CN102620905A - Device and method for identifying fluid type of high-pressure fluid in pipeline during rapid pressure change - Google Patents

Abstract

The invention discloses a method and device for identifying a fluid type of a high-pressure fluid in a pipeline during rapid pressure change, belonging to the field of fluid type identification. The device mainly comprises an incoming fluid pipeline, a fixed baffle, an incoming fluid pipeline joint, an incoming fluid sensor three-way joint, a high-pressure resistant quartz transparent tube, an outgoing fluid sensor three-way joint, an outgoing fluid pipeline joint, an outgoing fluid pipeline, an adjusting sleeve, an adjusting bolt with a through hole, an incoming fluid pressure sensor, an incoming fluid pressure transmitter, an outgoing fluid pressure sensor, an outgoing fluid pressure transmitter, a computer, a light source and a high-speed video camera. The pipeline is sealed by adjusting the screwing length of the adjusting sleeve and the adjusting bolt with the through hole. The method comprises the steps of: performing Hilbert-Huang transformation on pressure differential data to obtain each modulus energy ratio; then regarding each modulus energy ratio as input vectors of an Elman neural network to finish mapping from a characteristic space to a fluid type space. The method and device for identifying the fluid type of the high-pressure fluid in the pipeline during the rapid pressure change, disclosed by the invention, has advantages of simple structure, convenience in detachment and capability of identifying fluid type of high-pressure fluid in the pipeline rapidly and exactly during the rapid pressure change.

Description

Flow pattern recognition device and method when pipeline mesohigh hydrodynamic pressure changes fast

Technical field

The present invention relates to a kind of flow pattern recognition device and method, be specifically related to flow pattern recognition device and the method for a kind of pipeline mesohigh hydrodynamic pressure when changing fast.

Background technology

Pressure duct is prevalent in all trades and professions, when higher and pressure frequently shakes when fluid temperature in the pipeline, just has to consider the influence of flow pattern of gas-liquid two-phase flow.When the pressure of liquid in the pressure duct drops to saturated solution presasure when following, the free gas (like air) that originally is dissolved in the liquid will some discharge, and rapidly to discrete gas nuclear proliferation, forms micro-bubble.When pressure continues to drop to the saturated vapour pressure of liquid when following, the part in some liquid is gently heated up in a steamer composition and is begun to undergo phase transition, and the steam of generation is also assembled near gas nuclear rapidly, the formation minute bubbles.Like this, the liquid in the pressure duct just exists with the form of biphase gas and liquid flow, and major part is and the steam of liquid phase homogeneity in the bubble, and fraction is and the inhomogeneous free gas of liquid phase.On the contrary, when the pressure in the pressure duct is raised to when being higher than saturated vapour pressure, the steam in the bubble will become liquid rapidly, and free gas also can be dissolved in the liquid simultaneously.But the speed that the speed of gas dissolving discharges much smaller than gas, after steam all becomes liquid, some free gas bubble still in the liquid.Therefore frequently shake at pressure, particularly minimum pressure drops to the saturated vapour pressure of liquid when following in the pipeline, is accompanied by the generation of bubble and shattered to pieces in the pressure duct, and the medium in the pipeline just exists with the form of biphase gas and liquid flow.

Bubble can be accompanied by complicated physics, chemical phenomenon in generation, development and process of collapse, show strong vibration, noise and to the erosion of flow channel material etc.This can be to the measurement of fluid flow parameter in the pipeline and the very big influence of mobile generation of liquid.Strong vibration and noise can make parameter measurements lose contact with reality, and be extremely inaccurate, and this will make the measurement behavior lose meaning.If bubble has caused the formation of vapour lock in the pipeline, will have a strong impact on the serviceability of pipeline place system, even can cause its complete failure.When a large amount of generations of bubble and sustainable development, can have a strong impact on the tube wall generation, it is cellular or spongy that it is become.

Bubble is very big to the influence of pressure duct, and bubble quantity and bubble is different with distribution situation, and the difference of influence is also different.Be the validity and the security of assurance pressure duct work, and the needs of measuring, just need be distributed with more deep understanding to bubble wherein, promptly the flow pattern of biphase gas and liquid flow is discerned.Therefore, flow pattern of gas-liquid two-phase flow identification has obtained the concern of all trades and professions, develop very soon, and the new method appearance is constantly arranged.

Since Baker in 1954 drew out first flow regime map, flow regime map had become the main method of judging flow pattern in 20th century engineering.But flow regime map can only qualitative discrimination, and universality is bad, and usable range is very limited.

The various countries scholar is doing many research work aspect the change of flow relational expression, but owing to lack the further investigation of convection transformation mechanism, all there is certain error in the model of being set up, and the relational expression that different researchers obtains is also not quite identical.This has hindered the widespread use of change of flow criterion in flow pattern identification.

Modern flow type identification method mainly contains the direct method of measurement and the indirect method of measurement.The direct method of measurement mainly comprises ocular estimate, high-speed photography method etc.Ocular estimate and high-speed photography method are simple to operate, but flow pattern identification relies on experimenter's subjective judgement, lacks enough objective basis.The indirect method of measurement comprises process tomographic imaging method, pressure surge method and pressure-difference fluctuation method etc.The image reconstruction algorithm of process tomographic imaging method is complicated, poor, the soft field characteristic of reconstruction quality can't overcome; It is very big that force value is influenced by extraneous various factors, even pressure signal is flooded.In Method for Discriminating Gas-liquid Two Phase Flow, the pressure reduction method of identification has the advantage that can not be substituted, and can satisfy fast, accurately and be easy to requirement such as realization, and the pressure reduction method of identification has realistic meaning in pressure duct.

The pressure difference data that measures is carried out the processing of different modes, and its flow pattern accuracy of identification also can produce very big-difference.Data processing method commonly used has statistical analysis technique, Fourier transform, wavelet analysis, Hilbert-Huang transform (empirical modal decomposition), information fusion etc.Can directly adopt eigenwert to carry out flow pattern identification after data are handled, also can select sample that sorter is trained, and then the identification flow pattern.A kind of method in back can be more convenient, more promptly convection is discerned.Sorter commonly used has SVMs (SVM) and neural network etc.

Patent " based on the Method for Discriminating Gas-liquid Two Phase Flow and the flow pattern signal pickup assembly of information fusion " (application number 200610017091.5; Inventor: Zhou Yunlong etc.) introduced a kind of pressure-difference fluctuation signal and carried out small echo and handle to three different pressure spacings of horizontal pipeline of collecting; Form three wavelet packet information entropy proper vectors and as the input feature value of three RBF sub neural networks; Again the output of each sub neural network as evidence body independent of each other; Utilize the D-S evidence theory to carry out information fusion, thereby obtain the method for flow pattern identification and set up corresponding flow pattern signal pickup assembly.Patent " based on the horizontal tube Method for Discriminating Gas-liquid Two Phase Flow of Hilbert-Huang transform " (application number 200410017475.8; The inventor: shared energy ratio and the residual error average of the different natural mode of vibration of pressure difference signal set up fluidised form figure in the Venturi tube that the flow pattern recognition system of Sun Bin etc.) utilizing classical Venturi tube, differential pressure transmitter and computing machine to form provides, and proposed a kind of method of carrying out flow pattern identification.Patent " based on the oil gas water multiphase flow type identification method of principal component analysis (PCA) and SVMs " (application number 200610017090.0; Inventor: Sun Bin etc.) proposed a kind of the pressure signal of horizontal pipeline and the pressure difference signal of different pressure spacings to be carried out small echo denoising pre-service earlier; Carrying out empirical modal then decomposes; With the intrinsic mode function composition characteristic matrix that obtains, utilize principal component analysis (PCA) to obtain proper vector again, and with its input sample as SVM; Use SVM and accomplish mapping, finally realize the method for flow pattern identification from feature space to the flow pattern space.

Existing flow pattern signal pickup assembly mainly is made up of Venturi tube, pressure unit, computing machine or is made up of pressure ring, pressure unit, computing machine.The pressure that Venturi tube and pressure ring can bear is all not high enough, and (classical Venturi tube is generally less than 63MPa; The pressure ring is generally less than 50MPa); And the pressure of a lot of pressure ducts has all surpassed 80MPa; Even reached more than the 100MPa, and its pressure surge frequency even can reach more than tens hertz.Therefore, existing flow pattern signal pickup assembly can't be used in the pressure duct that hydrodynamic pressure changes fast.

Existing flow type identification method mainly is that technology such as utilization wavelet analysis, Hilbert-Huang transform (empirical modal decomposition), SVMs (SVM), neural network and information fusion are discerned flow pattern.Wavelet analysis can carry out the analysis of time domain and frequency domain simultaneously with resolution arbitrarily to signal, but wavelet basis is an empirical value with decomposing the number of plies, need relatively obtain through a large amount of experiments; Hilbert-Huang transform (empirical modal decomposition) has been drawn the wavelet transformation advantages of differentiating more; Overcome the difficulty that to choose wavelet basis in the wavelet transformation simultaneously; The characteristic of retention data itself in the process of non-linear, non-stationary signal being carried out linearity and tranquilization processing, and have good local adaptation's property; SVM can solve practical problemss such as small sample, non-linear and higher-dimension pattern-recognition preferably; And have advantages such as structure is simple relatively, fast convergence rate; But kernel function and model parameter be chosen in the performance that is determining SVM to a great extent, and can't find optimized parameter at present very exactly; Neural network has very strong nonlinear fitting ability, robustness and memory capability, and learning rules are simple, are convenient to computer realization; Information fusion technology may cause information redundancy, increases the burden of sorter, and influence identification accuracy must find suitable method to reject irrelevant information in the application.

Summary of the invention

The object of the invention is exactly the defective that overcomes above-mentioned prior art, designs a kind ofly can fast, accurately discern flow pattern recognition device and the method for pipeline mesohigh fluid when pressure changes fast, and concrete technical scheme is following:

Flow pattern recognition device when a kind of pipeline mesohigh hydrodynamic pressure changes fast, mainly comprise incoming flow pipeline, fixed dam, incoming flow pipe joint, incoming flow joint packing, come the flow sensor three-way connection, incoming flow soft washer, high pressure resistant quartzy transparent tube, go to flow soft washer, remove the flow sensor three-way connection, go to flow joint packing, go to flow pipe joint, go to flow pipeline, regulating sleeve, band through hole adjusting bolt, incoming-flow pressure sensor, incoming-flow pressure transmitter, incoming flow sensor packing ring, remove the flowing pressure sensor, remove the flowing pressure transmitter, remove flow sensor packing ring, computing machine, light source, high-speed camera.Incoming flow pipeline and incoming flow pipe joint weld together; The incoming flow pipe joint with come the flow sensor three-way connection to adopt to be threaded incoming flow pipe joint and come to be equipped with between the flow sensor three-way connection incoming flow joint packing; Come between flow sensor three-way connection and the high pressure resistant quartzy transparent tube incoming flow soft washer to be housed; High pressure resistant quartzy transparent tube and go to be equipped with between the flow sensor three-way connection and flow soft washer is come the flow sensor three-way connection, is gone the flow sensor three-way connection to adopt clamped-in style to be connected with high pressure resistant quartzy transparent tube; Go the flow sensor three-way connection to adopt and be threaded, remove the flow sensor three-way connection and go to flow to be equipped with between the pipe joint and flow joint packing with going to flow pipe joint; Go to flow pipe joint and go to flow pipeline welding together; Regulating sleeve with go to flow pipe joint and be close to, the internal thread of regulating sleeve matches with the external thread of the adjusting bolt of band through hole, is with the adjusting bolt sleeve of through hole going to flow on the pipeline; The incoming flow pipeline, go to flow the hole that pipeline passes the fixed dam two ends respectively, incoming flow pipe joint and fixed dam are close to, and the adjusting bolt and the fixed dam of band through hole are close to; The incoming-flow pressure sensor with come the flow sensor three-way connection to adopt to be threaded incoming-flow pressure sensor and come to be equipped with between the flow sensor three-way connection incoming flow sensor packing ring; The incoming-flow pressure transmitter is connected with the incoming-flow pressure sensor, and the incoming-flow pressure transmitter is connected with computing machine; Go the flowing pressure sensor to be threaded, remove the flowing pressure sensor and remove to be equipped with between the flow sensor three-way connection flow sensor packing ring with going the employing of flow sensor three-way connection; Go the flowing pressure transmitter to be connected, go the flowing pressure transmitter to be connected with computing machine with removing the flowing pressure sensor; Light source shines the fluid in the high pressure resistant quartzy transparent tube; High-speed camera and computing machine link together, and high-speed camera is taken the mobility status of fluid in the high pressure resistant quartzy transparent tube.

The screwing length of the present invention through adjustment regulating sleeve and the adjusting bolt of being with through hole realize the sealing between flow sensor three-way connection and the high pressure resistant quartzy transparent tube and remove the flow sensor three-way connection and high pressure resistant quartzy transparent tube between sealing.

Flow type identification method when a kind of pipeline mesohigh hydrodynamic pressure changes fast comprises the steps:

1) obtain data: the incoming-flow pressure sensor is measured pipeline inner fluid incoming-flow pressure; Go the flowing pressure sensor to go flowing pressure to measure to the pipeline inner fluid; Write down the flow pattern of fluid in the high pressure resistant quartzy transparent tube simultaneously with high-speed camera;

2) transmission data: the incoming-flow pressure that the incoming-flow pressure transmitter obtains the incoming-flow pressure sensor measurement is imported computing machine and is preserved; Go the flowing pressure transmitter to go flowing pressure input computing machine to what go that the flowing pressure sensor measurement obtains and preserve, high-speed camera deposits the flow pattern of noting in computing machine;

3) data are carried out pre-service: the incoming-flow pressure of getting off computer recording and corresponding time go flowing pressure to subtract each other to obtain corresponding pressure difference data;

4) sorter is trained: the corresponding pressure difference data of selecting to photograph with high-speed camera of typical flow pattern is carried out Hilbert-Huang transform; Obtain conversion spectrum; Calculate the energy ratio of each mode; And liken each mode energy to neural network into sample input Elman, and train, obtain the flow pattern criterion of identification;

5) actual flow pattern is discerned: the corresponding pressure difference data of flow pattern to be identified is carried out Hilbert-Huang transform; Obtain conversion spectrum; Calculate the energy ratio of each mode, and liken each mode energy to train neural network, can obtain corresponding flow pattern for the characteristic quantity input.

Of the present invention pressure difference data is carried out Hilbert-Huang transform, be to be decomposed into 8 natural mode of vibration components and 1 residual components to pressure difference data, and carry out Hilbert transform respectively, obtain hilbert spectrum, and then obtain shared energy ratio.

Elman neural network of the present invention be a kind of by input layer, middle layer (hidden layer), accept the recurrent nerve metanetwork that layer and output layer are formed; Being connected of its input layer, hidden layer and output layer is similar to feedforward network; The effect of signal transmission is only brought into play in the unit of input layer; The transport function of hidden layer unit can adopt linearity or nonlinear function; Accept layer be otherwise known as context layer or state layer, it is used for remembering the output valve of implicit unit previous moment, can be considered to one 1 step time-delay operator.The output of hidden layer is through accepting the time-delay and the storage of layer; From the input that is linked to hidden layer; Thisly make the Elman neural network have susceptibility, and the internal feedback network increased network itself and handled the ability of multidate information, thereby reached the purpose of dynamic modeling historical data from the couplet mode.

Flow pattern identification when apparatus and method of the present invention can be used for pipeline mesohigh hydrodynamic pressure and change fast.Its beneficial effect is embodied in:

1. flow pattern signal pickup assembly of the present invention designs to the pressure duct that hydrodynamic pressure changes fast; Whole apparatus structure is simple, and easy accessibility can be changed rapidly when high pressure resistant quartzy transparent tube is damaged; And through of the sealing of adjustment regulating sleeve with the screwing length assurance device of the adjusting bolt of band through hole; Prevent the leakage of fluid in the pressure duct, this device can be measured the pressure at pressure duct two places fast and accurately, and then draws the pressure reduction of respective distances;

2. flow type identification method of the present invention combines Hilbert-Huang transform and Elman neural network; After the completion of Elman neural metwork training; No longer utilize high-speed camera that tube fluid is taken, and directly utilize the pressure difference data that measures just can realize the biphase gas and liquid flow that pressure in the pressure duct changes is fast carried out flow pattern identification fast and accurately.

Description of drawings

Fig. 1 is the measurement mechanism structural representation.

1 is the incoming flow pipeline among the figure; 2 is fixed dam; 3 is the incoming flow pipe joint; 4 is the incoming flow joint packing; 5 for coming the flow sensor three-way connection; 6 is the incoming flow soft washer; 7 is high pressure resistant quartzy transparent tube; 8 for removing to flow soft washer; 9 for removing the flow sensor three-way connection; 10 for removing to flow joint packing; 11 for going to flow pipe joint; 12 for removing to flow pipeline; 13 is regulating sleeve; 14 adjusting bolts for the band through hole; 15 is the incoming-flow pressure sensor; 16 is the incoming-flow pressure transmitter; 17 is incoming flow sensor packing ring; 18 for removing the flowing pressure sensor; 19 for removing the flowing pressure transmitter; 20 for removing the flow sensor packing ring; 21 is computing machine; 22 is light source; 23 is high-speed camera.

Embodiment

Below in conjunction with accompanying drawing, further specify the present invention with embodiment.But this embodiment only is illustrative, and protection scope of the present invention does not receive the restriction of this embodiment.

As shown in Figure 1, the flow pattern recognition device when pipeline mesohigh hydrodynamic pressure of the present invention changes fast mainly comprises incoming flow pipeline 1, fixed dam 2, incoming flow pipe joint 3, incoming flow joint packing 4, comes flow sensor three-way connection 5, incoming flow soft washer 6, high pressure resistant quartzy transparent tube 7, go to flow soft washer 8, remove flow sensor three-way connection 9, go to flow joint packing 10, go to flow pipe joint 11, go to flow pipeline 12, regulating sleeve 13, band through hole adjusting bolt 14, incoming-flow pressure sensor 15, incoming-flow pressure transmitter 16, incoming flow sensor packing ring 17, remove flowing pressure sensor 18, remove flowing pressure transmitter 19, remove flow sensor packing ring 20, computing machine 21, light source 22, high-speed camera 23.Incoming flow pipeline 1 welds together with incoming flow pipe joint 3; Incoming flow pipe joint 3 with come flow sensor three-way connection 5 to adopt to be threaded incoming flow pipe joint 3 and come to be equipped with between the flow sensor three-way connection 5 incoming flow joint packing 4; Come between flow sensor three-way connection 5 and the high pressure resistant quartzy transparent tube 7 incoming flow soft washer 6 to be housed; High pressure resistant quartzy transparent tube 7 and go to be equipped with between the flow sensor three-way connection 9 and flow soft washer 8 is come flow sensor three-way connection 5, is gone flow sensor three-way connection 9 to adopt clamped-in styles to be connected with high pressure resistant quartzy transparent tube 7; Go flow sensor three-way connection 9 to adopt and be threaded, remove flow sensor three-way connection 9 and go to flow to be equipped with between the pipe joint 11 and flow joint packing 10 with going to flow pipe joint 11; Go to flow pipe joint 11 and remove to flow pipeline 12 and weld together; Regulating sleeve 13 with go to flow pipe joint 11 and be close to, the internal thread of regulating sleeve 13 matches with the external thread of the adjusting bolt 14 of band through hole, is with the adjusting bolt 14 of through hole to be enclosed within and flows on the pipeline 12; Incoming flow pipeline 1, go to flow the hole that pipeline 12 passes fixed dam 2 two ends respectively, incoming flow pipe joint 3 is close to fixed dam 2, and the adjusting bolt 14 of band through hole is close to fixed dam 2; Incoming-flow pressure sensor 15 with come flow sensor three-way connection 5 to adopt to be threaded incoming-flow pressure sensor 15 and come to be equipped with between the flow sensor three-way connection 5 incoming flow sensor packing ring 17; Incoming-flow pressure transmitter 16 is connected with incoming-flow pressure sensor 15, and incoming-flow pressure transmitter 16 is connected with computing machine 21; Go flowing pressure sensor 18 to be threaded, remove flowing pressure sensor 18 and remove to be equipped with between the flow sensor three-way connection 9 flow sensor packing ring 20 with going 9 employings of flow sensor three-way connection; Go flowing pressure transmitter 19 to be connected, go flowing pressure transmitter 19 to be connected with computing machine 21 with removing flowing pressure sensor 18; Fluid in 22 pairs of high pressure resistant quartzy transparent tube 7 of light source shines; High-speed camera 23 links together with computing machine 21, and the mobility status of fluid is taken in 23 pairs of high pressure resistant quartzy transparent tube 7 of high-speed camera.

Can realize the sealing between flow sensor three-way connection 5 and the high pressure resistant quartzy transparent tube 7 through the screwing length of adjustment regulating sleeve 13 and the adjusting bolt 14 of being with through hole after the assembling of flow pattern recognition device parts when pipeline mesohigh hydrodynamic pressure changes is fast accomplished and remove flow sensor three-way connection 9 and high pressure resistant quartzy transparent tube 7 between sealing.When the screwing length of regulating sleeve 13 and the adjusting bolt 14 of band through hole reduces; Because the adjusting bolt 14 of band through hole is close to the right-hand member of fixed dam 2; Can't move right; Regulating sleeve 13 will be moved to the left along going to flow pipeline 12; The power that is moved to the left will be delivered to and flow pipe joint 11; Because going to flow pipe joint 11 is that screw threads for fastening is connected with removing flow sensor three-way connection 9, power will be passed to flow sensor three-way connection 9, goes flow sensor three-way connection 9 to make it to be moved to the left through going to flow the high pressure resistant quartzy transparent tube 7 of soft washer 8 extruding; The power that is moved to the left will be passed to flow sensor three-way connection 5, incoming flow pipe joint 3 successively; The left end of baffle plate 2 blocks and can't continue to be moved to the left but incoming flow pipe joint 3 is fixed, finally will form come flow sensor three-way connection 5 with go flow sensor three-way connection 9 respectively through incoming flow soft washer 6 with go to flow the high pressure resistant quartzy transparent tube 7 of soft washer 8 extruding, realize the sealing between flow sensor three-way connection 5 and the high pressure resistant quartzy transparent tube 7 thus and remove flow sensor three-way connection 9 and high pressure resistant quartzy transparent tube 7 between sealing.

Flow pattern identification step of the present invention is following:

Obtain data: after the device connection finishes; The high-pressure fluid that cut-in pressure changes fast in pipeline; Give the parts that respectively need power supply power supply; 15 pairs of tube fluid incoming-flow pressure of incoming-flow pressure sensor are measured, and go 18 pairs of tube fluids of flowing pressure sensor to go flowing pressure to measure, and note the flow pattern of high pressure resistant quartzy transparent tube 7 medium simultaneously with high-speed camera 23;

The transmission data: the pipeline incoming-flow pressure that incoming-flow pressure transmitter 16 measures incoming-flow pressure sensor 15 is imported computing machine 21 and is preserved; The pipelines that go flowing pressure transmitter 19 handles to go flowing pressure sensor 18 to measure remove flowing pressure input computing machine 21 and preserve, and high-speed camera 23 deposits the flow patterns of noting in computing machine 21;

Data are carried out pre-service: the incoming-flow pressure of noting computing machine 21 and corresponding time go flowing pressure to subtract each other to obtain corresponding pressure difference data;

Sorter is trained: select 200 groups of corresponding pressure difference data of typical flow pattern that photograph with high-speed camera 23 to carry out Hilbert-Huang transform, obtain conversion spectrum, and then calculate the energy ratio of 8 modal components and 1 residual components; Remake the neural network for sample characteristics parameter input Elman, network is output as bubble flow (1000), intermittent flow (0100); Laminar flow (0010); Annular flow (0001) is accomplished the training to the Elman network, obtains the flow pattern criterion of identification;

Pressure difference data is carried out Hilbert-Huang transform, and then it is following to ask for the detailed process of each mode energy ratio:

(a) obtain the coenvelope F of pressure difference signal F (t) _s(t) and lower envelope F _x(t) mean value

$m=\frac{1}{2}[F_s\left(t\right)+F_x\left(t\right)]---\left(1\right)$

(b) ask for the difference F of F (t) and m ₁(t)

F ₁(t)＝F(t)-m (2)

(c) F ₁(t) regard F (t) as and repeat aforesaid operations, up to obtaining F _n(t), F _n(t) satisfy following two conditions

1) the extreme point quantity of signal equates with zero crossing quantity or differs one at the most;

The mean value of the upper and lower envelope that 2) is made up of signal local maximum and minimal value is zero.

F _n(t) be designated as C ₁, C ₁Be regarded as a mode function.

(d) try to achieve the difference of pressure difference data and modal components, be designated as R (t)

R(t)＝F(t)-C ₁ (3)

Regard R (t) as F (t), repeat aforesaid operations, obtain C successively ₂, C ₃, C ₄, C ₅, C ₆, C ₇, C ₈And R ₁(t) wherein

R ₁(t)＝F(t)-C ₈ (4)

So have

$F\left(t\right)=\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{C}_1+R_1\left(t\right)---\left(5\right)$

Be that former pressure difference data is broken down into 8 natural mode of vibration components and a residual components.

(e) to C ₁And R ₁(t) do Hilbert transform respectively and obtain analytic signal, with C ₁For example is explained

$H_1\left(t\right)=C_1\left(t\right)+i\frac{1}{\mathrm{\π}}{\∫}_{-\∞}^{+\∞}\frac{C_1\left(\mathrm{\τ}\right)}{t-\mathrm{\τ}}\mathrm{d\τ}=A_1\left(t\right)e^{i{\mathrm{\θ}}_1\left(t\right)}---\left(6\right)$

$A_1\left(t\right)={[{C_1\left(t\right)}^2+{\left[\frac{1}{\mathrm{\π}}{\∫}_{-\∞}^{+\∞}\frac{C_1\left(\mathrm{\τ}\right)}{t-\mathrm{\τ}}\mathrm{d\τ}\right]}^2]}^{\frac{1}{2}}---\left(7\right)$

${\mathrm{\θ}}_1\left(t\right)=\arctan \frac{\frac{1}{\mathrm{\π}}{\∫}_{-\∞}^{+\∞}\frac{C_1\left(\mathrm{\τ}\right)}{t-\mathrm{\τ}}\mathrm{d\τ}}{C_1\left(t\right)}---\left(8\right)$

H ₁(t) expression C ₁Do the analytic signal that Hilbert transform obtains, A ₁(t) expression amplitude, θ ₁(t) expression phase place.

R ₁(t) analytic signal is expressed as H _R(t), amplitude is expressed as A _R(t), phase meter is shown θ _R(t).

(f) calculate instantaneous frequency, with C ₁For example is explained

$f_1\left(t\right)=\frac{1}{2\mathrm{\π}}\frac{{\mathrm{d\θ}}_1\left(t\right)}{\mathrm{dt}}---\left(9\right)$

(g) after being done Hilbert transform, each natural mode of vibration component and residual components obtain

$F\left(t\right)=\mathrm{Re}[\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{A}_i\left(t\right)e^{j{\mathrm{\θ}}_1\left(t\right)}+A_R\left(t\right)e^{j{\mathrm{\θ}}_R\left(t\right)}]=\mathrm{Re}[\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{A}_i\left(t\right)e^{j{\∫\mathrm{\θ}}_1\left(t\right)\mathrm{dt}}+A_R\left(t\right)e^{j{\mathrm{\θ}}_R\left(t\right)}]---\left(10\right)$

Re representes to get real part.

The hilbert spectrum of pressure difference data does

$H(\mathrm{\θ},t)=\mathrm{Re}[\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{A}_i\left(t\right)e^{j{\∫\mathrm{\θ}}_1\left(t\right)\mathrm{dt}}+A_R\left(t\right)e^{j{\mathrm{\θ}}_R\left(t\right)}]---\left(11\right)$

(h) energy of each natural mode of vibration does

$E_i=\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{\left|A_i\left(t\right)\right|}^2---\left(12\right)$

The energy of remaining component does

$E_R=\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{\left|A_R\left(t\right)\right|}^2---\left(13\right)$

Gross energy does

$E=\underset{i=1}{\overset{8}{\mathrm{\Σ}}}{E}_i+E_R---\left(14\right)$

The energy ratio of each natural mode of vibration does

$B_i=\frac{E_i}{E}---\left(15\right)$

The energy ratio of remaining component does

$B_R=\frac{E_R}{E}---\left(16\right)$

Actual flow pattern is discerned: choose in addition 140 groups with the corresponding pressure difference data of flow pattern to be identified (40 groups of bubble flows, 60 groups of intermittent flows, 20 groups of laminar flows, 20 groups of annular flows) as test sample book; Liken energy to train Elman neural network respectively, can judge corresponding flow pattern by the output of network for the characteristic quantity input with corresponding 8 modal components and 1 residual components.

Claims (3)

1. the flow pattern recognition device when a pipeline mesohigh hydrodynamic pressure changes fast; Comprise incoming flow pipeline (1), fixed dam (2), incoming flow pipe joint (3), incoming flow joint packing (4), come flow sensor three-way connection (5), incoming flow soft washer (6), high pressure resistant quartzy transparent tube (7), go to flow soft washer (8), remove flow sensor three-way connection (9), go to flow joint packing (10), go to flow pipe joint (11), remove to flow pipeline (12), adjusting bolt (14), incoming-flow pressure sensor (15), incoming-flow pressure transmitter (16), the incoming flow sensor packing ring (17) of regulating sleeve (13), band through hole, remove flowing pressure sensor (18), remove flowing pressure transmitter (19), remove flow sensor packing ring (20), computing machine (21), light source (22), high-speed camera (23), it is characterized in that: incoming flow pipeline (1) welds together with incoming flow pipe joint (3); Incoming flow pipe joint (3) with come flow sensor three-way connection (5) to adopt to be threaded incoming flow pipe joint (3) and come to be equipped with between the flow sensor three-way connection (5) incoming flow joint packing (4); Come between flow sensor three-way connection (5) and the high pressure resistant quartzy transparent tube (7) incoming flow soft washer (6) to be housed; High pressure resistant quartzy transparent tube (7) and go to be equipped with between the flow sensor three-way connection (9) and flow soft washer (8) is come flow sensor three-way connection (5), is gone flow sensor three-way connection (9) and high pressure resistant quartzy transparent tube (7) to adopt clamped-in style to be connected; Go flow sensor three-way connection (9) to adopt and be threaded, remove flow sensor three-way connection (9) and go to flow to be equipped with between the pipe joint (11) and flow joint packing (10) with going to flow pipe joint (11); Go to flow pipe joint (11) and remove to flow pipeline (12) and weld together; Regulating sleeve (13) with go to flow pipe joint (11) and be close to, the internal thread of regulating sleeve (13) matches with the external thread of the adjusting bolt (14) of band through hole, is with the adjusting bolt (14) of through hole to be enclosed within and flows on the pipeline (12); Incoming flow pipeline (1), go to flow the hole that pipeline (12) passes fixed dam (2) two ends respectively, incoming flow pipe joint (3) is close to fixed dam (2), and the adjusting bolt (14) of band through hole is close to fixed dam (2); Incoming-flow pressure sensor (15) with come flow sensor three-way connection (5) to adopt to be threaded incoming-flow pressure sensor (15) and come to be equipped with between the flow sensor three-way connection (5) incoming flow sensor packing ring (17); Incoming-flow pressure transmitter (16) is connected with incoming-flow pressure sensor (15), and incoming-flow pressure transmitter (16) is connected with computing machine (21); Go flowing pressure sensor (18) to be threaded, remove flowing pressure sensor (18) and remove to be equipped with between the flow sensor three-way connection (9) flow sensor packing ring (20) with going flow sensor three-way connection (9) employing; Go flowing pressure transmitter (19) to be connected, go flowing pressure transmitter (19) to be connected with computing machine (21) with removing flowing pressure sensor (18); Light source (22) shines the fluid in the high pressure resistant quartzy transparent tube (7); High-speed camera (23) links together with computing machine (21), and high-speed camera (23) is taken the mobility status of fluid in the high pressure resistant quartzy transparent tube (7).

2. the flow pattern recognition device when pipeline mesohigh hydrodynamic pressure according to claim 1 changes fast is characterized in that: the screwing length through adjustment regulating sleeve (13) and the adjusting bolt (14) of being with through hole guarantee the sealing between flow sensor three-way connection (5) and the high pressure resistant quartzy transparent tube (7) and remove flow sensor three-way connection (9) and high pressure resistant quartzy transparent tube (7) between sealing.

3. the flow type identification method when a pipeline mesohigh hydrodynamic pressure changes fast is characterized in that, comprises the steps:

1) obtain data: incoming-flow pressure sensor (15) is measured pipeline inner fluid incoming-flow pressure; Go flowing pressure sensor (18) to go flowing pressure to measure, use the flow pattern of fluid in high-speed camera (23) the high pressure resistant quartzy transparent tube of record (7) simultaneously the pipeline inner fluid;

2) transmission data: the incoming-flow pressure that incoming-flow pressure transmitter (16) measures incoming-flow pressure sensor (15) is imported computing machine (21) and is preserved; Go flowing pressure transmitter (19) to go flowing pressure input computing machine (21) to what go that flowing pressure sensor (18) measures and preserve, high-speed camera (23) deposits the flow pattern of noting in computing machine (21);

3) data are carried out pre-service: the incoming-flow pressure of noting computing machine (21) and corresponding time go flowing pressure to subtract each other to obtain corresponding pressure difference data;

4) sorter is trained: the corresponding pressure difference data of selecting to photograph with high-speed camera (23) of typical flow pattern is carried out Hilbert-Huang transform; Obtain conversion spectrum; Calculate the energy ratio of each mode; And liken each mode energy to neural network into sample input Elman, and train, obtain the flow pattern criterion of identification;

5) actual flow pattern is discerned: the corresponding pressure difference data of flow pattern to be identified is carried out Hilbert-Huang transform; Obtain conversion spectrum; Calculate the energy ratio of each mode, and liken each mode energy to train neural network, obtain corresponding flow pattern for the characteristic quantity input.

Images