CN105759782A - Criticality-based vehicle fault diagnosis strategy construction method - Google Patents

Abstract

The invention discloses a criticality-based vehicle fault diagnosis strategy construction method. The method is characterized by comprising the following steps: 1, a heuristic function containing criticality factors is used, a test with single step optimization is selected, the optimization test starting position serves as a test starting node, corresponding test is carried out on the vehicle, and a sensor measurement result is acquired; 2, according to the output value of the sensor measurement result, a fault test correlation matrix is used for decomposing a state set node corresponding to the test output value to serve as the current branch node correlated with the node; and 3, if the branch node is a leaf node, the branch node is marked as an already-solved node, and a fault result is outputted, and if the branch node is a node not solved, the first step is carried out. The corrected relative criticality serves as a weight, an optimization objective function for average test cost is brought forward, and the total number of final intermediate nodes in the diagnosis strategy is reduced.

Description

A kind of car fault diagnosis construction of strategy method based on density of infection

Technical field

The present invention relates to car fault diagnosis technical field, particularly to a kind of car fault diagnosis construction of strategy method based on density of infection.

Background technology

Existing diagnosis policy developing algorithm is substantially made up of heuristic function and search strategy, and heuristic function determines the scope of application of algorithm, and search strategy determines the good and bad degree of the diagnosis policy obtained.Existing algorithm solve the many-valued test tested in reliable situation, multiple faults, multiloop, multi-mode, multistage maintenance, towards correction maintenance with towards the diagnosis policy optimization problem of life cycle management.Existing algorithm is disadvantageous in that: the optimization object function of algorithm only considered test spent time and cost element, to being likely to damage that equipment is caused in diagnosis process and the factor such as negative effect that cycle tests causes lacks and considers；When building diagnosis policy, it is based primarily upon the probability of malfunction attribute of fault mode, and in practical application, when abnormal alarm occurs in system test, during according to the optimal diagnosis strategy reasoning location fault obtained, often requiring that the fault mode to harm to the system is higher can quickly be detected and isolate, namely should consider the hazardness of fault mode when diagnosis policy builds, the intermediate node sum making the diagnosis policy of structure again is more few more good；Probability is worked as weight by the optimization object function of existing algorithm, in practical application, the normal probability of system accounts for major part, weight difference is big, when test execution expense is identical, the minimum diagnostic tree of average test expense selected is isolated to specify the intermediate node sum of level experience to be not necessarily minimum, and owing to comentropy function is ∩ type convex function, when system probability differs greatly, adopt the single step optimum test test that simply comentropy is maximum that the heuristic function optimizing based on probability obtains, compartment system state is not necessarily maximum, it is unfavorable for reducing the intermediate node sum of diagnosis policy.

Summary of the invention

The present invention has designed and developed a kind of car fault diagnosis construction of strategy method based on density of infection, solves the defect that in existing method for diagnosing faults, diagnosis node is too much, decreases intermediate node sum final in diagnosis policy, is diagnosed to be fault fast and accurately.

Technical scheme provided by the invention is:

A kind of car fault diagnosis construction of strategy method based on density of infection, comprises the following steps:

Step one, use and include the heuristic function of the density of infection factor, select the test that single step is optimum, and using optimum test original position as with test start node, vehicle is tested accordingly, obtains sensor measurement；

Step 2, output valve according to described sensor measurement, operational failure test correlation matrix decomposites the state set node corresponding to this test output valve as the relevant minor matters point of current and described node；

If step 3 minor matters point is leaf node, then this minor matters point of labelling is for solving node, exports fail result；If minor matters point is not for solving node, return step one.

Preferably, the heuristic function including the density of infection factor described in is expressed as:

$k^{*}=\arg \underset{j}{max}\{IG(X^h,t_j)/C\left(t_j\right)\}$

IG(X^h；t_j)=H_j1(X_j1)+H_j2(X_j2)+…+H_jn(X_jn)

$H_{ji}\left(X_{ji}\right)=-\frac{\stackrel{\‾}{S\left(X_{ji}\right)}}{S\left(X\right)}log\frac{\stackrel{\‾}{S\left(X_{ji}\right)}}{S\left(X\right)}$

Wherein, k^*Represent next step the best test when given system failure collection；N is test t_jOutput valve number；C(t_j) for performing test t_jExpense；H_ji(X_ji) in fault test correlation matrix with test t_jThe fault set X that i-th output result is relevant_jiAmount of diagnostic information；For fault set X_jiCorrection relative risk degree.

Preferably, described correction relative risk degree is:

$\stackrel{\‾}{S_i}=\frac{1+S_i^{\′}}{\underset{i=1}{\overset{m}{\Σ}}(1+S_i^{\′})}$

$S_i^{\′}=\frac{S_i}{\underset{i=1}{\overset{m}{\Σ}}{S}_i}$

Wherein, S_iFor fault mode f_iDensity of infection in unit interval, m is the quantity of fault mode.

Preferably, described fault test correlation matrix is fault-two-value test correlation matrix.

Preferably, described fault test correlation matrix is fault-many-valued test correlation matrix.

Preferably, described fault test correlation matrix is replaceable units-test correlation matrix.

Preferably, fault test carries out under 16 kinds of selected mode of operations.

Preferably, under selected mode of operation, described fault test can be derived that test result.

The invention has the beneficial effects as follows: revised relative risk degree is worked as weight by the present invention, it is proposed to the optimization object function of average test expense combine testing expense, time and fault mode and be likely to factors such as the damages that equipment causes.By object function it can be seen that when test execution expense is identical, when selecting the minimum diagnostic tree of average test expense, revised relative risk degree less than probability difference (normal condition probability is bigger), the intermediate node sum in diagnostic tree is also less.What the present invention proposed is local optimization searching strategy based on what adopt in the diagnosis policy developing algorithm of density of infection, result of calculation shows, with compared with the greedy search algorithm of probability, based in the searching algorithm of density of infection when test execution expense is identical, owing to revised density of infection difference is less, algorithm obtains the maximum test of topical diagnosis quantity of information and is often also to discriminate between the test that number of faults is maximum, advantageously reduces intermediate node sum final in diagnosis policy.

Accompanying drawing explanation

Fig. 1 is the diagnosis policy schematic diagram based on fault-two-value test correlation matrix of the present invention.

Fig. 2 is the diagnosis policy schematic diagram based on fault-many-valued test correlation matrix of the present invention.

Fig. 3 is the diagnosis policy schematic diagram based on replaceable units-test correlation matrix of the present invention.

Detailed description of the invention

Below in conjunction with accompanying drawing, the present invention is described in further detail, to make those skilled in the art can implement according to this with reference to description word.

The invention provides a kind of car fault diagnosis construction of strategy method based on density of infection, it is necessary first to set up car fault diagnosis model, use F={f₀,f₁,…,f_mRepresent all fault mode f in model_iSet；Use T={t₁,t₂,…,t_lRepresent all test t in model_jSet；Total m+1 of fault mode, total l of test, and 0≤i≤m, 1≤j≤l.

The object function of existing diagnosis policy optimization problem is improved, and integration test expense, time and fault mode are likely to factors such as the test damages that causes of object, it is proposed to the optimization object function of average test expense, it may be assumed that

$J=min\left\{\underset{i=0}{\overset{m}{\Σ}}\left\{\underset{j=1}{\overset{\left|p_i\right|}{\Σ}}{c}_{p_{i\left|j\right|}}\right\}\stackrel{\‾}{S}\left(f_i\right)\right\}$

The relative risk degree of system mode is used as the weight calculation average test expense of state, wherein p by above formula_iFor diagnosis policy is isolated to fault mode f_iPerformed cycle tests, | p_i| for testing quantity,For jth test execution cost,For fault mode f_iDensity of infection weight.

Assuming that fault mode collection F={f₀,f₁,…,f_mIn fault mode f_iDensity of infection in unit interval is S_i, then the relative risk degree of fault mode is S_i':

$S_i^{\′}=\frac{S_i}{\underset{i=1}{\overset{m}{\Σ}}{S}_i}$

In practical application, for avoiding working as S_iWhen=0, S_i' weight is 0, when namely system is normal, weight is 0, to S_i' it is modified:

$\stackrel{\‾}{S_i}=\frac{1+S_i^{\′}}{\underset{i=1}{\overset{m}{\Σ}}(1+S_i^{\′})}$

The heuristic function adopted under assigned work pattern is IG (X^h；t_j)/C(t_j), represent and perform test t_jThe diagnosis obtained

$k^{*}=\arg \underset{j}{max}\{IG(X^h,t_j)/C\left(t_j\right)\}$

IG(X^h；t_j)=H_j1(X_j1)+H_j2(X_j2)+…+H_jn(X_jn)

$H_{ji}\left(X_{ji}\right)=-\frac{\stackrel{\‾}{S\left(X_{ji}\right)}}{S\left(X\right)}log\frac{\stackrel{\‾}{S\left(X_{ji}\right)}}{S\left(X\right)}$

In formula, n is test t_jOutput valve number, is two-value test during n=2, is many-valued test during n=3, C (t_j) for performing test t_jExpense；H_ji(X_ji) in fault-test correlation matrix with test t_jThe state set X that i-th output result is relevant_jiAmount of diagnostic information.When being isolated to replaceable units, H_ji(X_ji) represent test t_jI-th output result correlation unit X_jiAmount of diagnostic information.Owing to entropy is ∩ type convex function, and revised relative risk degree diversity ratio probability is less, when test execution expense is identical, tests t_jOutput result is more many, H_ji(X_ji) more big, obtainable diagnostic message is more big, and the system mode number of differentiation is also more many.

Analyzed it can be seen that diagnosis policy developing algorithm is mainly made up of heuristic function and search strategy by present Research, diagnosis policy developing algorithm can have multiple optimizing strategy, as optimum in single step or multistep, global optimum's search strategy.The heuristic function based on density of infection proposed, the optimizing strategy under a certain pattern adopts single step optimum search, tests optimizing strategy and is:

1) it is F={f by initialization failure fuzzy set₀,f₁,…,f_mIt is defined as root node, namely do not solve node；

2) application heuristic function calculates the test that single step is optimum, will test as with node, then the state set node corresponding with this test output valve is decomposited as current and node minor matters point according to correlation matrix, and labelling minor matters point and node, judge whether current all minor matters points are leaf nodes, and namely state set need not decompose again；

3) if minor matters point is leaf node, then this minor matters point of labelling is for solving node.If all minor matters points are solves node, then, when tracing back to root node, algorithm terminates；If current all minor matters points still having and not solving node, jump to step 2).

It is configured to example with two-value diagnosis policy, first obtains fault-two-value test correlation matrix

$D=\left[\begin{array}{cccc}{d}_{01}& d_{02}& \mathrm{...}& d_{0n}\\ d_{11}& d_{12}& \mathrm{...}& d_{1n}\\ .& .& & .\\ .& .& \mathrm{...}& .\\ .& .& & .\\ d_{m1}& d_{m2}& \mathrm{...}& dmn\end{array}\right]$

In above-mentioned correlation matrix, different row represents different fault mode f_i, different test t is shown in different lists_j, and

The fault derived-two-value test correlation matrix is as shown in table 1.

Table 1

According to the density of infection data in table 1, the correction relative risk degree that calculated system mode set pair is answered is:

$\stackrel{\‾}{S}=\left\{0.0909,0.1050,0.1065,0.0931,0.1067,0.1018,0.0926,0.0957,0.1003,0.1074\right\}$

According to heuristic function formula, obtain each test heuristic function value as follows:

H={0.8862,0.7212,0.9603,0.9807,0.8648}

Wherein maximum max (IG (X^h；t_j)/C(t_j))=0.9807, corresponding test is t₄, namely select test t₆As with node, according to test output result, state set is divided into by crossing two fringe collection { f with difference₀,f₃,f₅,f₆,f₇,f₉And { f₁,f₂,f₄,f₈, i.e. minor matters point.Minor matters point is not for solve node, to minor matters point { f₀,f₃,f₅,f₆,f₇,f₉Application heuristic function, obtain max (IG (X^h；t_j)/C(t_j)) corresponding test is t₃.The like the diagnosis policy based on density of infection obtained, as shown in Figure 1.After obtaining diagnosis policy, according to the testing sequence in diagnosis policy, vehicle can be tested, fault set is isolated, to find final fault.

In another embodiment, carry out the structure of many-valued diagnosis policy, obtain fault-many-valued test correlation matrix, as shown in table 2.

Table 2

When adopting above-mentioned analysis method to be isolated to concrete fault mode, obtain the diagnosis policy based on density of infection, as shown in Figure 2.

In another embodiment, carry out the structure of unit diagnosis strategy, obtain fault-test correlation matrix, as shown in table 3.

Table 3

When adopting above-mentioned analysis method to be isolated to concrete LRU, H_ji(X_ji) represent test t_jThe amount of diagnostic information of i-th output result correspondence LRU, obtains the diagnosis policy based on density of infection as shown in Figure 3.

When carrying out have selected a certain test, it should carry out this test under specific pattern.The invention provides pattern in 16, as shown in table 4.

Table 4

By above-mentioned setting, make the diagnosis policy of vehicle trouble, when carrying out car fault diagnosis, according to the testing sequence of diagnosis policy, vehicle is tested, isolate partial fault pattern, until finding final fault mode.

Although embodiment of the present invention are disclosed as above, but listed utilization that it is not restricted in description and embodiment, it can be applied to various applicable the field of the invention completely, for those skilled in the art, it is easily achieved other amendment, therefore, under the general concept limited without departing substantially from claim and equivalency range, the present invention is not limited to specific details and shown here as the legend with description.

Claims (8)

1. the car fault diagnosis construction of strategy method based on density of infection, it is characterised in that comprise the following steps:

Step one, use and include the heuristic function of the density of infection factor, select the test that single step is optimum, and using optimum test original position as with test start node, vehicle is tested accordingly, obtains measurement result by sensor；

Step 2, output valve according to described sensor measurement, operational failure test correlation matrix decomposites the state set node corresponding to this test output valve as the relevant minor matters point of current and described node；

If step 3 minor matters point is leaf node, then this minor matters point of labelling is for solving node, exports fail result；If minor matters point is not for solving node, return step one.

2. the car fault diagnosis construction of strategy method based on density of infection according to claim 1, it is characterised in that described in include the heuristic function of the density of infection factor and be expressed as:

$k^{*}=\arg \underset{j}{\max }\{\mathrm{IG}(X^h,t_j)/C\left(t_j\right)\}$

IG(X^h；t_j)=H_j1(X_j1)+H_j2(X_j2)+…+H_jn(X_jn)

$H_{ji}\left(X_{ji}\right)=-\frac{\stackrel{\‾}{S\left(X_{ji}\right)}}{\stackrel{\‾}{S\left(X\right)}}log\frac{\stackrel{\‾}{S\left(X_{ji}\right)}}{\stackrel{\‾}{S\left(X\right)}}$

Wherein, k^*Represent next step the best test when given system failure collection；N is test t_jOutput valve number；C(t_j) for performing test t_jExpense；H_ji(X_ji) in fault test correlation matrix with test t_jThe fault set X that i-th output result is relevant_jiAmount of diagnostic information；For fault set X_jiCorrection relative risk degree.

3. the car fault diagnosis construction of strategy method based on density of infection according to claim 2, it is characterised in that described correction relative risk degree is:

$\stackrel{\‾}{S_i}=\frac{1+S_i^{\′}}{\underset{i=1}{\overset{m}{\Σ}}(1+S_i^{\′})}$

$S_i^{\′}=\frac{S_i}{\underset{i=1}{\overset{m}{\Σ}}{S}_i}$

Wherein, S_iFor fault mode f_iDensity of infection in unit interval, m is the quantity of fault mode.

4. the car fault diagnosis construction of strategy method based on density of infection according to claim 3, it is characterised in that described fault test correlation matrix is fault-two-value test correlation matrix.

5. the car fault diagnosis construction of strategy method based on density of infection according to claim 3, it is characterised in that described fault test correlation matrix is fault-many-valued test correlation matrix.

6. the car fault diagnosis construction of strategy method based on density of infection according to claim 3, it is characterised in that described fault test correlation matrix is replaceable units-test correlation matrix.

7. the car fault diagnosis construction of strategy method based on density of infection according to claim 3, it is characterised in that fault test carries out under 16 kinds of selected mode of operations.

8. the car fault diagnosis construction of strategy method based on density of infection according to claim 7, it is characterised in that described fault test can be derived that test result under selected mode of operation.