CN102445346A - Aeroengine ignition system fault diagnosis method by utilizing waveform matching and system thereof - Google Patents

Abstract

The invention relates to an aeroengine ignition system fault diagnosis method by utilizing waveform matching and a system thereof. The method comprises: a form mode characteristic of a detected waveform is solved as well as identification and preselection are carried out on N waveforms in a fault characteristic database set, wherein N is an integer that is larger than or is equal to 1; correlated identification calculation is carried out on the waveform that meets screening characteristic requirements by utilizing a cross correlation algorithm, an average absolute difference algorithm, and an average variance algorithm, so that the kind of the waveform is decided; an average amplitude difference is used as a threshold; when matching calculation is carried out in the whole set, types of reference waveforms that are outside the threshold range are abandoned, so that a searching range is reduced and an objective of information compression is achieved; and two-level information compression accelerates matching calculation; according to the first-level information compression, a screened threshold h0 is used to carry out pretreatment; and according to the second-level information compression, a screening characteristic vector is used to carry out pretreatment. In addition, a detection ignition signal, a high voltage sensor, a signal conditioning circuit, a data collection card and a PC computer are successively in series connection in the system. According to the invention, fault diagnosis efficiency is improved and maintenance time is reduced.

Description

Aeromotor ignition failure diagnostic method and system with Pohle shape coupling

Technical field

The present invention relates to a kind of aeromotor ignition failure diagnostic method.Particularly relate to a kind of aeromotor ignition failure diagnostic method and system with Pohle shape coupling.

Background technology

For each aeromotor; No matter be jet engine or piston engine; All need equip firing system, its effect is when the ground start engine, and firing system and starting system are coordinated co-operation; Make the fuel mixture ignition in firing chamber or the cylinder, to produce the condition of engine operate as normal; When flight, firing system works independently.This shows that firing system is the core of aircraft engine, whether whether this system design rationally reaches system works reliably significant to guaranteeing the flight of engine operate as normal and aircraft safety.

In the engine start process, there is the possibility of starting or loss of ignition.Along with the aircraft prolongation of working time, parts such as harness, ignition exciter unit, ignition electric nozzle can damage gradually or power descends, and can cause the unsuccessful phenomenon of igniting.These faults and problem have reduced engines ground and aerial starting performance, and are difficult to find at the engines ground dead ship condition, have reduced fault diagnosis efficient, have prolonged the boat inspection time.In-flight ingition is failed even can be caused engine cut-off, causes the generation of aircraft accident.Therefore, the exploitation of the test and diagnostic system of aviation firing system has great importance to the security and the economy of aircarrier aircraft.

Summary of the invention

Technical matters to be solved by this invention is, a kind of diagnosis efficiency that improves is provided, and reduces servicing time and manpower, utilizes the aeromotor ignition failure diagnostic method and the system of Waveform Matching.

The technical scheme that the present invention adopted is: a kind of aeromotor ignition failure diagnostic method and system with Pohle shape coupling.Utilize the aeromotor ignition failure diagnostic method of Waveform Matching, comprise and carry out following steps in order:

(1) the shape mode characteristic of obtaining tested waveform is discerned preliminary election to N waveform in the set of fault signature storehouse, and wherein, N is the integer more than or equal to 1;

(2) meet being correlated with to discern and calculating the final decision generic of screening feature request with cross correlation algorithm, average absolute difference algorithm peace mean square deviation algorithm;

(3) utilize average amplitude differences as threshold value, when the coupling in universal class was calculated, all reference waveform classifications that drops on outside the threshold range were all abandoned, thereby dwindle the hunting zone, reach the purpose of Information Compression;

(4) the two-stage Information Compression is quickened coupling calculating, and first order Information Compression is carried out pre-service with the screening threshold value, and the second-level message compression is carried out pre-service with the screening proper vector.

The described identification preliminary election of step 1 is: establish f (t) for the input waveform, should import the waveform discretize according to sampling thheorem, obtain sequence:

{f(t _j)}＝{f(t ₀)，f(t ₁)，L，f(t _p-1)}，j＝0，1，2，L，p-1

Be made as F={F _i(t) }, i=1,2, L, N are known reference waveform sets, wherein F _i(t) discrete turning to

{F _i(t _j)}＝{F _i(t ₀)，F _i(t ₁)，L，F _i(t _p-1)}，j＝1，2，L，p-1

Identification input waveform f (t) comes down to waveform f (t) is mated calculating as a template to the element in the F set one by one, finds out and f (t) degree of correlation soprano, is f (t) and belongs to one type with it together.

The described cross correlation algorithm of step 2 adopts following formula to calculate:

$G_i=\frac{K_{f\left(t\right)},F_i\left(t\right)}{{\mathrm{\σ}}_{\left[f\left(t\right)\right]}g{\mathrm{\σ}}_{\left[F_i\left(t\right)\right]}}$

$=\frac{\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}[f\left(t_j\right)-\stackrel{\‾}{f\left(t\right)}]}{\sqrt{\frac{1}{p-1}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}{[f\left(t_j\right)-\stackrel{\‾}{f\left(t\right)}]}^2}}$

$=\frac{F_i\left(t_j\right)-\stackrel{\‾}{F_i\left(t\right)}}{\sqrt{\frac{1}{p-1}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}[F_i\left(t_j\right)-\stackrel{\‾}{F_i\left(t\right)}]}}$

Wherein be K _{F (t)}, F _i(t) f (t) and F _i(t) covariance.σ _{[f (t)]}Be f (t) and F _i(t) covariance. Be f (t) and F _i(t) covariance.

Be respectively f (t) and F _i(t) mean value.

The described mean absolute difference of step 2 adopts following formula to calculate:

$G_i^{\′}=\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}|f\left(t_j\right)-F_i\left(t_j\right)|$

Described average variance adopts following formula to calculate:

$G_i^{\′\′}=\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}{[f\left(t_j\right)-F_i\left(t_j\right)]}^2$

Judgment criterion is: if G _i=max [G _j], j ∈ I, then f (t) F _i(t).If G ' _i=min [G ' _j], j ∈ I, then f (t) F _i(t).If G " _i=min [G " _j], j ∈ I, then f (t) F _i(t).Here I={1,2, L, N} representes similar.

The calculating choosing method of the described threshold value of step 3 is:

If f (t) is certain a type of waveform among the F, generate by certain error model superposition, and these error models are known that therefore, f (t) is certain classification among the F,

Like this,,, obtain the waveform sets R to be identified that becomes mapping relations with it of coincidence loss model through calculating by reference waveform set F,

If set F does

F＝(F ₁，F ₂，L，F _i，L，F _n)

Process is to F _iK kind error superposition after, intended waveform matrix R to be identified one to one

R＝(f ₁，f ₂，L，f _i，L，f _n)

ER＝(Ef ₁，Ef ₂，L，Ef _i，L，Ef _n)

EF＝(EF ₁，EF ₂，L，EF _i，L，EF _n)

In the formula, f _i-by F _i" intending to be identified " waveform by the generation of known error model

-known wave mode wave amplitude average

-" intending to be identified " waveform wave amplitude average

E＝EF-ER＝(E ₁，E ₂，L，E _i，L，E _n)

In the formula: E _iBe F _iWith I's _iThe mean wave amplitude.Make H ₀=max (E _i), H then ₀Be selected threshold value.

The described two-stage Information Compression of step 4 is quickened coupling and calculated, and is specific as follows:

At first use operator H ^*Act on the F set and obtain the screening proper vector

i＝1，2，L，r

And calculate H ₀,

First order Information Compression: with screening threshold value H ₀, carry out pre-service, if

Jettisoning waveform F then _i(t), because two average amplitudes differ too big.If Then change the second-level message compression over to, said second-level message compression: with the screening proper vector

Carry out pre-service, if

Correlation computations described in then carry out step 2,

Then abandon waveform F _i(t), continue F _I+1(t) repeat above-mentioned screening process.

A kind of test and diagnostic system of aeromotor firing system includes the tested ignition signal of serial connection successively, high pressure sensor, and signal conditioning circuit, data collecting card and PC, wherein,

Described tested ignition signal links to each other with high pressure sensor, from the high-tension ignition waveform of ignition electric nozzle acquisition;

High pressure sensor links to each other with signal conditioning circuit, will become 10V with interior low tension from the high-tension electricity of the 25kV of ignition electric nozzle output;

Signal conditioning circuit links to each other with data collecting card, carries out shaping, conversion, Filtering Processing to high pressure sensor output signal, becomes standard signal;

Data collecting card links to each other with PC, carries out the AD conversion, becomes digital signal to simulating signal, in the input PC, so that carry out data processing;

PC shows with data processing the ignition signal that collects and draws the fault diagnosis conclusion.

Aeromotor ignition failure diagnostic method and the system that utilizes Waveform Matching of the present invention; Be that fluctuation characteristic with the rising of the amplitude expectation value of waveform and waveform, decline, level constitutes screening threshold value and proper vector; Waveform sets to known is screened, only to satisfying that threshold value requires and just carrying out the correlation computations of template matches with the identical waveform of proper vector.The present invention reduces servicing time for improving fault diagnosis efficient, and is significant to the security and the continuous airworthi ness that improve aeromotor.

Description of drawings

Fig. 1 is the aeromotor ignition failure diagnostic method process flow diagram that utilizes Waveform Matching of the present invention;

Fig. 2 is the test and diagnostic system chart that is used for the aeromotor firing system of the inventive method.

Embodiment

Below in conjunction with instance and accompanying drawing the aeromotor ignition failure diagnostic method and the system of Waveform Matching of utilizing of the present invention made detailed description.

The aeromotor ignition failure diagnostic method of Waveform Matching and the system of utilizing of the present invention utilizes high pressure sensor to obtain tested ignition signal; Make the reference position of waveform identical with the reference position and the final position of fault signature library template with final position; Utilize cross correlation algorithm, average absolute difference algorithm peace mean square deviation algorithm carries out template matches, utilizes the threshold value screening to discern preliminary election; Improve operation efficiency; All typical fault waveform arrays in tested waveform array and the fault signature storehouse are carried out the data point Waveform Matching, select certain fault-tolerant scope as manner of comparison, the failure judgement type.

As shown in Figure 1, the aeromotor ignition failure diagnostic method that utilizes Waveform Matching of the present invention comprises as carrying out following steps in order:

(1) the S0 stage: the shape mode characteristic of obtaining tested waveform is discerned preliminary election to N waveform in the set of fault signature storehouse, and wherein, N is the integer more than or equal to 1;

Described identification preliminary election is: establish f (t) for the input waveform, should import the waveform discretize according to sampling thheorem, obtain sequence:

{f(t _j)}＝{f(t ₀)，f(t ₁)，L，f(t _p-1)}，j＝0，1，2，L，p-1 (1)

Be made as F={F _i(t) }, i=1,2, L, N are known reference waveform sets, wherein F _i(t) discrete turning to

{F _i(t _j)}＝{F _i(t ₀)，F _i(t ₁)，L，F _i(t _p-1)}，j＝1，2，L，p-1 (2)

Identification input waveform f (t) comes down to waveform f (t) is mated calculating as a template to the element in the F set one by one, finds out and f (t) degree of correlation soprano, is f (t) and belongs to one type with it together.

(2) the S1 stage: meet being correlated with to discern and calculating its generic of final decision of screening feature request with cross correlation algorithm, average absolute difference algorithm peace mean square deviation algorithm;

The described cross correlation algorithm of step 2 adopts following formula to calculate:

$G_i=\frac{K_{f\left(t\right)},F_i\left(t\right)}{{\mathrm{\σ}}_{\left[f\left(t\right)\right]}g{\mathrm{\σ}}_{\left[F_i\left(t\right)\right]}}$

$=\frac{\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}[f\left(t_j\right)-\stackrel{\‾}{f\left(t\right)}]}{\sqrt{\frac{1}{p-1}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}{[f\left(t_j\right)-\stackrel{\‾}{f\left(t\right)}]}^2}}$

$=\frac{F_i\left(t_j\right)-\stackrel{\‾}{F_i\left(t\right)}}{\sqrt{\frac{1}{p-1}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}[F_i\left(t_j\right)-\stackrel{\‾}{F_i\left(t\right)}]}}---\left(3\right)$

Wherein be K _{F (t)}, F _i(t) f (t) and F _i(t) covariance.σ _{[f (t)]}Be f (t) and F _i(t) covariance.

Be f (t) and F _i(t) covariance.

Be respectively f (t) and F _i(t) mean value.

Described mean absolute difference adopts following formula to calculate:

$G_i^{\′}=\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}|f\left(t_j\right)-F_i\left(t_j\right)|---\left(4\right)$

Described average variance adopts following formula to calculate:

$G_i^{\′\′}=\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}{[f\left(t_j\right)-F_i\left(t_j\right)]}^2---\left(5\right)$

Judgment criterion is: if G _i=max [G _j], j ∈ I, then f (t) F _i(t).If G ' _i=min [G ' _j], j ∈ I, then f (t) F _i(t).If G " _i=min [G " _j], j ∈ I, then f (t) F _i(t).Here I={1,2, L, N} representes similar.

(3) the S2 stage: utilize average amplitude differences as threshold value, when the coupling in universal class was calculated, all reference waveform classifications that drops on outside the threshold range were all abandoned, thereby dwindle the hunting zone, reach the purpose of Information Compression;

The calculating choosing method of described threshold value is:

If f (t) is certain a type of waveform among the F, generate (for example f (t) various error components and interference when measuring etc.) by certain error model superposition, and these error models are known, therefore, f (t) is certain classification among the F,

Like this,,, obtain the waveform sets R to be identified that becomes mapping relations with it of coincidence loss model through calculating by reference waveform set F,

If set F does

F＝(F ₁，F ₂，L，F _i，L，F _n)

Process is to F _iK kind error superposition after, intended waveform matrix R to be identified one to one

R＝(f ₁，f ₂，L，f _i，L，f _n)

ER＝(Ef ₁，Ef ₂，L，Ef _i，L，Ef _n)

EF＝(EF ₁，EF ₂，L，EF _i，L，EF _n)

In the formula, f _i-by F _i" intending to be identified " waveform by the generation of known error model

-known wave mode wave amplitude average (6)

-" intending to be identified " waveform wave amplitude average (7)

E＝EF-ER＝(E ₁，E ₂，L，E _i，L，E _n) (8)

In the formula: E _iBe F _iWith I's _iThe mean wave amplitude.Make H ₀=max (E _i), H then ₀Be selected threshold value.

(4) the S3 stage: the two-stage Information Compression is quickened coupling and is calculated, and first order Information Compression is carried out pre-service with the screening threshold value, and the second-level message compression is carried out pre-service with the screening proper vector.

Described two-stage Information Compression is quickened coupling and calculated, and is specific as follows:

At first use operator H ^*Act on the F set and obtain the screening proper vector

i＝1，2，L，r (9)

And calculate H ₀,

First order Information Compression: with screening threshold value H ₀, carry out pre-service, if

Jettisoning waveform F then _i(t), because two average amplitudes differ too big.If

Then change the second-level message compression over to, said second-level message compression: with the screening proper vector

Carry out pre-service, if

The correlation computations of (3), (4), (5) formula in then carry out step 2,

Then abandon waveform F _i(t), continue F _I+1(t) repeat above-mentioned screening process.

As shown in Figure 2, the test and diagnostic system of aeromotor firing system provided by the invention is by tested ignition signal 1, high pressure sensor 2, and signal conditioning circuit 3, data collecting card 4 composes in series with PC 5 successively.Wherein,

Described tested ignition signal 1 links to each other with high pressure sensor 2, is the high-tension ignition waveform that the ignition electric nozzle place obtains.

High pressure sensor 2 links to each other with signal conditioning circuit 3, and it makes the high-tension electricity of the 25kV of ignition electric nozzle output become 10V with interior low tension.

Signal conditioning circuit 3 links to each other with data collecting card 4, and it can carry out shaping, conversion, Filtering Processing to high pressure sensor 2 output signals, becomes standard signal.

Data collecting card 4 links to each other with PC 5, and it can carry out the AD conversion, becomes digital signal to simulating signal, in the input PC, so that carry out data processing.

5 pairs of ignition signals that collect of PC show with data processing and draw the fault diagnosis conclusion.

Claims (7)

1. an aeromotor ignition failure diagnostic method that utilizes Waveform Matching is characterized in that, comprises and carries out following steps in order:

(1) the shape mode characteristic of obtaining tested waveform is discerned preliminary election to N waveform in the set of fault signature storehouse, and wherein, N is the integer more than or equal to 1;

(2) meet being correlated with to discern and calculating the final decision generic of screening feature request with cross correlation algorithm, average absolute difference algorithm peace mean square deviation algorithm;

(3) utilize average amplitude differences as threshold value, when the coupling in universal class was calculated, all reference waveform classifications that drops on outside the threshold range were all abandoned, thereby dwindle the hunting zone, reach the purpose of Information Compression;

(4) the two-stage Information Compression is quickened coupling calculating, and first order Information Compression is carried out pre-service with the screening threshold value, and the second-level message compression is carried out pre-service with the screening proper vector.

2. the aeromotor ignition failure diagnostic method that utilizes Waveform Matching according to claim 1 is characterized in that, the described identification preliminary election of step 1 is: establish f (t) for the input waveform, should import the waveform discretize according to sampling thheorem, obtain sequence:

{f(t _j)}＝{f(t ₀)，f(t ₁)，L，f(t _p-1)}，j＝0，1，2，L，p-1

Be made as F={F _i(t) }, i=1,2, L, N are known reference waveform sets, wherein F _i(t) discrete turning to

{F _i(t _j)}＝{F _i(t ₀)，F _i(t ₁)，L，F _i(t _p-1)}，j＝1，2，L，p-1

Identification input waveform f (t) comes down to waveform f (t) is mated calculating as a template to the element in the F set one by one, finds out and f (t) degree of correlation soprano, is f (t) and belongs to one type with it together.

3. the aeromotor ignition failure diagnostic method that utilizes Waveform Matching according to claim 1 is characterized in that, the described cross correlation algorithm of step 2 adopts following formula to calculate:

$G_i=\frac{K_{f\left(t\right)},F_i\left(t\right)}{{\mathrm{\σ}}_{\left[f\left(t\right)\right]}g{\mathrm{\σ}}_{\left[F_i\left(t\right)\right]}}$

$=\frac{\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}[f\left(t_j\right)-\stackrel{\‾}{f\left(t\right)}]}{\sqrt{\frac{1}{p-1}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}{[f\left(t_j\right)-\stackrel{\‾}{f\left(t\right)}]}^2}}$

$=\frac{F_i\left(t_j\right)-\stackrel{\‾}{F_i\left(t\right)}}{\sqrt{\frac{1}{p-1}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}[F_i\left(t_j\right)-\stackrel{\‾}{F_i\left(t\right)}]}}$

Wherein be K _{F (t)}, F _i(t) f (t) and F _i(t) covariance.σ _{[f (t)]}Be f (t) and F _i(t) covariance.t

Be f (t) and F _i(t) covariance.

Be respectively f (t) and F _i(t) mean value.

4. the aeromotor ignition failure diagnostic method that utilizes Waveform Matching according to claim 1 is characterized in that, the described mean absolute difference of step 2 adopts following formula to calculate:

$G_i^{\′}=\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}|f\left(t_j\right)-F_i\left(t_j\right)|$

Described average variance adopts following formula to calculate:

$G_i^{\′\′}=\frac{1}{p}\underset{j=0}{\overset{p-1}{\mathrm{\Σ}}}{[f\left(t_j\right)-F_i\left(t_j\right)]}^2$

Judgment criterion is: if G _i=max [G _j], j ∈ I, then f (t) F _i(t).If G ' _i=min [G ' _j], j ∈ I, then f (t) F _i(t).If G " _i=min [G " _j], j ∈ I, then f (t) F _i(t).Here I={1,2, L, N} representes similar.

5. the aeromotor ignition failure diagnostic method that utilizes Waveform Matching according to claim 1 is characterized in that the calculating choosing method of the described threshold value of step 3 is:

If f (t) is certain a type of waveform among the F, generate by certain error model superposition, and these error models are known that therefore, f (t) is certain classification among the F,

Like this,,, obtain the waveform sets R to be identified that becomes mapping relations with it of coincidence loss model through calculating by reference waveform set F,

If set F does

F＝(F ₁，F ₂，L，F _i，L，F _n)

Process is to F _iK kind error superposition after, intended waveform matrix R to be identified one to one

R＝(f ₁，f ₂，L，f _i，L，f _n)

ER＝(Ef ₁，Ef ₂，L，Ef _i，L，Ef _n)

EF＝(EF ₁，EF ₂，L，EF _i，L，EF _n)

In the formula, f _i--by F _i" intending to be identified " waveform by the generation of known error model

-known wave mode wave amplitude average

-" intending to be identified " waveform wave amplitude average

E＝EF-ER＝(E ₁，E ₂，L，E _i，L，E _n)

In the formula: E _iBe F _iWith I's _iThe mean wave amplitude.Make H ₀=max (E _i), H then ₀Be selected threshold value.

6. the aeromotor ignition failure diagnostic method that utilizes Waveform Matching according to claim 1 is characterized in that, the described two-stage Information Compression of step 4 is quickened coupling and calculated, and is specific as follows:

At first use operator H ^*Act on the F set and obtain the screening proper vector

i＝1，2，L，r

And calculate H ₀,

First order Information Compression: with screening threshold value H ₀, carry out pre-service, if

Jettisoning waveform F then _i(t), because two average amplitudes differ too big.If

Then change the second-level message compression over to, said second-level message compression: with the screening proper vector

Carry out pre-service, if Correlation computations described in then carry out step 2,

Then abandon waveform F _i(t), continue F _I+1(t) repeat above-mentioned screening process.

7. a test and diagnostic system that adopts the aeromotor firing system of the described method of claim 1 is characterized in that, includes the tested ignition signal 1 of serial connection successively, high pressure sensor 2, and signal conditioning circuit 3, data collecting card 4 and PC 5, wherein,

Described tested ignition signal 1 links to each other with high pressure sensor 2, from the high-tension ignition waveform of ignition electric nozzle acquisition;

High pressure sensor 2 links to each other with signal conditioning circuit 3, will become 10V with interior low tension from the high-tension electricity of the 25kV of ignition electric nozzle output;

Signal conditioning circuit 3 links to each other with data collecting card 4, carries out shaping, conversion, Filtering Processing to high pressure sensor 2 output signals, becomes standard signal;

Data collecting card 4 links to each other with PC 5, carries out the AD conversion, becomes digital signal to simulating signal, in the input PC, so that carry out data processing;

5 pairs of ignition signals that collect of PC show with data processing and draw the fault diagnosis conclusion.

Images