponding to the fault set waiting to diagnose, calculate the number of the different elements that each test point comprises, wherein maximum one is decided to be N _m; Last according to formula calculate the lower limit of present node.

Set labels.T1 is the decision method of efficient solution: if set labels.T1 all covers the fault of isolation diagnostic, then gathering labels.T1 is efficient solution; Otherwise set labels.T1 is not efficient solution.

The not diagnosable decision method of ambiguity group that set labels.T1 establishes is: first determine that present node still can not by the ambiguity group of the independent completely isolated fault opened according to set labels.T1, then for matrix labels.MC, check whether each ambiguity group exists not diagnosable situation successively, the row vector that whether there are at least two row during the matrix labels.MC namely corresponding to ambiguity group fault is capable is identical; If existed, then the ambiguity group that set labels.T1 establishes is not diagnosable, otherwise the ambiguity group that set labels.T1 establishes is diagnosable.

In above-mentioned formula (1), Is_diag_ambiguity function is implemented as follows: for the determined fault ambiguity group waiting to distinguish further of current selected measuring point labels.T1, if the test in present node labels.MC can not be distinguished it, then this function exports Isdiagn is 1, is leaf node; Otherwise it is 0 that function exports, and is not leaf node.

If new_U < is U, and upper limit measuring point collection is an efficient solution, then replacing solution is upper limit measuring point collection, and replacement U is new_U.

If labels.T1 be an efficient solution and new_U≤U time, when solution is empty, or when in labels.T1, measure-point amount is less than solution, then replacing solution is labels.T1, and replacement U is new_U.

Step 7: present node is split into two child nodes, comprises:

71) from the remaining test point of present node, an optional test point, is added in set labels.T1, and in current sub-system labels.MC, is deleted the row corresponding to this test point, obtained a split vertexes.

72) this optional test point is joined in set labels.T2, and in matrix labels.MC, delete the row corresponding to this test point, thus obtain another node after dividing.

73) add in set labels.set dividing two nodes obtained, and therefrom delete present node.

74) step 3 is returned.

Implementation process of the present invention is elaborated for wave filter.As shown in Figure 3, the integer coding fault diagnosis matrix corresponding to it as shown in Figure 4 for the circuit of wave filter.The fault that its row is corresponding different, row then corresponding different measuring point.The measuring point of matrix element corresponding to this element of 1 expression can detect this fault.Otherwise be 0.

Initialization labels={MC, T ₁, T ₂be primal system, the corresponding primary fault diagnostic matrix of its labels.MC, labels.T1, labels.T2 are sky.U deposits the minimum number of upper limit test corresponding to point set of the current number found, and is initialized as infinity, the corresponding current number corresponding to upper limit measuring point collection connecing node and find of new_U.If new_U < U and new_U is not 0, then U is replaced with new_U.T1 deposits the upper limit measuring point collection of present node.Detailed search node figure as shown in Figure 7.

To the initial matrix application rule 1 of node 0, find that fault 3 and fault 11 liang of row only have the 4th test point different successively.Thus measuring point 4 must be selected into labels.T1, otherwise cannot distinguish fault 3 and fault 11.As a same reason, fault 4 be distinguished and fault 19 must select test point 3, fault 4 be distinguished and fault 21 must select 7, fault 8 be distinguished and fault 11 must select 2, fault 10 be distinguished and fault 12 must select 1.Thus, after application rule 1, obtain labels.T1=[1,2,3,4,7,11], labels.MC is not as added dash area in Fig. 6.Application rule 2, because labels.MC does not now have the row of full 0 or complete 1, thus rule 2 does not carry out any operation.When upper limit measuring point collection is selected, because now labels.T1=[1,2,3,4,7,11] can isolate 22 faults, the fault of not yet keeping apart has 2.For the fault of not yet isolating, consider remaining test point [5,6,8,9,10,12], the diagnostic matrix formed is not as added dash area in Fig. 6, select the measuring point that wherein isolated fault number is maximum, when having the isolated fault ability of multiple measuring point identical simultaneously, then select first in the middle of them, as shown in frame empty in Fig. 6, here select test point 8, can realize whole fault isolation afterwards, thus, upper limit measuring point collection is [1,2,3,4,7,8,11], the upper limit is 7; Then calculation flag yn ₁(whether 1abels.T1 is efficient solution), yn ₂(whether being leaf node).Two marks of this node are respectively 0,0, thus need to divide it.The method of division is the test point selected arbitrarily in labels.MC, is added in labels.T1 and labels.T2, namely performs respectively and is selected into and deletion action.Select measuring point 5 in example, execution is selected into operation and obtains node 1, and perform deletion action and obtain node 2, its subsystem matrix as shown in Figure 8.

For node 1, it indicates yn ₁, yn ₂be respectively 0, now U is updated to 7.Due to yn ₂=0, so need to divide it, select measuring point 10, obtain node 3,4, its subsystem matrix as shown in Figure 8 b.

For node 2, it indicates yn ₁, yn ₂be respectively 0,1, due to lower limit LB=8 > 7 thus this node be defined as leaf node, thus deleted.

For node 3, it indicates yn ₁, yn ₂be respectively 0, due to yn ₂=0, so need to divide it, select measuring point 9, obtain node 5,6, its subsystem matrix as shown in Figure 8 c.

For node 4, it indicates yn ₁, yn ₂be respectively 1,1, thus this node is defined as leaf node, thus deleted.

For node 5, it indicates yn ₁, yn ₂be respectively 0, due to yn ₂=0, so need to divide it, select measuring point 6, obtain node 7,8, its subsystem matrix as shown in figure 8d.

For node 6, it indicates yn ₁, yn ₂be respectively 1,1, thus this node is defined as leaf node, thus deleted.

For node 7, it indicates yn ₁, yn ₂be respectively 0, due to yn ₂=0, so need to divide it, select measuring point 8, obtain node 9,10, its subsystem matrix as figure 8 e shows.

For node 8, it indicates yn ₁, yn ₂be respectively 1,1, due to lower limit LB=8 > 7, thus this node is defined as leaf node, thus deleted.

For node 9, it indicates yn ₁, yn ₂be respectively 0,1 due to Isdiagn=1, namely considers not yet tracing trouble, and residue node can not be distinguished, thus yn ₂=1, so this node is defined as leaf node, thus deleted.

For node 10, it indicates yn ₁, yn ₂be respectively 1,1, thus this node is defined as leaf node, thus deleted.

Hereto, searching algorithm terminates, algorithm search to optimum solution totally three: the T1 being respectively node 0 place, node 4, node 6, the labels.T1 at node 10 place, here the T1 at node 0 place and the labels.T1 at node 10 place is identical, is thus taken as one.

The above; be only the present invention's preferably embodiment, but protection scope of the present invention is not limited thereto, is anyly familiar with those skilled in the art in the technical scope that the present invention discloses; the change that can expect easily or replacement, all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection domain of claim.

Claims (1)

1. the preferred implementation method of system testing, is characterized in that described method comprises:

Step 1: treat examining system and carry out initialization;

Setting present node correspondence treats that the fault diagnosis matrix of examining system is labels.MC, at present node place, oneself the test point set of choosing is labels.T1, present node place determines that the test point set do not selected is labels.T2, and pending node set is labels_set;

Step 2: treat examining system fault diagnosis matrix labels.MC and carry out pre-service;

If more than two row in system fault diagnosis matrix to be measured and two row identical, then the fault corresponding to it is claimed to be of equal value; In fault of equal value, only need retain a representing fault, all the other faults are then defined as redundant fault, delete redundant fault;

If two row and two in system fault diagnosis matrix to be measured arrange above identical, then claim the test corresponding to it to be of equal value, in equivalence test, only need retain a representative and test, all the other tests are then defined as redundancy testing, delete redundancy testing;

Step 3: judge whether set labels_set is null set, if set labels_set is null set, then terminates; Otherwise, perform step 4;

Step 4: select a node as present node from set labels_set, pre-service is carried out to present node, comprising:

41) by present node institute essential test point and must delete test point add to respectively set labels.T1 and set labels.T2 in;

42) try to achieve make set labels.T1 still can not independent isolating diagnosis fault ambiguity group, then for each ambiguity group, if present node remains selectable test point, namely present node correspondence is treated in each row of fault diagnosis matrix labels.MC of examining system, there are row, make in this ambiguity group two faults only have these row to be difference, then the test point corresponding to these row necessarily should be selected in set labels.T1;

43) if present node correspondence treat the fault diagnosis matrix labels.MC of examining system exist each element be entirely 0 or be entirely 1 row, then by these row add to set labels.T2 in, namely these row to diagnosis without any meaning, deleted;

Step 5: calculate the upper limit new_U of present node, specifically finds out the test that in the test that present node comprises, diagnosis capability is the strongest and joins in set labels.T1; The test that described diagnosis capability is the strongest refers to that this test point adds the test in labels.T1, its can exclusive diagnosis isolation the number of defects maximum;

Step 6: judge whether present node is leaf node, if present node is leaf node, then returns step 3; Otherwise, perform step 7;

Judge present node be whether leaf node specifically, when meeting following condition for the moment, present node is leaf node;

1) if set labels.T1 and set labels.T2 comprises all test points;

2) if the lower limit of present node exceedes the upper limit of current record;

The computing method of the lower limit of present node are, first with the number of defects that the independent failure total number deduction set labels.T1 in primal system can diagnose, namely obtain the fault F that present node waits to diagnose _m; Again in matrix labels.MC, in the row matrix corresponding to the fault set waiting to diagnose, calculate the number of the different elements that each test point comprises, wherein maximum one is decided to be N _m; Last according to formula calculate the lower limit of present node;

3) if set labels.T1 is efficient solution;

Set labels.T1 is the decision method of efficient solution: if set labels.T1 all covers the fault of isolation diagnostic, then gathering labels.T1 is efficient solution; Otherwise set labels.T1 is not efficient solution;

4) if the set labels.T1 ambiguity group of establishing is not diagnosable;

The not diagnosable decision method of ambiguity group that set labels.T1 establishes is: first determine that present node still can not by the ambiguity group of the independent completely isolated fault opened according to set labels.T1, then for matrix labels.MC, check whether each ambiguity group exists not diagnosable situation successively, the row vector that whether there are at least two row during the matrix labels.MC namely corresponding to ambiguity group fault is capable is identical; If existed, then the ambiguity group that set labels.T1 establishes is not diagnosable, otherwise the ambiguity group that set labels.T1 establishes is diagnosable;

5) if the lower limit of present node is greater than the row rest_tp of matrix labels.MC;

Step 7: present node is split into two child nodes, comprises:

71) from the remaining test point of present node, an optional test point, is added in set labels.T1, and in current sub-system labels.MC, is deleted the row corresponding to this test point, obtained a split vertexes;

72) this optional test point is joined in set labels.T2, and in matrix labels.MC, delete the row corresponding to this test point, thus obtain another node after dividing;

73) add in set labels_set dividing two nodes obtained, and therefrom delete present node;

74) step 3 is returned.