Summary of the invention
The object of this invention is to provide a kind of failure symptom extracting method based on vector spectrum, with solve existing employing single channel signal carry out acquisition process cause False Rate large problem.
The present invention is for solving the problems of the technologies described above and providing a kind of failure symptom extracting method based on vector spectrum, and this extracting method comprises the following steps:
1) vibration data of orthogonal two passages in same cross section is gathered;
2) carry out vector spectrum filtering to the vibration data of gathered two passages, the master extracted on vector spectrogram under different frequency multiplication shakes vector value;
3) Fuzzy processing is carried out to the vector value that shakes of the master under extracted each frequency multiplication, determine the degree of membership of each frequency multiplication;
4) failure symptom information is extracted, to determine type belonging to vibration fault according to determined degree of membership.
Described step 3) be adopt to rise the determination that ridge shape distribution function carries out degree of membership.
The described ridge shape distribution function that rises is:
Wherein x is vibration amplitude, a and b is the threshold value of different stage.
Described step 4) in the failure symptom that extracts be that degree of membership is greater than fault corresponding to setting degree of membership threshold value.
Described setting degree of membership threshold value is the half of maximum membership degree.
Described step 1) be realize measuring by arranging two mutually perpendicular sensors in same cross section.
The invention has the beneficial effects as follows: first the present invention measures the vibration original signal of orthogonal two passages in same cross section, then carry out vector spectrum filtering to the vibration data of gathered two passages, the master extracted on vector spectrogram under different frequency multiplication shakes vector value; And Fuzzy processing is carried out to the vector value that shakes of the master under extracted each frequency multiplication, determine the degree of membership of each frequency multiplication; Extract prognostic information according to determined degree of membership, determine type belonging to vibration fault.The present invention adopts twin-channel vector spectrum analytical approach can reflect the spread spectrum scenarios of vibration signal objective reality, in conjunction with rational Fuzzy distribution function, can significantly improve the reliability of fault diagnosis.
Embodiment
Below in conjunction with accompanying drawing, the specific embodiment of the present invention is further described.
The present invention is by measuring the vibration original signal of orthogonal two passages in same cross section, the spread spectrum scenarios of rotor is reflected by vector spectrum analytical approach objective reality, in conjunction with rational Fuzzy distribution function, the reliability of fault diagnosis can be significantly improved, as shown in Figure 3, detailed process is as follows for its flow process:
1. gather the vibration original signal of orthogonal two passages in same cross section
This step realizes by sensor installation on orthogonal two passages in same cross section.The present embodiment is described for certain power plant steam turbine rotor oscillation signal, the layout of sensor is according to the feature of rotor, pass the two ends adopting and be arranged symmetrically in rotor, install sketch as shown in Figure 1, the vibration data of this power plant steam turbine rotor measuring point 2 position can adopt orthogonal electric vortex sensor measuring to go out.
2. the vibration data of pair two passages gathered carries out vector spectrum filtering, and the master extracted on vector spectrogram under different frequency multiplication shakes vector value.
According to the binary channels current vortex data that measuring point 2 collects, carry out vector spectrum analysis, wherein for X to for frequency spectrum, as shown in Fig. 2-a, 1 frequency multiplication value is maximum, and be attended by 2 frequencys multiplication, other frequency multiplication values are very little, draw it may is rotor bow fault according to expertise; Analyze from Y-direction frequency spectrum separately, as shown in Fig. 2-b, 2 frequencys multiplication values are maximum, and other frequencys multiplication smaller, draws it may is misalign fault according to expertise.Vector spectrum analysis is carried out to x, y two passages, as shown in fig. 2-c, can find out, 2 frequency multiplication values are maximum, be attended by 1 frequency multiplication, draw comprehensively to be inferred as according to expertise and misalign fault, analyze x separately, y channel frequency spectrum figure, likely obtain different inferred results, therefore, the approach that vector spectrum Filtering Analysis is highly effective raising fault diagnosis accuracy is carried out to x, y passage.
3. the vector value that shakes of the master under pair each frequency multiplication extracted carries out Fuzzy processing, determines the degree of membership of each frequency multiplication.
When vibration amplitude reaches a certain setting danger warning value b, degree of membership is defined as 1, and when vibration amplitude is lower than a certain setting alarming value a, degree of membership is defined as 0.According to this, adopt ambiguity function to process the main vector value that shakes, the Fuzzy distribution function that this enforcement adopts is for rise ridge shape distribution function, and concrete form is as follows:
Amplitude under main frequency multiplication (1/4X, 1/2X, 1X, 2X, 3X, 4X, 5X) on spectrogram is carried out Fuzzy processing, determines the degree of membership of each frequency multiplication, form sign fuzzy vector.According to vector spectrum characteristic, a, b value that vector Pu Sheng ridge shape distribution function is corresponding should be 0.3 ~ 0.5 times during single channel.
(4) fuzzy matrix
Fuzzy matrix R is the important step embodying expert system fault diagnosis level, and be the important component part of expert reasoning machine, blurring mapping formula b=aR is sign fuzzy vector a (a
1, a
2..., a
m) and fault fuzzy vector b (b
1, b
2..., b
n) bridge.
(5) setting of threshold value
For effectively ensureing the reliability that fault is inferred, avoiding the generation of judging by accident, needing setting degree of membership threshold value, when degree of membership exceedes given threshold value, prove to infer effectively.General setting threshold value is the half of maximum membership degree value.
(6) checking and conclusion
In order to verify the validity that vector spectrum sign is fuzzy, at the scene steam turbine has carried out multiple authentication, experimental procedure is carried out according to above step.Single channel frequency spectrum sign Fuzzy Processing result and binary channels sign Fuzzy Processing result are analyzed by experiment.Comparing result is as shown in table 1.
Table 1
Infer and experimental analysis according to expert, determine a fuzzy relationship matrix r (remarks: through exchanging with expert, fuzzy matrix R renewal is as follows :)
According to blurring mapping formula b=aR, derive fault fuzzy vector, be subordinate to angle value accordingly as shown in table 2:
Table 2 (remarks: the data of table 2 upgrade as follows according to fuzzy matrix R)
|
Uneven |
Misalign |
Bending |
Oil whirl |
Oil whip |
X is to frequency spectrum |
0.8534 |
0.5648 |
0.9769 |
0.8534 |
0.9460 |
Y-direction frequency spectrum |
0 |
0.9156 |
0.3662 |
0 |
0.2747 |
Vector spectrum |
0.5212 |
0.9175 |
0.8256 |
0.5212 |
0.7495 |
According to threshold value setting rule, X, to threshold value δ=0.9769/2=0.4885 corresponding to frequency spectrum, finally determines that X is flexural failure to fault most probable; In like manner, threshold value δ=0.4578 that Y-direction frequency spectrum is corresponding, finally determines that Y-direction fault most probable is for misaligning fault; Threshold value δ=0.4588 corresponding according to vector spectrum, final confirmation fault is for misaligning fault.
Known by comparative analysis, compared to single channel frequency spectrum sign blur method, extract prognostic information more comprehensively based on the vector spectrum sign blur method rising ridge shape distribution function, expert system fault diagnosis result reliability is higher.