CN106846305B - A kind of boiler combustion stability monitoring method based on the more characteristics of image of flame - Google Patents
A kind of boiler combustion stability monitoring method based on the more characteristics of image of flame Download PDFInfo
- Publication number
- CN106846305B CN106846305B CN201710017831.3A CN201710017831A CN106846305B CN 106846305 B CN106846305 B CN 106846305B CN 201710017831 A CN201710017831 A CN 201710017831A CN 106846305 B CN106846305 B CN 106846305B
- Authority
- CN
- China
- Prior art keywords
- image
- combustion
- coal dust
- flame
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20032—Median filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Photometry And Measurement Of Optical Pulse Characteristics (AREA)
Abstract
The invention belongs to boiler of power plant combustion stability detection field more particularly to a kind of boiler combustion stability monitoring methods based on the more characteristics of image of flame.Pass through all kinds of image feature values for extracting flame combustion video: the combustion zone index of oscillation, coal dust edge contour change difference, coal dust edge contour similarity and use RBF neural, a flameholding sex index is obtained, combustion stability classification and differentiation are carried out according to exponential size.The present invention is extracted from flame video obtains the combustion stability that 3 characteristic value sequences judge flame, realizes the combustion stability based on flame video image and determines automatically in real time.Be conducive to thermal power station's monitoring boiler combustion situation, operations staff is instructed to adjust coal-supplying amount.Technical guarantee is provided for unit safety, stabilization, economical operation.
Description
Technical field
The invention belongs to boiler of power plant combustion stability detection fields more particularly to a kind of based on the more characteristics of image of flame
Boiler combustion stability monitoring method.
Background technique
In the economy and safety of the stable relation that heat power station boiler coal-ash burns to boiler operatiopn.Stability
Method of discrimination mainly establishes model according to parameter relevant to burning, differentiates combustion stability indirectly.But due to influencing to burn
Factor it is numerous, the complexity of combustion position, practical application effect is undesirable.
In recent years, flame image collection monitoring system has been widely used in station boiler, but since site environment is complicated,
Measurement error is larger.So flame video image is also mainly used in judgement kindling fire extinguishing at the scene at present, it can't accomplish base
Combustion stability is monitored automatically in dynamic video image.
Summary of the invention
To solve the above-mentioned problems, the present invention provides a kind of sides that boiler combustion stability is judged according to video image
Method, for detecting the combustion position of burner hearth flame automatically.
The method includes characteristics extractions and judgement of stability two parts;Specially
Step 1 does image preprocessing to each frame image from furnace flame burning video, extracts the spy of each frame image
Value indicative, including the combustion zone index of oscillation, coal dust edge contour similarity and coal dust edge contour length difference;
Step 2 trains flame combustion determination of stability model using RBF neural, to judge that furnace flame burns
Stability, the range where its output parameter determine stability.
Described image pretreatment, including
Step 101, the single-frame images that t moment is extracted from furnace flame video, and by its gray processing, obtain image fg
(i,j,t);
Step 102, to image fg(i, j, t) carries out median filtering, obtains image fm(i,j,t);
Step 103, to image fm(i, j, t) does gray scale stretching processing, enhances picture contrast, obtains image fen(i,j,
t)。
The combustion zone index of oscillation can reflect that the fluctuation severe degree of combustion zone, extracting method include
Step 201, the image f for obtaining step 103 processingen(i, j, t) according to gray threshold 100,200 from secretly to it is bright according to
It is secondary to be divided into three gray levels and respectively represent unburned area, the area Chu Ran and burning-out zone, obtain image fl(i,j,t);Unburned area, that is, coal dust
It does not burn immediately after into burner hearth, absorbs radiation energy in furnace, because without issuing visible light;After the area Chu Ran, that is, coal dust enters burner hearth
It is gradually heated, starts to burn, and start to release light and heat, from the grey scale change that can be seen in horizontal direction in figure;After-flame
Area, that is, coal dust is under the heating of flame and high-temperature flue gas, completely burned, releases a large amount of light and heat, shows height in the picture
Brightness;
Step 202, through the above steps 1 and step 201 obtain two width combustion zone segmented image fl(i,j,t0) and fl
(i,j,t1);The selection of two images interval frame number combines flame flicking frequency and video frame rate, 5 frame of two images interval
When, differential effect is best;
Step 203 calculates fl(i,j,t0) and fl(i,j,t1) difference take absolute value fdiff(i, j, Δ t)=| fl(i,j,
t0)-fl(i,j,t1)|;
Step 204, statistical picture fdiff(i, j, the respective pixel number of three gray levels in Δ t), n0,n1,n2Respectively
Represent it is constant before and after combustion zone locating for pixel, change the i.e. unburned area in 1 grade of combustion zone to the area Chu Ran, the area Chu Ran to unburned area,
The area Chu Ran to burning-out zone, burning-out zone to the area Chu Ran and variation 2 grades of combustion zones, that is, unburned area to burning-out zone, burning-out zone to unburned area
Pixel number, change at the pixel of 2 grades of combustion zones that combustion fluctuation is big, change the fluctuation at the pixel of 1 grade of combustion zone
Degree is taken second place, and the degree of fluctuation before and after gray level at constant pixel is minimum, is burnt most stable;
Step 205, computational stability indexF is each combustion zone pixel
The weighted sum of the ratio of the point total pixel number of number Zhan;ω0、ω1、ω2For its weight coefficient, n is total of pixel in picture
Number;It since the pixel quantity for usually changing 2 gray levels is less, but is affected to combustion stability, so in order to amplify ash
Degree changes influence of the violent pixel to the combustion zone index of oscillation, enables ω2Equal to 16.Change the pixel of 1 gray level to combustion
It burns stability also to have an impact, enables ω1=4.And flameholding at the locating constant pixel of gray level, enable ω0=0.
The coal dust edge contour similarity can reflect the degree of fluctuation on coal dust sharp side, for judging that combustion stability is good
Bad, extracting method includes
Step 301 obtains step 103 processing, extracts binarization threshold according to Otsu algorithm, by image binaryzation,
Obtain image fbinary(i,j,t0) and fbinary(i,j,t1), interval is consistent with the combustion zone index of oscillation counting period, is spaced 5
Frame;
Step 302 detects f using Canny operatorbinary(i,j,t0) and fbinary(i,j,t1) coal dust profile sequence L (i,
t0)、L(i,t1);
Step 303, from L (i, t0) and L (i, t1) in choose longest profile Lmax(t0)、Lmax(t1) represent t0Moment and interval
T after 5 frame images1The longest profile at moment;
Step 304 is to Lmax(t0), Lmax(t1) seek 7 geometric invariant moment I1~I7;
I1=y20+y02
I3=(y30+3y)2+(3y21-y03)2
I4=(y30+y12)2+(y21+y03)2
I5=(y30-y12)(y30+y12)[(y30+y12)2-3(y21+y03)2]+
(3y21-y03)(y21+y30)[3(y30+y12)2-(y21+y03)2]
I6=(y20-y02)[(y30+y12)2-(y21+y03)2]+4y11(y30+y12)(y21+y03)
I7=3 (y21+y03)(y30+y12)[(y30+y12)2-(3y21-y03)2]+
(y30-3y12)(y21+y30)[3(y30+y12)2-(y21+y03)2]
Wherein:
In formula,
In formula, p, q=0,1,2 ...;
The wherein centre of moment (x0,y0) are as follows:
mpq=∫ ∫ xpyqF (x, y) dxdy p, q=0,1,2 ...;
Step 305 obtains coal dust edge contour similarity M;
Wherein, k (t)=sign (log10|Ii(t0)|)。
Its extracting method of coal dust edge contour length difference includes
Step 401 seeks the profile L obtained by step 303max(t0)、Lmax(t1) profile length
Wherein, n is the sum of the pixel on profile sequence L, (xj,yj) and (xj+1,yj+1) jth in profile is respectively represented,
+ 1 pixel of jth;
The coal dust edge contour length difference of step 402, the two images of 5 frame of counting period
Δ d=| d (t0)-d(t1)|。
The flame combustion stability distinguishing model, mode input extraction step are
Characteristic value described in step 1 is done 20 frame slips and is averaged by step 501;
Step 502, according to flame flicking frequency and operations staff's experience, choose before 2 seconds to current time totally 2 seconds when
The image of interior all frames calculates characteristic value to each image;Mean value, maximum-lowest difference in 2 seconds is calculated each characteristic value
Value, variance.Obtain input vector
For the combustion zone index of oscillationfH-L,Respectively represent in 2 seconds the combustion zone index of oscillation of totally 50 frame images
Mean value, maximum-minimal difference, variance;
For coal dust edge contour similarity,MH-L,Respectively represent in 2 seconds totally 50 frame image coal dust edge wheel
Mean value, the maximum-minimal difference, variance of wide similarity;
For coal dust edge contour length difference,dH-L,Respectively represent in 2 seconds totally 50 frame image coal dust edge
Mean value, the maximum-minimal difference, variance of profile length difference;
Step 503, to above-mentioned inputDo normalized
Obtain final input.
The flame combustion stability distinguishing model is a RBF nerve network;Its input layer number
It is consistent with input vector element number for 9;Its hidden layer node number is 9 × 2+1=19;Its output layer node number is
1。
The kernel function center of the RBF neural hidden layer is total under 5 kinds of flameholding states of artificial division
19 groups of inputs.
Its output parameter of the RBF neural represents stability, the parameter of output represented in 0~1.5 it is highly stable,
The parameter of output is represented 1.5~2.5 to be stablized, and the parameter of output is represented 2.5~3.5 generally to be stablized, and the parameter of output is 3.5
~4.5 represent unstable, and the parameter of output represents very unstable in 4.5~+∞.
The training sample size of the RBF neural is 5 × 100=500;By highly stable, stable, general, unstable
Fixed, very unstable each 100 samples composition of five classes operating condition;Above-mentioned highly stable, stable, general, unstable, very unstable five class
The output of operating condition is respectively labeled as 1,2,3,4,5.
Beneficial effect
The characteristics of present invention is burnt according to burner hearth internal flame, proposes a kind of boiler combustion stability based on characteristic value
Method of discrimination can determine automatically boiler combustion stability shape according to the changing features of the before and after frames image of video image in real time
Condition is conducive to thermal power station's monitoring boiler combustion situation, operations staff is instructed to adjust coal-supplying amount, is unit safety, stabilization, economic fortune
Row provides technical guarantee.
Detailed description of the invention
Fig. 1 invention broad flow diagram
The unburned area of Fig. 2 t moment, the area Chu Ran, burning-out zone segmented image
The difference image of 5 frame of the interval Fig. 3
Fig. 4 Otsu binarization result
Fig. 5 extracts to obtain longest profile
Fig. 6 RBF neural network structure figure
Fig. 7 stablizes neural network under combustion conditions and exports
Neural network exports under Fig. 8 rough burning operating condition
The output of Fig. 9 model is divided with stability
Specific embodiment
The invention proposes a kind of boiler combustion stability monitoring method based on the more characteristics of image of flame, the method packets
Include characteristics extraction and judgement of stability two parts;
Step 1 extracts each frame image from furnace flame burning video, and pretreatment operation, processing step are carried out to it
Suddenly include;
Step 101, the single-frame images that t moment is extracted from furnace flame video, and by its gray processing, obtain image fg
(i,j,t);
Step 102, to image fg(i, j, t) carries out median filtering, obtains image fm(i,j,t);
Step 103, to image fm(i, j, t) does gray scale stretching processing, enhances picture contrast, obtains image fen(i,j,
t);
Step 2 seeks the combustion zone index of oscillation to image after pretreatment;
Step 201, as shown in Figure 1, by image fen(i, j, t) is according to gray threshold 100,200 from being secretly divided into bright
It is (white in figure that three gray levels respectively represent unburned area (grid lines region in figure), the area Chu Ran (figure bend region) and burning-out zone
Color region), obtain image fl(i,j,t);Unburned area, that is, coal dust enters after burner hearth not to burn immediately, absorbs radiation energy in furnace,
Because without issuing visible light;The area Chu Ran, that is, coal dust is gradually heated after entering burner hearth, starts to burn, and start to release light and
Heat, from the grey scale change that can be seen in horizontal direction in figure;Burning-out zone, that is, coal dust is under the heating of flame and high-temperature flue gas, completely
Burning, releases a large amount of light and heat, shows high brightness in the picture;
Step 202, through the above steps 1 and step 201 obtain two width combustion zone segmented image fl(i,j,t0) and fl
(i,j,t1).The selection of two images interval frame number combines flame flicking frequency and video frame rate, 5 frame of two images interval
When, differential effect is best, as shown in Figure 3.
Step 203 calculates fl(i,j,t0) and fl(i,j,t1) difference take absolute value fdiff(i, j, Δ t)=| fl(i,j,
t0)-fl(i,j,t1)|;
Step 204, statistical picture f as shown in Figure 2diff(i, j, the respective pixel number of three gray levels in Δ t), n0,
n1,n2The front and back of combustion zone locating for pixel constant (grid lines region in figure) is respectively represented, changing 1 grade of combustion zone, (unburned area arrives
The area Chu Ran, the area Chu Ran to unburned area, the area Chu Ran to burning-out zone, burning-out zone to the area Chu Ran.Figure bend region) and 2 grades of variation
Combustion zone (unburned area to burning-out zone, burning-out zone to unburned area.White area in figure) pixel number, change 2 grades of combustion zones
Pixel at combustion fluctuation it is big, the degree of fluctuation changed at the pixel of 1 grade of combustion zone is taken second place, constant picture before and after gray level
Degree of fluctuation at vegetarian refreshments is minimum, burns most stable.
Step 205, computational stability indexF is each combustion zone pixel
The weighted sum of the ratio of the point total pixel number of number Zhan.ω0、ω1、ω2For its weight coefficient, n is total of pixel in picture
Number.It since the pixel quantity for usually changing 2 gray levels is less, but is affected to combustion stability, so in order to amplify ash
Degree changes influence of the violent pixel to the combustion zone index of oscillation, enables ω2Equal to 16.Change the pixel of 1 gray level to combustion
It burns stability also to have an impact, enables ω1=4.And flameholding at the locating constant pixel of gray level, enable ω0=0.
Step 3 seeks coal dust edge contour similarity to image after pretreatment, and extracting method includes:
Step 301 obtains step 103 processing, extracts binarization threshold according to Otsu algorithm, by image binaryzation,
Obtain image fbinary(i,j,t0) and fbinary(i,j,t1), interval is consistent with the combustion zone index of oscillation counting period, is spaced 5
Frame;
Step 302 detects f using Canny operatorbinary(i,j,t0) and fbinary(i,j,t1) coal dust profile sequence L (i,
t0)、L(i,t1);
Step 303, from L (i, t0) and L (i, t1) in choose longest profile Lmax(t0)、Lmax(t1) represent t0Moment and interval
T after 5 frame images1The longest profile at moment;
Step 304, to Lmax(t0), Lmax(t1) seek 7 geometric invariant moment I1~I7;
I1=y20+y02
I3=(y30+3y)2+(3y21-y03)2
I4=(y30+y12)2+(y21+y03)2
I5=(y30-y12)(y30+y12)[(y30+y12)2-3(y21+y03)2]+
(3y21-y03)(y21+y30)[3(y30+y12)2-(y21+y03)2]
I6=(y20-y02)[(y30+y12)2-(y21+y03)2]+4y11(y30+y12)(y21+y03)
I7=3 (y21+y03)(y30+y12)[(y30+y12)2-(3y21-y03)2]+
(y30-3y12)(y21+y30)[3(y30+y12)2-(y21+y03)2]
Wherein:
In formula,
In formula, p, q=0,1,2 ....
The wherein centre of moment (x0,y0) are as follows:
mpq=∫ ∫ xpyqF (x, y) dxdy p, q=0,1,2 ...;
Step 305 obtains coal dust edge contour similarity M, as shown in Figure 4.
Wherein, k (t)=sign (log10|Ii(t0)|)。
Step 4 seeks coal dust edge contour length difference to image after pretreatment, its extracting method as shown in Figure 5 includes:
Step 401 seeks the profile L obtained by step 303max(t0)、Lmax(t1) profile length
Wherein, n is the sum of the pixel on profile sequence L, (xj,yj) and (xj+1,yj+1) jth in profile is respectively represented,
+ 1 pixel of jth.
The coal dust edge contour length difference of step 402, the two images of 5 frame of counting period
Δ d=| d0)-d(t1)|。
Step 5 trains flame combustion Stability Model using RBF neural, to judge furnace flame flameholding
Property, the range judgement stability where its output parameter, Fig. 6 is RBF neural network structure figure.
The flame combustion Stability Model, input extraction step are
Characteristic value described in step 1 is done 20 frame slips and is averaged by step 501;
Step 502, according to flame flicking frequency and operations staff's experience, choose before 2 seconds to current time totally 2 seconds when
The image of interior all frames calculates characteristic value to each image.Mean value, maximum-lowest difference in 2 seconds is calculated each characteristic value
Value, variance.Obtain input vector
For the combustion zone index of oscillationfH-L,Respectively represent in 2 seconds the combustion zone index of oscillation of totally 50 frame images
Mean value, maximum-minimal difference, variance;
For coal dust edge contour similarity,MH-L,Respectively represent in 2 seconds totally 50 frame image coal dust edge wheel
Mean value, the maximum-minimal difference, variance of wide similarity;
For coal dust edge contour length difference,dH-L,Respectively represent in 2 seconds totally 50 frame image coal dust edge wheel
Mean value, the maximum-minimal difference, variance of wide length difference.
Step 503, to above-mentioned inputDo normalized
Obtain final input.
It is step 504, hidden as RBF neural using total 19 groups of inputs under 5 kinds of flameholding states of artificial division
Kernel function center containing layer
Step 505 is trained network, and training sample size is 5 × 100=500.By it is highly stable, stable, one
As, unstable, each 100 samples composition of very unstable five classes operating condition.It is above-mentioned highly stable, stable, general, unstable, very not
The output for stablizing five class operating conditions is respectively labeled as 1,2,3,4,5.
Step 406, to RBF neural input data, output parameter represents stability, and the parameter of output is 0~1.5
Inside represent highly stable, the parameter of output is represented 1.5~2.5 to be stablized, the parameter of output 2.5~3.5 represent it is general stablize,
The parameter of output represents unstable 3.5~4.5, and the parameter of output represents very unstable in 4.5~+∞.Fig. 9 is model output
It is divided with stability.
Two groups experimental result is shown in Fig. 7 in Figure of description, Fig. 8.Fig. 7 is that combustion stablized video clip carries out real-time stabilization
Property differentiate after, model output value.Fig. 8 is model output value after the video clip of combustion instability carries out real-time stability differentiation.
It can be observed that model output value is small when flameholding, curve fluctuation is small, and judgement of stability maintains essentially in " steady
It is fixed " and " general ".Unstable combustion timing model output value is larger, curve fluctuation frequently, " highly unstable ", it is " unstable " go out
Now frequently.
The experimental results showed that model is effective to the differentiation of boiler combustion stability.
Claims (6)
1. a kind of boiler combustion stability monitoring method based on the more characteristics of image of flame, which is characterized in that the method includes
Characteristics extraction and judgement of stability two parts;Specially
Step 1 does image preprocessing to each frame image from furnace flame burning video, extracts the feature of each frame image
Value, including the combustion zone index of oscillation, coal dust edge contour similarity and coal dust edge contour length difference;
Step 2 trains flame combustion determination of stability model using RBF neural, to judge furnace flame flameholding
Property, the range where its output parameter determines stability;
Described image pretreatment, including
Step 101, the single-frame images that t moment is extracted from furnace flame video, and by its gray processing, obtain image fg(i, j,
t);
Step 102, to image fg(i, j, t) carries out median filtering, obtains image fm(i, j, t);
Step 103, to image fm(i, j, t) does gray scale stretching processing, enhances picture contrast, obtains image fen(i, j, t);
The combustion zone index of oscillation can reflect that the fluctuation severe degree of combustion zone, extracting method include
Step 201, the image f for obtaining step 103 processingen(i, j, t) is according to gray threshold 100,200 from secretly successively dividing to bright
Unburned area, the area Chu Ran and burning-out zone are respectively represented for three gray levels, obtains image fl(i, j, t);Unburned area, that is, coal dust enters
It does not burn immediately after burner hearth, absorbs radiation energy in furnace, because without issuing visible light;The area Chu Ran, that is, coal dust enters after burner hearth gradually
It is heated, starts to burn;Burning-out zone, that is, coal dust is under the heating of flame and high-temperature flue gas, completely burned;
Step 202, through the above steps 1 and step 201 obtain two width combustion zone segmented image fl(i, j, t0) and fl(i,
J, t1);The selection of two images interval frame number combines flame flicking frequency and video frame rate;
Step 203 calculates fl(i, j, t0) and fl(i, j, t1) difference take absolute value fdiff(i, j, Δ t)=| fl(i, j, t0)-
fl(i, j, t1)|;
Step 204, statistical picture fdiff(i, j, the respective pixel number of three gray levels in Δ t), n0, n1, n2It respectively represents
It is constant before and after combustion zone locating for pixel, change the i.e. unburned area in 1 grade of combustion zone to the area Chu Ran, the area Chu Ran to unburned area, just combustion
Picture of the area to burning-out zone, burning-out zone to the area Chu Ran and variation 2 grades of combustion zones, that is, unburned area to burning-out zone, burning-out zone to unburned area
Vegetarian refreshments number, changes at the pixel of 2 grades of combustion zones that combustion fluctuation is big, changes the degree of fluctuation at the pixel of 1 grade of combustion zone
Take second place, the degree of fluctuation before and after gray level at constant pixel is minimum;
Step 205, computational stability indexF is each combustion zone pixel number
The weighted sum of the ratio of the total pixel number of Zhan;ω0、ω1、ω2For its weight coefficient, n is the total number of pixel in picture;For
Influence of the violent pixel of amplification grey scale change to the combustion zone index of oscillation, enables ω2Equal to 16;Change the picture of 1 gray level
Vegetarian refreshments also has an impact to combustion stability, enables ω1=4;And flameholding at the locating constant pixel of gray level, enable ω0=0;
The coal dust edge contour similarity can reflect the degree of fluctuation on coal dust sharp side, for judging combustion stability quality,
Extracting method includes
Step 301, the image f for obtaining step 103 processingen(i, j, t) extracts binarization threshold according to Otsu algorithm, will scheme
As binaryzation, image f is obtainedbinary(i, j, t0) and fbinary(i, j, t1), interval and combustion zone index of oscillation counting period one
It causes, 5 frames of interval;
Step 302 detects f using Canny operatorbinary(i, j, t0) and fbinary(i, j, t1) coal dust profile sequence L (i, t0)、L
(i, t1);
Step 303, from L (i, t0) and L (i, t1) in choose longest profile Lmax(t0)、Lmax(t1) represent t05 frames of moment and interval
T after image1The longest profile at moment;
Step 304 is to Lmax(t0), Lmax(t1) seek 7 geometric invariant moment I1~I7;
I1=y20+y02
I3=(y30+3y)2+(3y21-y03)2
I4=(y30+y12)2+(y21+y03)2
I5=(y30-y12)(y30+y12)[(y30+y12)2-3(y21+y03)2]+
(3y21-y03)(y21+y30)[3(y30+y12)2-(y21+y03)2]
I6=(y20-y02)[(y30+y12)2-(y21+y03)2]+4y11(y30+y12)(y21+y03)
I7=3 (y21+y03)(y30+y12)[(y30+y12)2-(3y21-y03)2]+
(y30-3y12)(y21+y30)[3(y30+y12)2-(y21+y03)2]
Wherein:
In formula,P+q=2,3 ...;
In formula, p, q=0,1,2 ...;
The wherein centre of moment (x0,y0) are as follows:
mpq=∫ ∫ xpyqF (x, y) dxdy, p, q=0,1,2 ...;
Step 305 obtains coal dust edge contour similarity M;
Wherein, k (t)=sign (log10|Ii(t0)|);
The extracting method of the coal dust edge contour length difference includes
Step 401 seeks the profile L obtained by step 303max(t0)、Lmax(t1) profile length
Wherein, n is the sum of the pixel on profile sequence L, (xj, yj) and (xj+1, yj+1) respectively represent jth in profile, jth+
1 pixel;
The coal dust edge contour length difference of step 402, the two images of 5 frame of counting period
Δ d=| d (t0)-d(t1)|。
2. monitoring method according to claim 1, which is characterized in that the flame combustion determination of stability model, mould
Type inputs extraction step
Characteristic value described in step 1 is done 20 frame slips and is averaged by step 501;
Step 502, according to flame flicking frequency and operations staff's experience, chose before 2 seconds in totally 2 second time of current time
The image of all frames calculates characteristic value to each image;To each characteristic value calculate 2 seconds in mean value, maximum-minimal difference,
Variance;Obtain input vector
For the combustion zone index of oscillation,fH-L,Respectively represent in 2 seconds the equal of the combustion zone index of oscillation of totally 50 frame images
Value, maximum-minimal difference, variance;
For coal dust edge contour similarity,MH-L,Respectively represent in 2 seconds that totally 50 frame image coal dust edge contours are similar
Mean value, the maximum-minimal difference, variance of degree;
For coal dust edge contour length difference,dH-L,Respectively represent in 2 seconds that totally 50 frame image coal dust edge contours are long
Spend mean value, the maximum-minimal difference, variance of difference;
Step 503, to above-mentioned inputNormalized is done to obtain finally
Input.
3. monitoring method according to claim 1, which is characterized in that the flame combustion determination of stability model is one
RBF nerve network;Its input layer number is 9 consistent with input vector element number;Its node in hidden layer
Mesh is 9 × 2+1=19;Its output layer node number is 1.
4. monitoring method according to claim 3, which is characterized in that in the kernel function of the RBF neural hidden layer
The heart is to amount to 19 groups of inputs under 5 kinds of flameholding states of artificial division.
5. monitoring method according to claim 1, which is characterized in that its output parameter of the RBF neural represents steady
Qualitative, the parameter of output represents highly stable in 0~1.5, and the parameter of output is represented 1.5~2.5 to be stablized, the parameter of output
2.5~3.5 represent it is general stablize, the parameter of output represents unstable 3.5~4.5, and the parameter of output is in 4.5~+∞ generation
Table is very unstable.
6. monitoring method according to claim 1, which is characterized in that the training sample size of the RBF neural is 5
× 100=500;It is made of highly stable, stable, general, unstable, very unstable each 100 samples of five classes operating condition;It is above-mentioned non-
The output of normal stable, stable, general, unstable, very unstable five classes operating condition is respectively labeled as 1,2,3,4,5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710017831.3A CN106846305B (en) | 2017-01-11 | 2017-01-11 | A kind of boiler combustion stability monitoring method based on the more characteristics of image of flame |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710017831.3A CN106846305B (en) | 2017-01-11 | 2017-01-11 | A kind of boiler combustion stability monitoring method based on the more characteristics of image of flame |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106846305A CN106846305A (en) | 2017-06-13 |
CN106846305B true CN106846305B (en) | 2019-09-10 |
Family
ID=59118120
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710017831.3A Expired - Fee Related CN106846305B (en) | 2017-01-11 | 2017-01-11 | A kind of boiler combustion stability monitoring method based on the more characteristics of image of flame |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106846305B (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110501914B (en) * | 2018-05-18 | 2023-08-11 | 佛山市顺德区美的电热电器制造有限公司 | Security monitoring method, equipment and computer readable storage medium |
CN109214332A (en) * | 2018-08-31 | 2019-01-15 | 华北电力大学 | A kind of combustion stability method of discrimination based on furnace flame image fractal characteristic |
CN109539291B (en) * | 2018-10-18 | 2019-12-13 | 吉林建筑大学 | Waste heat storage and recovery method for multi-heat-source heat dissipation |
RU2713850C1 (en) * | 2018-12-10 | 2020-02-07 | Федеральное государственное бюджетное учреждение науки Институт теплофизики им. С.С. Кутателадзе Сибирского отделения Российской академии наук (ИТ СО РАН) | Fuel combustion modes monitoring system by means of torch images analysis using classifier based on convolutional neural network |
CN109754019B (en) * | 2019-01-10 | 2020-11-06 | 燕山大学 | Method for continuously monitoring combustion condition of boiler |
CN111079537B (en) * | 2019-11-18 | 2023-09-26 | 中冶赛迪技术研究中心有限公司 | Method, system, machine-readable medium and equipment for identifying smelting working conditions of converter |
CN111706866B (en) * | 2020-06-22 | 2022-11-11 | 赵莉莉 | Carbonization monitoring method and system based on household garbage |
CN112396101A (en) * | 2020-11-16 | 2021-02-23 | 西安热工研究院有限公司 | Intelligent image analysis and decision-making method based on firepower plant hearth flame television |
CN114046533B (en) * | 2021-08-31 | 2024-03-12 | 汉谷云智(武汉)科技有限公司 | Pulverized coal furnace combustion optimization method based on flame analysis optimization |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102206727A (en) * | 2011-05-31 | 2011-10-05 | 湖南镭目科技有限公司 | Converter steelmaking endpoint determination method and system, control method and control system |
CN103343978A (en) * | 2013-06-07 | 2013-10-09 | 长沙理工大学 | Method for predicting and diagnosing pulverized coal flame stability |
TWI421721B (en) * | 2010-12-09 | 2014-01-01 | Ind Tech Res Inst | A method for combustion flames diagnosis |
-
2017
- 2017-01-11 CN CN201710017831.3A patent/CN106846305B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI421721B (en) * | 2010-12-09 | 2014-01-01 | Ind Tech Res Inst | A method for combustion flames diagnosis |
CN102206727A (en) * | 2011-05-31 | 2011-10-05 | 湖南镭目科技有限公司 | Converter steelmaking endpoint determination method and system, control method and control system |
CN103343978A (en) * | 2013-06-07 | 2013-10-09 | 长沙理工大学 | Method for predicting and diagnosing pulverized coal flame stability |
Also Published As
Publication number | Publication date |
---|---|
CN106846305A (en) | 2017-06-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106846305B (en) | A kind of boiler combustion stability monitoring method based on the more characteristics of image of flame | |
CN105678295B (en) | Gas heating flame method of real-time based on the analysis of ROI the average image | |
CN107609470B (en) | Method for detecting early smoke of field fire by video | |
CN108548199B (en) | A kind of intelligent fume exhauster wind speed is adjusted a wage scale method and device | |
CN101750152B (en) | Method for representing and diagnosing combustion instability | |
CN109460708A (en) | A kind of Forest fire image sample generating method based on generation confrontation network | |
CN107326137B (en) | Blast funnace hot blast stove burns stove process operating parameters multistage matching optimization method at times | |
CN105046868A (en) | Fire early warning method based on infrared thermal imager in narrow environment | |
CN110222633A (en) | City solid waste burning process combusts operating mode's switch method based on flame image color feature extracted | |
CN106682635A (en) | Smoke detecting method based on random forest characteristic selection | |
CN102121906A (en) | Smoke detection system as well as intelligent control system and intelligent control method of range hood | |
CN103077394B (en) | Method for automatically monitoring flame combustion stability | |
CN109614908A (en) | A kind of flame combustion state detection system and detection method | |
TW201224823A (en) | A method for combustion flames diagnosis | |
CN110717495B (en) | Solid waste incineration working condition identification method based on multi-scale color moment characteristics and random forest | |
CN103256620A (en) | Multi-information-fusion intelligent flame detecting device and detecting method thereof | |
CN102323213B (en) | On-line recognition method for flame free radial based fuel type | |
CN103425959A (en) | Flame video detection method for identifying fire hazard | |
CN112699801A (en) | Fire identification method and system based on video image | |
CN106778816A (en) | Combustion stability method of discrimination based on burning mixed coefficint and fuzzy diagnosis | |
CN109919071A (en) | Flame identification method based on infrared multiple features combining technology | |
CN114937036A (en) | Blast furnace equipment operation evaluation method and system based on artificial intelligence | |
CN116805065B (en) | Intelligent management method for monitoring data of coal powder heating furnace burner | |
CN105912872B (en) | Hot-fluid mixed effect measure in a kind of coal dust and stove based on burning image | |
CN109214332A (en) | A kind of combustion stability method of discrimination based on furnace flame image fractal characteristic |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190910 Termination date: 20200111 |