CN102706534A - Gas-liquid two-phase flow pattern recognition method - Google Patents

Abstract

The invention provides a gas-liquid two-phase flow pattern recognition method based on interval intuition fuzzy set similarity measure. As a membership grade function and a non-membership grade function are defined, supporting and opposing information is taken into consideration, and suspicious information is added, the recognition difficult problem of mutually crossed characteristic parameters of flow patterns of the two-phase flow is solved; meanwhile, as the membership grade and the non-membership grade in an interval intuition fuzzy set are represented through intervals, flexibility and practicability are realized in case of the treatment of fuzzification, indeterminacy and the like, the essential characteristics of the objective world can be described and reflected, and the accuracy and reliability of flow pattern recognition of the two-phase flow are improved further.

Description

A kind of Method for Discriminating Gas-liquid Two Phase Flow

Technical field

The present invention relates to the two-phase flow measurement technical field, especially relate to Method for Discriminating Gas-liquid Two Phase Flow based on interval intuitionistic Fuzzy Sets similarity measure.

Background technology

Biphase gas and liquid flow extensively is present among the modern industries productions such as water conservancy, electric power, oil, chemical industry, and its flow pattern has very important influence to related industries Equipment Design, operation and security.But in the engineering practical application, because the phase interface effect of the distribution situation of biphase gas and liquid flow flow media and two-phase flow complicacy is unclear fully as yet so far, two-phase flow is hydromechanical research emphasis and difficult point always.In the characterisitic parameter of research two-phase flow; The research of flow pattern with confirm it is top priority; It not only influences Two-phase flow characteristic and performances such as heat transfer, mass transfer, and influences the accurate measurement of other parameter of two-phase flow system, therefore; Carry out two phase flow pattern ONLINE RECOGNITION new principle and Study on new method and have very important scientific meaning, have significant industrial application value simultaneously.

Because the importance of two phase flow pattern identification, domestic and international many researchists have done number of research projects to this.The work of studying the flow pattern ONLINE RECOGNITION has the earliest proposed the method according to the probability density function identification flow pattern of pressure surge with artificial representatives such as Hubbard in 1966.Develop into today, flow pattern of gas-liquid two-phase flow Classification and Identification technology commonly used has nerual network technique, chaology, fractal theory, information fusion technology and Fuzzy Processing etc.Neural network has been widely used in the two phase flow pattern Study of recognition, and has obtained good recognition effect owing to have simple structure and self-learning function.But there is the so-called problem concerning study of crossing usually in neural network, is suitable for the small sample training, and along with the increase of sample number, the training difficulty can increase sharply, and network convergence speed is slack-off, thus big limitations the raising of its extensive performance.Method such as chaology and Fuzzy Processing also is applied in the flow pattern identification of two-phase flow by some scholars, but these fuzzy reasonings are fairly simple, and for the cross one another flow pattern of characteristic parameter, said method still can not be discerned well.

Though done a large amount of research in Chinese scholars aspect the flow pattern of gas-liquid two-phase flow identification, the reliability and the repeatability of flow pattern ONLINE RECOGNITION method are lower, still can not solve the industrial flow problem well.Main cause is because the variation of flow pattern is a complex random process, and the division of flow pattern is that a fuzzy character property is described, and the various characteristic parameters of flow pattern are all serious to intersect each other.So, how to handle the serious cross one another problem of various characteristic parameters of flow pattern effectively, be to solve present two phase flow pattern recognition accuracy and the low difficult point of reliability; And traditional method all can't solve the cross one another identification difficult problem of the various characteristic parameters of two phase flow pattern at present.

Summary of the invention

To above-mentioned technological deficiency, the present invention proposes a kind of flow pattern of gas-liquid two-phase flow identification new method based on interval intuitionistic Fuzzy Sets similarity measure, to solve the cross one another identification difficult problem of the various characteristic parameters of two phase flow pattern.

In order to solve the problems of the technologies described above, the present invention proposes a kind of Method for Discriminating Gas-liquid Two Phase Flow based on interval intuitionistic Fuzzy Sets similarity measure.

A kind of Method for Discriminating Gas-liquid Two Phase Flow comprises the steps:

11) utilize high speed photography to obtain the flow image of biphase gas and liquid flow;

12) utilize digital image processing techniques to extract bubble area x1 in the two-phase flow binary image, width x2 and height x3 characteristic parameter;

13) based on two-phase flow characteristics defined five standard flow patterns: bubbly flow

plug flow

stratified flow

slug flow and fog annular flow

14) with the membership function and the non-membership function of the bubble area x1 that obtains, width x2 and height x3 characteristic parameter substitution flow of bubble

slug flow

stratified flow

slug flow

and mist annular flow

; Said membership function and non-membership function are respectively

and

and

and wherein, k=1 ~ 5;

15) repeating step 12) thereby and step 14) confirm membership function and non-membership function interval range

and

and

and

and

and and

He

He

He

wherein, k=1 ~ 5;

16) utilize normalization method treatment step 15) membership function that obtains and the bound of non-membership function, make satisfied:

${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_i\right)=[{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^L\left(x\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^U\left(x_i\right)]\&Subset;\left[\mathrm{0,1}\right],$ ${\tilde{v}}_{\tilde{A}k}\left(x_i\right)=[{\tilde{v}}_{\tilde{A}k}^L\left(x_i\right),{\tilde{v}}_{\tilde{A}k}^U\left(x_i\right)]\&Subset;\left[\mathrm{0,1}\right],$ Wherein, k=1 ~ 5, i=1,2,3; Then ${\tilde{A}}_k=\{<x_1,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_1\right),{\tilde{v}}_{\tilde{A}k}\left(x_1\right)>,<x_2,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_2\right),{\tilde{v}}_{\tilde{A}k}\left(x_2\right)>,<x_3,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_3\right),{\tilde{v}}_{\mathrm{Ak}}\left(x_3\right)>|x_1,x_2,x_3\&Element;X\},$ Wherein, k=1 ~ 5 are represented five kinds of basic flow patterns of biphase gas and liquid flow respectively

With Interval intuitionistic Fuzzy Sets;

17) set up the weighting cosine similarity measure of forming by each flow pattern more characteristic parameters, confirm weighting coefficient w _i∈ [0,1]; Calculate flow pattern to be identified With each standard flow pattern

With

Between weighting cosine similarity measure: according to maximum membership grade principle, if

(k=1 ~ 5, i=1 ~ 5), then two interval intuitionistic Fuzzy Sets

With

Between weighting cosine similarity measure maximum, angle is minimum, then

Be under the jurisdiction of relatively

Further, establishing X is nonempty set, then claims:

$\tilde{A}=\{<x,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}\left(x\right),{\tilde{v}}_{\tilde{A}}\left(x\right)>|x\&Element;X\}---\left(1\right)$

Be interval intuitionistic Fuzzy Sets, note is made IVIFS (X), wherein:

${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}\left(x\right)=[{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)]\&Subset;\left[\mathrm{0,1}\right]---\left(2\right)$

${\tilde{v}}_{\tilde{A}}\left(x\right)=[{\tilde{v}}_{\tilde{A}}^L\left(x\right),{\tilde{v}}_{\tilde{A}}^U\left(x\right)]\&Subset;\left[\mathrm{0,1}\right]---\left(3\right)$

With

Be respectively that x belongs to

Membership function and non-membership function,

With

The upper and lower bound of representing membership function respectively,

With The lower limit and the upper limit of representing non-membership function respectively, and must satisfy ${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)+{\tilde{v}}_{\tilde{A}}^U\left(x\right)\&le;1,$ Order ${\widetilde{\mathrm{\&pi;}}}_{\tilde{A}}^L{=1-\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)-{\tilde{v}}_{\tilde{A}}^U\left(x\right);$ ${\widetilde{\mathrm{\&pi;}}}_{\tilde{A}}^U{=1-\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x\right)-{\tilde{v}}_{\tilde{A}}^L\left(x\right);$ Then claim

For x belongs to

The hesitation degree.

Further, make up vector:

$V_{\tilde{A}}\left(x_i\right)=\{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right),{\tilde{v}}_{\tilde{A}}^L\left(x_i\right),{\tilde{v}}_{\tilde{A}}^U\left(x_i\right)\}---\left(4\right)$

$V_{\tilde{B}}\left(x_i\right)=\{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right),{\tilde{v}}_{\tilde{B}}^L\left(x_i\right),{\tilde{v}}_{\tilde{B}}^U\left(x_i\right)\}---\left(5\right)$

Make up the cosine similarity measure:

$C_{\mathrm{IVIFS}}(\tilde{A},\tilde{B})=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^L\left(x_i\right){\tilde{v}}_{\tilde{B}}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^U\left(x_i\right){\tilde{v}}_{\tilde{B}}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{U^2}\left(x_i\right)}}---\left(6\right)$

Make up weighting cosine similarity measure:

${\mathrm{Cw}}_{\mathrm{IVIFS}}(\tilde{A},\tilde{B})=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^L\left(x_i\right){\tilde{v}}_{\tilde{B}}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^U\left(x_i\right){\tilde{v}}_{\tilde{B}}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{U^2}\left(x_i\right)}}---\left(7\right)$

Wherein, w _i∈ [0,1], i=1,2 ..., n, and

Further, said step 17) concrete steps comprise:

41) great amount of samples collection

j of extraction two phase flow pattern is concrete sample number, calculates the wherein pairing bubble area x1 of each flow pattern, width x2 and height x3;

42) according to membership function and the non-membership function of the bubble area x1 that has confirmed in the step 12), width x2 and height x3;

is the basis with the average weighted coefficient, the weighting cosine similarity measure in the calculating sample set between each concrete flow pattern

and each standard flow pattern

and

:

${\mathrm{Cw}}_{\mathrm{IVIFS}}(\widetilde{A_k},\widetilde{B_j})=\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}{w}_i\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^L\left(x_i\right){\tilde{v}}_{\tilde{B}j}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^U\left(x_i\right){\tilde{v}}_{\tilde{B}j}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}j}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}j}^{U^2}\left(x_i\right)}}---\left(8\right)$

43) calculate by formula (8)

K=1 ~ 5 are differentiated according to maximum membership grade principle

The flow pattern classification, recognition result and visual observation are relatively suitably revised weighting coefficient according to recognition result then again, thus through revising the weighting coefficient of confirming each degree of membership and non-membership function: w once more _i∈ [0,1],

Improve precision.

Beneficial effect of the present invention is: traditional Method for Discriminating Gas-liquid Two Phase Flow can not be handled the serious cross one another problem of various characteristic parameters of flow pattern effectively, can not solve present two phase flow pattern recognition accuracy and the low difficult point of reliability well; Method for Discriminating Gas-liquid Two Phase Flow based on interval intuitionistic Fuzzy Sets similarity measure of the present invention; Owing to defined membership function and non-membership function; Not only considered the information of supporting and opposing; And increased suspicious degree information, solved the cross one another identification difficult problem of characteristic parameter between two phase flow pattern; Degree of membership and the non-degree of membership concentrated owing to interval intuitionistic fuzzy are simultaneously represented with the interval; Have more dirigibility and practicality at aspects such as handling ambiguity and uncertainty; It more can describe and reflect the essential characteristic of objective world, has further improved the accuracy rate and the reliability of two phase flow pattern identification.

Description of drawings

Fig. 1 is the technical route figure that the present invention discerns flow pattern of gas-liquid two-phase flow;

Fig. 2 is an experimental provision synoptic diagram of the present invention;

Fig. 3 is the flow regime map that obtains during the present invention tests.

Embodiment

To combine accompanying drawing and specific embodiment that the present invention is done further explanation below.

The present invention is in order to provide a kind of similarity measure method of interval intuitionistic Fuzzy Sets; Cosine similarity measure in the common fuzzy set is extended to interval intuitionistic fuzzy concentrates, set up cosine similarity measure that interval upper lower limit value constituted and weighting cosine similarity measure by membership function and non-membership function; The weighting cosine similarity measure that foundation is made up of many characteristics such as bubble area, width and height in the two phase flow pattern image provides a kind of flow pattern of gas-liquid two-phase flow identification new method.

The principle of the invention is following:

The interval of definition intuitionistic Fuzzy Sets

If X is nonempty set, then claim:

$\tilde{A}=\{<x,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}\left(x\right),{\tilde{v}}_{\tilde{A}}\left(x\right)>|x\&Element;X\}---\left(a\right)$

Be interval intuitionistic Fuzzy Sets, note is made IVIFS (X), wherein:

${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}\left(x\right)=[{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)]\&Subset;\left[\mathrm{0,1}\right]---\left(b\right)$

${\tilde{v}}_{\tilde{A}}\left(x\right)=[{\tilde{v}}_{\tilde{A}}^L\left(x\right),{\tilde{v}}_{\tilde{A}}^U\left(x\right)]\&Subset;\left[\mathrm{0,1}\right]---\left(c\right)$

With Be respectively that x belongs to

Membership function and non-membership function,

With

The upper and lower bound of representing membership function respectively, With

The lower limit and the upper limit of representing non-membership function respectively, and must satisfy ${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)+{\tilde{v}}_{\tilde{A}}^U\left(x\right)\&le;1.$

Order ${\widetilde{\mathrm{\&pi;}}}_{\tilde{A}}^L{=1-\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)-{\tilde{v}}_{\tilde{A}}^U\left(x\right)---\left(d\right)$

${\widetilde{\mathrm{\&pi;}}}_{\tilde{A}}^U{=1-\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x\right)-{\tilde{v}}_{\tilde{A}}^L\left(x\right)---\left(e\right)$

Called

is the x belong hesitation degree.

(2) to establish two IVIFSs

and

cosine similarity measure between the constructed vectors:

$V_{\tilde{A}}\left(x_i\right)=\{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right),{\tilde{v}}_{\tilde{A}}^L\left(x_i\right),{\tilde{v}}_{\tilde{A}}^U\left(x_i\right)\}---\left(f\right)$

$V_{\tilde{B}}\left(x_i\right)=\{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right),{\tilde{v}}_{\tilde{B}}^L\left(x_i\right),{\tilde{v}}_{\tilde{B}}^U\left(x_i\right)\}---\left(g\right)$

Make up the cosine similarity measure:

$C_{\mathrm{IVIFS}}(\tilde{A},\tilde{B})=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^L\left(x_i\right){\tilde{v}}_{\tilde{B}}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^U\left(x_i\right){\tilde{v}}_{\tilde{B}}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{U^2}\left(x_i\right)}}---\left(h\right)$

(3) make up weighting cosine similarity measure

${\mathrm{Cw}}_{\mathrm{IVIFS}}(\tilde{A},\tilde{B})=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^L\left(x_i\right){\tilde{v}}_{\tilde{B}}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^U\left(x_i\right){\tilde{v}}_{\tilde{B}}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{U^2}\left(x_i\right)}}---\left(i\right)$

Wherein, w _i∈ [0,1], i=1,2 .., n, and

The cosine similarity is the tolerance of angle (cosine) between two vectors in fact, if the cosine similarity is 1, then angle is 0 ° between two vectors, and except that size (length), two vectors are identical; If the cosine similarity is 0, then angle is 90 ° between two vectors, shows between these two vectors not comprise any similar characteristic.The cosine similarity measure has not only embodied the similarity relation between the vector, and has comprised the changing condition of vectorial inner element.

Concrete steps of the present invention comprise:

Definite weighting cosine similarity measure of being made up of the many characteristics of two phase flow pattern is according to similarity measure identification flow pattern

(1) the definition flow pattern of gas-liquid two-phase flow is interval intuitionistic Fuzzy Sets

If

and

representes five kinds of basic flow patterns of biphase gas and liquid flow respectively: flow of bubble; Slug flow; Stratified flow (comprising laminar flow and wave flow); Slug flow and mist annular flow.

(2) utilize digital image processing techniques to extract bubble shape characteristic in the flow pattern of gas-liquid two-phase flow

Obtain the flow image of biphase gas and liquid flow through high speed photography; Utilize existing poor shadow algorithm to eliminate background noise, utilize process of iteration that image is carried out optimal threshold and cut apart and obtain binary image, for the glutinous situation about connecting of bubble; Employing is based on morphologic dividing method; Handle area x1, width x2 and the height x3 characteristic parameter that extracts bubble in the binary image in the back through form, above-mentioned concrete steps have been utilized existing technology, here repeat no more.

(3) confirm the interval range of each membership function and non-membership function

Concrete steps are following:

1. the membership function and the non-membership function that define relevant bubble area x1, width x2 and the height x3 characteristic parameter of flow pattern

are respectively

and and

and wherein, k=1 ~ 5;

2. obtain the characteristic parameters such as area x1, width x2 and height x3 of bubble in the biphase gas and liquid flow binary image through step (2) digital image processing techniques;

3. judge two phase flow pattern through ocular estimate, thus the corresponding relation of characteristic parameter such as the x1 that 2. determination step obtains, x2 and x3 and actual flow pattern;

4. repeating step 2. with step 3.; Extract great amount of samples and confirm the membership function of characteristic parameters such as bubble area x1, width x2 and height x3 and the bound of non-membership function:

and

and and and

and

and

He

He

He

wherein, k=1 ~ 5;

5. utilize the membership function that 4. the normalization method treatment step obtain and the bound of non-membership function, make satisfied:

${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_i\right)=[{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^L\left(x\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^U\left(x_i\right)]\&Subset;\left[\mathrm{0,1}\right],$ ${\tilde{v}}_{\tilde{A}k}\left(x_i\right)=[{\tilde{v}}_{\tilde{A}k}^L\left(x_i\right),{\tilde{v}}_{\tilde{A}k}^U\left(x_i\right)]\&Subset;\left[\mathrm{0,1}\right],$

Wherein, k=1 ~ 5, i=1,2,3; Then,

${\tilde{A}}_k=\{<x_1,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_1\right),{\tilde{v}}_{\tilde{A}k}\left(x_1\right)>,<x_2,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_2\right),{\tilde{v}}_{\tilde{A}k}\left(x_2\right)>,<x_3,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_3\right),{\tilde{v}}_{\mathrm{Ak}}\left(x_3\right)>|x_1,x_2,x_3\&Element;X\},$

Wherein, K=1 ~ 5 are represented the five kinds of basic flow patterns

of biphase gas and liquid flow and the interval intuitionistic Fuzzy Sets of respectively.

(4) set up the weighting cosine similarity measure of forming by each flow pattern more characteristic parameters

1. great amount of samples collection

j that extracts two phase flow pattern is concrete sample number, calculates the wherein pairing bubble area x1 of each flow pattern, width x2 and height x3;

2. according to membership function and the non-membership function of the bubble area x1 that has confirmed in the step (3), width x2 and height x3;

is the basis with the average weighted coefficient, the weighting cosine similarity measure in the calculating sample set between each concrete flow pattern

and each standard flow pattern

and

:

${\mathrm{Cw}}_{\mathrm{IVIFS}}(\widetilde{A_k},\widetilde{B_j})=\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}{w}_i\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^L\left(x_i\right){\tilde{v}}_{\tilde{B}j}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^U\left(x_i\right){\tilde{v}}_{\tilde{B}j}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}j}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}j}^{U^2}\left(x_i\right)}}---\left(j\right)$

3. calculate k=1 ~ 5 by formula (j), according to the flow pattern classification that maximum membership grade principle is differentiated

.In reality, can be further, with recognition result and visual observation relatively, again weighting coefficient is suitably revised according to recognition result then, thereby revised the weighting coefficient of each degree of membership and non-membership function: w _i∈ [0,1], Make it more accurate.

(5) any given flow pattern is carried out flow pattern identification

If flow pattern to be identified is that

is according to maximum membership grade principle; if

(k=1 ~ 5; I=1 ~ 5); Then the weighting cosine similarity measure between two interval intuitionistic Fuzzy Sets and

is maximum; Angle is minimum, and then

is under the jurisdiction of

relatively

Referring to Fig. 1, implementation process of the present invention is: at first utilize high speed photography to obtain the flow image of biphase gas and liquid flow, utilize digital image processing techniques to extract characteristic parameters such as the area x1 of bubble in the two-phase flow binary image, width x2 and height x3; Then according to the flow pattern characterizing definition five class standard flow patterns of two-phase flow: flow of bubble

Slug flow Stratified flow

(comprising laminar flow and wave flow), slug flow

With the mist annular flow And the relevant membership function of definite each standard flow pattern and the interval range of non-membership function; Set up the weighting cosine similarity measure of forming by each flow pattern more characteristic parameters then, confirm weighting coefficient w _i∈ [0,1]; Calculate flow pattern to be identified then

With each standard flow pattern

With

Between weighting cosine similarity measure: at last according to maximum membership grade principle, if

(k=1 ~ 5, i=1 ~ 5), then two interval intuitionistic Fuzzy Sets

With

Between weighting cosine similarity measure maximum, angle is minimum, then Be under the jurisdiction of relatively

Fig. 2 is the experimental provision described in the present invention, and this experimental provision mainly comprises two parts: fluid control systems and image capturing system.

Fluid control systems mainly is made up of air compressor, gas-holder, water pump, water tank and testing tube.Test used liquid phase from water pump, gas phase is a pressurized air, after pressure regulator, control valve are injected mixer, flows into transparent test pipeline section.It is 50mm transparent organic glass pipe that testing tube is selected internal diameter for use, and volumetric flow of gas is 0 ~ 8.25 * 10 ^-3m ³/ s, discharge is 0 ~ 3.33 * 10 ^-3m ³/ s.In actual production process, exist gas, the liquid two-phase of many high pressure; And adopt metallic conduit mostly; Need this moment one section see-through section or transparent window that can bear high pressure to be set, therefore need the transparent pipeline that adopts special materials such as sapphire, quartz glass or boronation glass to make in the appropriate location of pipeline.

Image capturing system mainly comprises illuminator, camera, image pick-up card and computing machine.Wherein, the illuminator light source uses the three primary colours light pipe of 5000K colour temperature, and light is stable, flicker free; Image recording system adopts the U.S. UP-900 digital ccd video camera of lining by line scan, and its resolution is 1392 * 1040, and frame frequency is 15FPS, and shutter speed is 1/15 ~ 1/31000s; Image pick-up card adopts the PCI1422 of America NI company, 16M onboard memory.In the image capture process,,, take the shade of bubble so adopt backlighting because liquid and bubble all are transparent.In order to obtain satisfied picture quality, on the rear side plexi-glass tubular of bubble flow pattern, cover with two-layer drawing and used template, make illumination patterns even.

Referring to Fig. 3, in the horizontal checkout pipe, collected size and be 760 * 275 typical flow pattern image.Can know by observing, (a) be flow of bubble, in continuous liquid phase, contains the minute bubbles of dispersion, and bubble trends towards flowing along the pipeline top; (b) be slug flow, contain the bullet shaped air pocket that flows in the pipeline top in the liquid phase; (c) be stratified flow, gas flows along the pipeline top liquid in that duct bottom is mobile, has level and smooth interphase or gas-liquid two-phase interphase to be wavy between the two; (d) be slug flow, in the slug flow, when steam bubble concentration increases, the effect of effect in being become, small bubble aggregates into large sparkle, and diameter increases gradually.When steam bubble dia approaches ips, form slug flow; (e) be the mist annular flow, gas entrainment drop and is partly flowed in pipeline center, and liquid forms liquid film and flows forward along tube wall, and partially liq flows downward along inner-walls of duct, and when gas flow rate was accelerated, gas-liquid two-phase presented disperse state.More than be that the observer directly confirms flow pattern according to the form of two-phase flow flow image, but, also must from the gained flow image, extract characteristic in order to make the calculating function carry out flow pattern identification automatically, and then according to characteristics of image identification flow pattern.

The above only is a preferred implementation of the present invention; Should be pointed out that for those skilled in the art, under the prerequisite that does not break away from the present invention's design; Can also make some improvement and retouching, these improvement and retouching also should be regarded as in the protection domain of the present invention.

Claims (4)

1. a Method for Discriminating Gas-liquid Two Phase Flow is characterized in that, comprises the steps:

11) utilize high speed photography to obtain the flow image of biphase gas and liquid flow;

12) utilize digital image processing techniques to extract bubble area x1 in the two-phase flow binary image, width x2 and height x3 characteristic parameter;

13) based on two-phase flow characteristics defined five standard flow patterns: bubbly flow

plug flow

stratified flow

slug flow

and fog annular flow

14) with the membership function and the non-membership function of the bubble area x1 that obtains, width x2 and height x3 characteristic parameter substitution flow of bubble

slug flow stratified flow

slug flow

and mist annular flow ; Said membership function and non-membership function are respectively

and

and and

wherein, k=1 ~ 5;

15) repeating step 12) thereby and step 14) confirm membership function and non-membership function interval range

and and

and

and

and

He He

and

He

wherein, k=1 ~ 5;

16) utilize normalization method treatment step 15) membership function that obtains and the bound of non-membership function, make satisfied:

${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_i\right)=[{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^L\left(x\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^U\left(x_i\right)]\&Subset;\left[\mathrm{0,1}\right],$ ${\tilde{v}}_{\tilde{A}k}\left(x_i\right)=[{\tilde{v}}_{\tilde{A}k}^L\left(x_i\right),{\tilde{v}}_{\tilde{A}k}^U\left(x_i\right)]\&Subset;\left[\mathrm{0,1}\right],$ Wherein, k=1 ~ 5, i=1,2,3; Then ${\tilde{A}}_k=\{<x_1,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_1\right),{\tilde{v}}_{\tilde{A}k}\left(x_1\right)>,<x_2,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_2\right),{\tilde{v}}_{\tilde{A}k}\left(x_2\right)>,<x_3,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}\left(x_3\right),{\tilde{v}}_{\mathrm{Ak}}\left(x_3\right)>|x_1,x_2,x_3\&Element;X\},$ Wherein, k=1 ~ 5 are represented five kinds of basic flow patterns of biphase gas and liquid flow respectively

With

Interval intuitionistic Fuzzy Sets;

17) set up the weighting cosine similarity measure of forming by each flow pattern more characteristic parameters, confirm weighting coefficient w _i∈ [0,1]; Calculate flow pattern to be identified With each standard flow pattern

With

Between weighting cosine similarity measure: according to maximum membership grade principle, if

(k=1 ~ 5, i=1 ~ 5), then two interval intuitionistic Fuzzy Sets

With

Between weighting cosine similarity measure maximum, angle is minimum, then

Be under the jurisdiction of relatively

2. a kind of Method for Discriminating Gas-liquid Two Phase Flow according to claim 1 is characterized in that, establishing X is nonempty set, then claims:

$\tilde{A}=\{<x,{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}\left(x\right),{\tilde{v}}_{\tilde{A}}\left(x\right)>|x\&Element;X\}---\left(1\right)$

Be interval intuitionistic Fuzzy Sets, note is made IVIFS (X), wherein:

${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}\left(x\right)=[{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)]\&Subset;\left[\mathrm{0,1}\right]---\left(2\right)$

${\tilde{v}}_{\tilde{A}}\left(x\right)=[{\tilde{v}}_{\tilde{A}}^L\left(x\right),{\tilde{v}}_{\tilde{A}}^U\left(x\right)]\&Subset;\left[\mathrm{0,1}\right]---\left(3\right)$

With

Be respectively that x belongs to

Membership function and non-membership function, With The upper and lower bound of representing membership function respectively,

With

The lower limit and the upper limit of representing non-membership function respectively, and must satisfy ${\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)+{\tilde{v}}_{\tilde{A}}^U\left(x\right)\&le;1,$ Order ${\widetilde{\mathrm{\&pi;}}}_{\tilde{A}}^L{=1-\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x\right)-{\tilde{v}}_{\tilde{A}}^U\left(x\right);$ ${\widetilde{\mathrm{\&pi;}}}_{\tilde{A}}^U{=1-\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x\right)-{\tilde{v}}_{\tilde{A}}^L\left(x\right);$ Then claim

For x belongs to The hesitation degree.

3. a kind of Method for Discriminating Gas-liquid Two Phase Flow according to claim 1 is characterized in that, makes up vector:

$V_{\tilde{A}}\left(x_i\right)=\{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right),{\tilde{v}}_{\tilde{A}}^L\left(x_i\right),{\tilde{v}}_{\tilde{A}}^U\left(x_i\right)\}---\left(4\right)$

$V_{\tilde{B}}\left(x_i\right)=\{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right),{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right),{\tilde{v}}_{\tilde{B}}^L\left(x_i\right),{\tilde{v}}_{\tilde{B}}^U\left(x_i\right)\}---\left(5\right)$

Make up the cosine similarity measure:

$C_{\mathrm{IVIFS}}(\tilde{A},\tilde{B})=\frac{1}{n}\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^L\left(x_i\right){\tilde{v}}_{\tilde{B}}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^U\left(x_i\right){\tilde{v}}_{\tilde{B}}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{U^2}\left(x_i\right)}}---\left(6\right)$

Make up weighting cosine similarity measure:

${\mathrm{Cw}}_{\mathrm{IVIFS}}(\tilde{A},\tilde{B})=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_i\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^L\left(x_i\right){\tilde{v}}_{\tilde{B}}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^U\left(x_i\right){\tilde{v}}_{\tilde{B}}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}}^{U^2}\left(x_i\right)}}---\left(7\right)$

Wherein, w _i∈ [0,1], i=1,2 ..., n, and

4. a kind of Method for Discriminating Gas-liquid Two Phase Flow according to claim 1 is characterized in that, said step 17) concrete steps comprise:

41) great amount of samples collection

j of extraction two phase flow pattern is concrete sample number, calculates the wherein pairing bubble area x1 of each flow pattern, width x2 and height x3;

42) according to membership function and the non-membership function of the bubble area x1 that has confirmed in the step 12), width x2 and height x3; is the basis with the average weighted coefficient, the weighting cosine similarity measure in the calculating sample set

between each concrete flow pattern

and each standard flow pattern

and

:

${\mathrm{Cw}}_{\mathrm{IVIFS}}(\widetilde{A_k},\widetilde{B_j})=\underset{i=1}{\overset{3}{\mathrm{\&Sigma;}}}{w}_i\frac{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^L\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^L\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^U\left(x_i\right){\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^U\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^L\left(x_i\right){\tilde{v}}_{\tilde{B}j}^L\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^U\left(x_i\right){\tilde{v}}_{\tilde{B}j}^U\left(x_i\right)}{\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{A}k}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{A}k}^{U^2}\left(x_i\right)}\sqrt{{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^{L^2}\left(x_i\right)+{\widetilde{\mathrm{\&mu;}}}_{\tilde{B}j}^{U^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}j}^{L^2}\left(x_i\right)+{\tilde{v}}_{\tilde{B}j}^{U^2}\left(x_i\right)}}---\left(8\right)$

43) calculate by formula (8)

K=1 ~ 5 are differentiated according to maximum membership grade principle

The flow pattern classification, recognition result and visual observation are relatively suitably revised weighting coefficient according to recognition result then again, thus through revising the weighting coefficient of confirming each degree of membership and non-membership function: w once more _i∈ [0,1],

Improve precision.

Images