CN101900589B - Air-entrainment liquid flow measuring method based on mass flowmeter - Google Patents

Abstract

The invention discloses an air-entrainment liquid flow measuring method based on a mass flowmeter. The device comprises a metering duct, a pressure sensor, a mass flowmeter, an A/D conversion card and a computer, wherein pressure signals measured by the pressure sensor, instantaneous mass flow signals measured by the mass flowmeter, medium density signals and medium temperature signals are sent into a computer through the A/D conversion card. The measuring method comprises the following three basic steps: (1) feature extraction of pressure fluctuation signal; (2) training of BP neural network; and (3) flow measurement and error correction. The invention has the advantages of simple structure, convenient installation, high parameter detection precision, favorable real-time performance, high reliability, easy realization and the like, and is suitable for measuring air-entrainment liquids.

Description

Air-entrainment liquid flow measuring method based on mass flowmeter

Technical field

The invention belongs to the fluid measurement technical field, be specifically related to a kind of air-entrainment liquid flow measuring method based on mass flowmeter and neural network.

Background technology

Gas enclosure liquid extensively is present in departments such as oil, chemical industry, metallurgy, the energy, power, light industry, and the ubiquity of its application and importance are impelled developing rapidly of this area research work.Gas enclosure liquid is meant the pipeline inner fluid because the conversion between design or the liquids and gases that cause of flow operating mode or comprise the medium of two states.The flow of gas enclosure liquid is the parameter that has important science and engineering significance in the gas-liquid two-phase streaming system; Metering, control, reliability and efficient etc. to the biphase gas and liquid flow application system all have significant impact; For a long time, be the research direction in two-phase flow field always.

Chinese scholars has been carried out a large amount of research to air-entrainment liquid flow measuring method, and wherein main method has mass flowmeter method, single-phase flowmeter method, measurement of correlation method etc.

Mass flowmeter is also claimed driect type mass flowmeter, is characterized in directly measuring the mass rate of fluid and the density that need not to measure fluid, thereby is attempted being used for directly measuring the mass rate of gas enclosure liquid by many researchers.The mass flowmeter method is a kind of mature methods, and measuring accuracy is higher, combines with branch phase content measuring equipments such as mass rate void fraction (claiming mass dryness fraction again) or voidages, can satisfy higher measurement requirement.This method is used to measure the low air void mass rate and has obtained certain progress.

Single-phase flowmeter method is that the single-phase flow flow measurement instrument is applied to the method for air-entrainment liquid flow in measuring, because these single-phase flowmeters comparative maturity all in theory research and practical application makes this method in commercial Application, accepted more easily.According to the difference of single-phase flow meter, this method can be divided into two single-phase flowmeter combined methods, single-phase flowmeter and densitometer combined methods and fluctuation signal method of characteristic etc.

The measurement of correlation method is to be the two-phase flow measurement method that the basis constitutes with the correlation technique.This method can be used for measuring the flow of any fluid system in theory, and the scope of measurement flow rate is very wide, so the correlation flowmeters method provides a kind of strong technological means for solving two-phase flowmeter.This technological advantage is to adopt different sensor to obtain fluid flow signal, is equipped with other measurement means again, can constitute various flow measuring systems, realizes non-contact measurement.But related flow measurement technology still exists some problem needs further to inquire at present, and for example the physical significance of relevant speed is still unclear, and the cross correlation function peak value is difficult confirms that correlation flowmeters are demarcated and still had certain difficulty etc.

Coriolis mass flowmeter is arrived petroleum and petrochemical industry and other commercial production field by promotion and application widely because of its high measuring accuracy, reliability and low-down maintenance cost.Coriolis mass flowmeter can accurately be measured the single-phase flow flow, and precision can reach 0.2%.Yet, in the plurality of applications field, all exist a large amount of gas enclosure liquid (or being called biphase gas and liquid flow).The researcher thinks that higher gas contains under the rate situation at present; The pipe vibrations are unstable; Be difficult to obtain stable signal; Measuring error is bigger when measuring the gas enclosure liquid medium in view of Coriolis mass flowmeter, and measurement result can not well satisfy the requirement in petroleum and petrochemical industry and other commercial production field, and Coriolis mass flowmeter seldom is considered and is applied to the measurement of gas enclosure liquid medium flow.

The flow of gas enclosure liquid is inseparable with its voidage, can utilize voidage to set up gas enclosure liquid quality flow measurement error correction model.Because the difference of characteristics such as gas phase and density of liquid phase, viscosity in the biphase gas and liquid flow, phase interface motion and the alternate reasons such as interaction of gas-liquid are accompanied by the pressure surge phenomenon.The wave characteristic of pressure signal can be used for estimating the voidage of gas enclosure liquid.

Summary of the invention

The objective of the invention is to overcome the above-mentioned deficiency of prior art, a kind of air-entrainment liquid flow measuring method based on mass flowmeter is provided.Method parameter accuracy of detection provided by the invention is high, and real-time is good, and is reliable, is easy to realize.Be applicable to the measurement of air containing fluid.

For this reason, the present invention adopts following technical scheme:

A kind of air-entrainment liquid flow measurement mechanism based on mass flowmeter; Comprise metering pipeline (1), pressure transducer (2), Coriolis mass flowmeter (3), A/D transition card (4), computing machine (5); On metering pipeline (1), be provided with pressure transducer (2), Coriolis mass flowmeter (3) successively; A/D transition card (4) links to each other with pressure transducer (2), Coriolis mass flowmeter (3), and computing machine (5) links to each other with A/D transition card (4).

The present invention is based on the air-entrainment liquid flow measuring method of mass flowmeter and neural network, include following basic step:

(1) Pressure Fluctuation Signal feature extraction: analyze the time-frequency characteristic of Pressure Fluctuation Signal, the spectrum value on the fractal dimension of extraction Pressure Fluctuation Signal, root mean square, the coefficient of skew, the different frequency range forms a vector, the constitutive characteristic vector;

(2) BP neural metwork training: design the BP neural network as requested,, confirm neural network weight and threshold value with the proper vector training BP neural network that step 1) obtains;

(3) void fraction calculates: the pressure surge data input step 1 that collects in real time), step 1) is extracted the proper vector of pressure surge data.Then, with the neural network that the proper vector input of extracting trains, obtain the void fraction φ of the corresponding gas enclosure liquid of these pressure surge data _g

(4) flow measurement error correction: according to ρ _Pre=a ₁φ _g+ b ₁Calculate the hybrid density of gas enclosure liquid, in the formula, ρ _PreBe the hybrid density of gas enclosure liquid, a ₁, b ₁Be hybrid density calculating formula coefficient, confirm, be stored in the computing machine φ through the experimental data off-line _gBe the void fraction of gas enclosure liquid, by above-mentioned steps 3) confirm.According to m _Pre=a ₂φ _g+ b ₂Calculate the mixing quality flow of gas enclosure liquid, in the formula, m _PreBe the mixing quality flow of gas enclosure liquid, a ₂, b ₂Be mixing quality flowmeter formula coefficient, confirm, be stored in the computing machine through the experimental data off-line.If | ρ _Mea-ρ _Pre|≤U _ρAnd | m _Mea-m _Pre|≤U _m, so

In the formula, ρ _MeaBe the pseudodensity that Coriolis mass flowmeter records, U _ρBe the uncertainty of density measure, be stored in the computing machine m _MeaBe the apparent mass flow that Coriolis mass flowmeter records, U _mBe the uncertainty of mass flow measurement, be stored in the computing machine m _LiquidBe the liquid quality flow after proofreading and correct, ρ _lBe fluid density, a ₃, b ₃For liquid quality flow correction coefficient, be stored in the computing machine.If | ρ _Mea-ρ _Pre|≤U _ρAnd | m _Mea-m _Pre|＞U _m, so $m_{\mathrm{Mea}}^{*}=a_4\&CenterDot;{\mathrm{\&rho;}}_{\mathrm{Mea}}+{\mathrm{<figure-callout\; id="4"\; label="b"\; filenames="HSA00000098972300041.png"\; state="\{\{state\}\}">b</figure-callout>}}_4$ And

In the formula,

Be the reliable quality flow value of estimating according to density measurement, a ₄, b ₄Be reliable quality flow estimator coefficient.If | ρ _Mea-ρ _Pre|＞U _ρAnd | m _Mea-m _Pre|≤U _m, so ${\mathrm{\&rho;}}_{\mathrm{Mea}}^{*}=\frac{1}{a_4}\&CenterDot;m_{\mathrm{Mea}}-\frac{b_4}{a_4}$ And $m_{\mathrm{Liquid}}=\frac{m_{\mathrm{Mea}}}{{\mathrm{<figure-callout\; id="3"\; label="b"\; filenames="HSA00000098972300041.png"\; state="\{\{state\}\}">b</figure-callout>}}_3+a_3\&CenterDot;\frac{{\mathrm{\&rho;}}_{\mathrm{Mea}}^{*}}{{\mathrm{\&rho;}}_l}},$ In the formula, Be the believable density value of estimating according to mass flow measurement.If | ρ _Mea-ρ _Pre|＞U _ρAnd | m _Mea-m _Pre|＞U _m, so In the formula,

Be the liquid quality flow after the last measurement point correction.

Above-mentioned steps 2) neural network weight and the threshold value that train in are stored in the computing machine.

Above-mentioned steps 2) select three layers BP neural network in, wherein the input layer number is 2～7, and the number of hidden nodes is 4～12, and output layer node number is 1.

Beneficial effect of the present invention and advantage are to need not to adopt high efficient gas and liquid separator to carry out gas-liquid separation, adopts pressure sensor and Coriolis mass flowmeter multiple measurement air-entrainment liquid flow.Utilize the time-frequency characteristic of Pressure Fluctuation Signal and the void fraction that neural net method is confirmed gas enclosure liquid; Utilize Coriolis mass flowmeter to measure instantaneous mass flow, hybrid density and temperature, utilize the void fraction that calculates that the measurement result of Coriolis mass flowmeter is proofreaied and correct.This apparatus structure is simple, and is easy for installation.This measuring method parameter detecting precision is high, and real-time is good, is easy to realize.Be applicable to the measurement of air containing fluid.

Description of drawings

Fig. 1 is the air-entrainment liquid flow measurement mechanism structural representation based on mass flowmeter and neural network;

Fig. 2 is a BP neural network structure synoptic diagram;

Fig. 3 is a software program flow chart of the present invention;

Fig. 4 is void fraction and gas enclosure liquid mixing density relationship figure;

Fig. 5 is void fraction and gas enclosure liquid mixing mass rate graph of a relation;

The graph of a relation that falls with flow falls in Fig. 6 for density;

Fig. 7 is believable density measurement and reliable quality flow measurements graph of a relation;

Fig. 8 is an air-entrainment liquid flow measuring error correcting result.

Embodiment

Air-entrainment liquid flow measurement mechanism based on mass flowmeter has metering pipeline (1); On metering pipeline (1), be provided with pressure transducer (2), Coriolis mass flowmeter (3) successively; A/D transition card (4) links to each other with pressure transducer (2), Coriolis mass flowmeter (3), and computing machine (5) links to each other with A/D transition card (4).

Present embodiment to volumetric flow of gas from 0.29m ³/ h is to 0.905m ³The air-entrainment liquid flow measuring error of/h is proofreaied and correct.

1) Pressure Fluctuation Signal feature extraction

The first step of the present invention is for extracting the time-frequency characteristic of Pressure Fluctuation Signal, the constitutive characteristic vector.The proper vector of Pressure Fluctuation Signal is made up of the spectrum value on the fractal dimension of pressure signal, root mean square, the coefficient of skew, the different frequency range; Be complicated nonlinear relationship between the voidage of this proper vector and gas enclosure liquid, this nonlinear relationship is confirmed by neural network.The proper vector that is input as Pressure Fluctuation Signal of neural network, neural network is output as the void fraction of gas enclosure liquid.Proper vector by Pressure Fluctuation Signal is extracted is as follows.Each line data is a gas enclosure liquid space rate value characteristic of correspondence vector.

The proper vector of Pressure Fluctuation Signal

1.4641 3.0498 1.1858 ?0.94 19.05 39.97

1.2823 3.0414 1.3620 ?2.46 22.13 43.44

1.4906 3.2716 0.2807 ?0.21 24.84 33.95

1.4895 3.7372 -0.0312 0.29 21.34 35.49

The gas enclosure liquid space rate value vector that the Pressure Fluctuation Signal proper vector is corresponding

0.028 0.041 0.056 0.073

2) BP neural metwork training

Second step of the present invention is the weights and the threshold value of training BP neural network.Because of three layers of BP neural network have the learning ability of approaching any nonlinear function, so the present invention adopts three layers of BP neural network.The input layer number of three layers of BP neural network can be 2～7, and the number of hidden nodes can be 4～12, and output layer node number is 1.In the present embodiment, the input number of nodes of three layers of BP neural network is 6, and the number of hidden nodes is 7, and output layer node number is 1.

Before carrying out gas enclosure liquid space rate estimation, at first use last one and go on foot the Pressure Fluctuation Signal proper vector and the gas enclosure liquid space rate value vector that obtain the BP neural network that designs is trained, confirm the weights and the threshold value of neural network.The parameter of neural network training is following: learning rate is 0.6, and the inertia scale factor is 0.9, and 1000 step of network iteration back obtains satisfied convergence effect.

3) void fraction calculates

The BP neural network is through step 2) train after, just can be used for the calculating of gas enclosure liquid space rate.The data that the gas enclosure fluid pressure fluctuation signal that collects is carried out behind the time frequency analysis are input to the BP neural network that trains, and the output result of neural network is exactly the voidage value φ of this gas enclosure liquid _g

In the present embodiment, measure the gas enclosure fluid pressure fluctuation signal under two kinds of mobility status, behind time frequency analysis, obtain two proper vectors

V1＝[1.3876 3.4301 -0.6309 0.84 29.28 41.69]′

V2＝[1.4433 3.5710 0.4102 ?0.35 24.70 35.82]′

With above-mentioned two proper vector input steps 2) in the BP network that trains, the neural network that proper vector V1 is corresponding is output as 0.048, and the neural network that proper vector V2 is corresponding is output as 0.065, thereby has obtained the voidage value φ under these two kinds of mobility status _gBe respectively 0.048 and 0.065.

4) flow measurement error correction

Gas enclosure liquid void fraction and its hybrid density relational expression are following:

ρ _pre＝a ₁·φ _g+b ₁ (1)

In the present embodiment, a ₁=-0.6181, b ₁=1.0082, φ _gNeural network weight and the threshold value utilizing Pressure Fluctuation Signal and be stored in computing machine are confirmed ρ _PreBe gas enclosure liquid mixing density according to gas enclosure liquid space rate value prediction.

Gas enclosure liquid void fraction and its mixing quality discharge relation formula are following:

m _pre＝a ₂·φ _g+b ₂ (2)

In the present embodiment, a ₂=-161.08, b ₂=174.24, m _PreBe gas enclosure liquid mixing mass rate according to gas enclosure liquid space rate value prediction.

In gas enclosure liquid (or being called biphase gas and liquid flow); Owing to exist a shape and be distributed in phase interface all variable at random in time and the space; And alternately in fact have a relative velocity of can not ignore, cause two-phase flow have than single-phase flow complicated the flow characteristics of Duoing, measuring accuracy obviously descended when Coriolis mass flowmeter was measured this gas enclosure liquid; But the pseudodensity that Coriolis mass flowmeter records is relevant with the mass flow measurement error.Therefore, can utilize this pseudodensity correcting fluid physique measuring mass flow value.

Definition:

$\mathrm{\&Delta;\&rho;}=\frac{{\mathrm{\&rho;}}_l-{\mathrm{\&rho;}}_{\mathrm{mea}}}{{\mathrm{\&rho;}}_l}\&times;100\%---\left(3\right)$

$\mathrm{\&Delta;m}=\frac{m_{\mathrm{liquid}}-m_{\mathrm{mea}}}{m_{\mathrm{liquid}}}\&times;100\%---\left(4\right)$

Wherein, Δ ρ representes that ρ falls in density _lAnd ρ _MeaRepresent the density of testing liquid and the pseudodensity that Coriolis mass flowmeter records respectively, Δ m representes that flow falls, m _LiquidAnd m _MeaThe apparent mass flow that the Coriolis mass flow instrumentation got before the mass rate of testing liquid and the correction after expression was proofreaied and correct respectively.

If the measurement of Coriolis mass flowmeter is more stable; So; Gas enclosure liquid mixing density according to voidage prediction should be more or less the same with the pseudodensity that Coriolis mass flowmeter records, and the gas enclosure liquid quality flow of prediction and the apparent mass flow that records also should be more or less the same.That is to say, if satisfy

|ρ _mea-ρ _pre|≤U _ρ (5)

|m _mea-m _pre|≤U _m (6)

So, according to relational expression (3), relational expression (4) and great deal of experiment data, have

$m_{\mathrm{liquid}}=\frac{{\mathrm{<figure-callout\; id="3"\; label="m\; mea\; b"\; filenames="HSA00000098972300041.png"\; state="\{\{state\}\}">m</figure-callout>}}_{\mathrm{mea}}}{b_{}}---\left(7\right)$

In the present embodiment, a ₃=1.5017, b ₃=0.6427, U _ρ=0.0011, U _m=0.9683, m _MeaAnd ρ _MeaRecord by Coriolis mass flowmeter.

If

|ρ _mea-ρ _pre|≤U _ρ (8)

|m _mea-m _pre|＞U _m (9)

So,

$m_{\mathrm{mea}}^{*}=a_4\&CenterDot;{\mathrm{\&rho;}}_{\mathrm{mea}}+b_4---\left(10\right)$

$m_{\mathrm{liquid}}=\frac{m_{\mathrm{mea}}^{*}}{{\mathrm{<figure-callout\; id="3"\; label="b"\; filenames="HSA00000098972300041.png"\; state="\{\{state\}\}">b</figure-callout>}}_3+a_3\&CenterDot;\frac{{\mathrm{\&rho;}}_{\mathrm{mea}}}{{\mathrm{\&rho;}}_l}}---\left(11\right)$

In the present embodiment, a ₄=259.91, b ₄=-87.83.

If

|ρ _mea-ρ _pre|＞U _ρ (12)

|m _mea-m _pre|≤U _m (13)

So,

${\mathrm{\&rho;}}_{\mathrm{mea}}^{*}=\frac{1}{a_4}\&CenterDot;m_{\mathrm{mea}}-\frac{b_4}{a_4}---\left(14\right)$

$m_{\mathrm{liquid}}=\frac{{\mathrm{<figure-callout\; id="3"\; label="m\; mea\; b"\; filenames="HSA00000098972300041.png"\; state="\{\{state\}\}">m</figure-callout>}}_{\mathrm{mea}}}{b_{}}---\left(15\right)$

If

|ρ _mea-ρ _pre|＞U _ρ (16)

|m _mea-m _pre|＞U _m (17)

So

$m_{\mathrm{liquid}}=m_{\mathrm{liquid}}^{*}---\left(18\right)$

In the formula,

is the liquid quality flow after last measurement point is proofreaied and correct.

Claims (4)

1. the air-entrainment liquid flow measuring method based on mass flowmeter is characterized in that, comprises the following steps:

(1) Pressure Fluctuation Signal feature extraction: analyze the time-frequency characteristic of Pressure Fluctuation Signal, the spectrum value on the fractal dimension of extraction Pressure Fluctuation Signal, root mean square, the coefficient of skew, the different frequency range forms a vector, the constitutive characteristic vector;

(2) BP neural metwork training: design the BP neural network as requested,, confirm neural network weight and threshold value with the proper vector training BP neural network that step 1) obtains;

(3) void fraction calculates: the pressure surge data input step 1 that collects in real time); Step 1) is extracted the proper vector of pressure surge data; Then, with the neural network that the proper vector input of extracting trains, obtain the void fraction φ of the corresponding gas enclosure liquid of these pressure surge data _g

(4) flow measurement error correction: according to ρ _Pre=a ₁φ _g+ b ₁Calculate the hybrid density of gas enclosure liquid, in the formula, ρ _PreBe the hybrid density of gas enclosure liquid, a ₁, b ₁Be hybrid density calculating formula coefficient, confirm, be stored in the computing machine φ through the experimental data off-line _gBe the void fraction of gas enclosure liquid, by above-mentioned steps 3) confirm, according to m _Pre=a ₂φ _g+ b ₂Calculate the mixing quality flow of gas enclosure liquid, in the formula, m _PreBe the mixing quality flow of gas enclosure liquid, a ₂, b ₂Be mixing quality flowmeter formula coefficient, confirm, be stored in the computing machine through the experimental data off-line, if | ρ _Mea-ρ _Pre|≤U _ρAnd | m _Mea-m _Pre|≤U _m, so

In the formula, ρ _MeaBe the pseudodensity that Coriolis mass flowmeter records, U _ρBe the uncertainty of density measure, be stored in the computing machine m _MeaBe the apparent mass flow that Coriolis mass flowmeter records, U _mBe the uncertainty of mass flow measurement, be stored in the computing machine m _LiquidBe the liquid quality flow after proofreading and correct, ρ _lBe fluid density, a ₃, b ₃For liquid quality flow correction coefficient, be stored in the computing machine, if | ρ _Mea-ρ _Pre|≤U _ρAnd | m _Mea-m _Pre|＞U _m, so

And

In the formula, Be the reliable quality flow value of estimating according to density measurement, a ₄, b ₄Be reliable quality flow estimator coefficient, if | ρ _Mea-ρ _Pre|＞U _ρAnd | m _Mea-m _Pre|≤U _m, so

And

In the formula,

Be the believable density value of estimating according to mass flow measurement, if | ρ _Mea-ρ _Pre|＞U _ρAnd | m _Mea-m _Pre|＞U _m, so

In the formula,

Be the liquid quality flow after the last measurement point correction.

2. a kind of air-entrainment liquid flow measuring method based on mass flowmeter according to claim 1 is characterized in that above-mentioned steps 2) in the neural network weight and the threshold value that train be stored in the computing machine.

3. a kind of air-entrainment liquid flow measuring method according to claim 1 based on mass flowmeter; It is characterized in that above-mentioned steps 2) the middle BP neural network of selecting three layers; Wherein the input layer number is 2～7, and the number of hidden nodes is 4～12, and output layer node number is 1.

4. a kind of air-entrainment liquid flow measuring method based on mass flowmeter according to claim 1 is characterized in that above-mentioned steps 4) in flow measurement error correction formula be stored in the computing machine.

Images