CN114510974B - Intelligent recognition method for gas-liquid two-phase flow pattern in porous medium - Google Patents
Intelligent recognition method for gas-liquid two-phase flow pattern in porous medium Download PDFInfo
- Publication number
- CN114510974B CN114510974B CN202210102131.5A CN202210102131A CN114510974B CN 114510974 B CN114510974 B CN 114510974B CN 202210102131 A CN202210102131 A CN 202210102131A CN 114510974 B CN114510974 B CN 114510974B
- Authority
- CN
- China
- Prior art keywords
- differential pressure
- pressure signal
- flow pattern
- power
- feature vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000005514 two-phase flow Effects 0.000 title claims abstract description 41
- 239000007788 liquid Substances 0.000 title claims abstract description 31
- 238000000034 method Methods 0.000 title claims abstract description 25
- 239000013598 vector Substances 0.000 claims abstract description 49
- 238000005259 measurement Methods 0.000 claims abstract description 7
- 238000012706 support-vector machine Methods 0.000 claims description 32
- 230000003595 spectral effect Effects 0.000 claims description 26
- 230000006870 function Effects 0.000 claims description 23
- 238000005516 engineering process Methods 0.000 claims description 16
- 238000009826 distribution Methods 0.000 claims description 15
- 238000003909 pattern recognition Methods 0.000 claims description 13
- 238000001228 spectrum Methods 0.000 claims description 11
- 238000012549 training Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 7
- 230000001186 cumulative effect Effects 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 3
- 230000008092 positive effect Effects 0.000 abstract description 2
- 230000001737 promoting effect Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 10
- 239000012530 fluid Substances 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000012567 pattern recognition method Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000009776 industrial production Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 239000007791 liquid phase Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N7/00—Analysing materials by measuring the pressure or volume of a gas or vapour
- G01N7/10—Analysing materials by measuring the pressure or volume of a gas or vapour by allowing diffusion of components through a porous wall and measuring a pressure or volume difference
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Measuring Volume Flow (AREA)
- Aerodynamic Tests, Hydrodynamic Tests, Wind Tunnels, And Water Tanks (AREA)
Abstract
The invention discloses an intelligent recognition method for a gas-liquid two-phase flow pattern in a porous medium, which comprises the following steps: 1. acquiring differential pressure signal data; 2. extracting the characteristics of the differential pressure signals obtained by measurement, and generating characteristic parameters; 3. constructing a feature vector; 4. judging a corresponding flow pattern at the moment by utilizing an SVM intelligent recognition module according to the input feature vector and a preset recognition model; 5. an output flow pattern; the method has the advantages of low cost, easy realization, high accuracy and good real-time performance; the method realizes the rapid identification of the gas-liquid two-phase flow pattern of the porous medium, and has positive effects of perfecting the gas-liquid two-phase flow pattern identification theory and promoting the gas-liquid two-phase flow detection.
Description
Technical Field
The invention belongs to the technical field of fluid measurement, applies a feature extraction technology and a machine learning technology to gas-liquid two-phase flow pattern recognition, and particularly relates to an in-tube porous medium gas-liquid two-phase flow pattern recognition method based on a feature extraction and SVM intelligent recognition module.
Background
The gas-liquid two-phase flow phenomenon is a very common phenomenon in daily life and industrial production. The flow pattern characterizes the distribution of two fluids in the two-phase flow, and is one of the important parameters of the two-phase flow. The difference in flow patterns can result in significant differences between the heat and mass transfer properties, flow characteristics, and other flow heat transfer parameters of the fluid. In industrial production, the flow pattern can be identified rapidly and effectively, the product quality can be improved, the potential safety hazard is reduced, and the energy is saved effectively. Therefore, the method has important significance for researching the two-phase flow pattern recognition technology.
At present, some domestic scholars have developed related researches on two-phase flow pattern recognition technology, but the two-phase flow pattern recognition technology is mainly aimed at the relatively simple two-phase flow of a flow channel. For example, the application publication number is CN 107402116A, and the name is a gas-liquid two-phase flow pattern recognition method and a detection device. The invention develops a flow pattern identification method and a measurement device thereof which are specially aimed at the turbulence of gas-liquid two-phase flow. For example, the application publication number is CN 112113742A, and the gas-liquid two-phase flow pattern online identification method is based on GRU neural network. The method utilizes a throttling device and a deep learning development board to identify the two-phase flow pattern in the smooth pipeline. For example, the application publication number is CN 104330336A, and the name is a gas-liquid two-phase flow pattern recognition method based on ICA and SVM. The method is based on ICA technology and SVM technology, and is used for identifying two-phase flow patterns in a venturi tube. The invention is researched aiming at smooth pipelines with simpler internal structures, and because the pore channels of the porous medium have the characteristics of flexibility and randomness, the flow characteristics of the fluid in the porous medium are very complex, and have great difference with the flow of the light pipe. Therefore, the two-phase flow pattern recognition technology based on smooth pipeline development cannot be directly applied to the field of porous medium two-phase flow pattern recognition, and an effective technical method for realizing the flow pattern recognition of the gas-liquid two-phase flow pattern in the porous medium is necessary.
The gas-liquid phases of different flow patterns pass through the pressure transmitter in their own flow patterns, and this difference necessarily results in a difference in the differential pressure signal between the time domain and the frequency domain. The probability density function can reflect the distribution range, fluctuation intensity and other time domain characteristics of the signals, and the power spectrum density is widely used for calculating the strength of the natural frequency of the signals, so that the frequency domain characteristics of the signals can be reflected well. Thus, the probability density function and the power spectral density may be the basis for differential pressure signal feature extraction.
The Support Vector Machine (SVM) technology is a machine learning algorithm based on a statistical learning theory, and has strong classification recognition capability. Based on SVM technology, experimental data and computer technology, the flow pattern recognition module with continuous optimization potential can be trained and developed. The module has the basic function of rapidly identifying the flow pattern of the porous medium gas-liquid two-phase flow, and can enlarge the identification range and improve the identification precision by machine learning in a mode of supplementing experimental data.
Disclosure of Invention
The invention aims to provide an intelligent recognition method for a gas-liquid two-phase flow pattern in a porous medium, so as to make up for the defects of the related technology, realize rapid recognition of the gas-liquid two-phase flow pattern of the porous medium, and have positive effects of perfecting a gas-liquid two-phase flow pattern recognition theory and promoting gas-liquid two-phase flow detection.
In order to achieve the above purpose, the invention adopts the following technical scheme:
a gas-liquid two-phase flow pattern intelligent identification method in a porous medium comprises the following steps:
step 1: acquiring differential pressure signal data;
obtaining an inlet-outlet pressure difference signal of the porous medium pipeline through a pressure transmitter arranged at an inlet and an outlet of the porous medium pipeline;
step 2: extracting features of the differential pressure signals obtained through measurement, and generating feature vectors; the specific feature extraction steps are as follows:
step 2.1: calculating time domain characteristic parameters of the differential pressure signal, and obtaining probability density function f of the differential pressure signal firstly x (x) The method comprises the steps of carrying out a first treatment on the surface of the Then 2 required domain feature parameters are calculated: standard deviation sigma of differential pressure signal, deviation S of probability density function of differential pressure signal k ：
Step 2.1.1: calculating to obtain a probability density function of the differential pressure signal; for a set of time-varying differential pressure signals x i Let its cumulative distribution function be F x (x) If there is a measurable function f x (x) The method meets the following conditions:
then the differential pressure signal x i Is a continuous random variable, and f x (x) Is a differential pressure signal x i Probability density functions of (2);
step 2.1.2: the standard deviation sigma of the differential pressure signal is calculated according to the following formula:
wherein x is i Representing the original differential pressure signal,representing the average value of the original differential pressure signal in the whole time interval, wherein n is the number of data signal points;
step 2.1.3: calculating the bias S of the probability density function of the differential pressure signal according to the following formula k ：
Wherein mu 3 Is x i The third-order center distance of the device is calculated by the following formula:
step 2.2: calculating frequency domain characteristic parameters of the differential pressure signals; the power spectral density P of the signal is calculated first f And on the basis of the above, the required 3 characteristic parameters are obtained: maximum power corresponding to frequency f max Standard deviation sigma of power spectral density P Power efficient distribution range R P ；
Step 2.2.1: calculating to obtain the power spectrum density P of the differential pressure signal f
Given a time-varying differential pressure signal x (t), the power spectral density P of the signal f Is defined as:
wherein the method comprises the steps ofIs the square norm of the fourier transform of x (t) calculated using the fast fourier transform;
step 2.2.2: calculating to obtain frequency domain characteristic parameter of differential pressure signal, maximum power corresponding to frequency f max ：
The abscissa corresponding to the maximum value in the power spectrum density of the differential pressure signal is the maximum power corresponding to the frequency f max ；
Step 2.2.3: calculating standard deviation sigma of power spectrum density P ：
Wherein P is i Representing the corresponding power spectral density value for each frequency bin,representing a weighted average of the power spectral densities of all frequency bins;
step 2.2.4: calculating the effective distribution range R of power P
Solving the total power of the signals; and summing the power from 0Hz, and when the summed power reaches 99% of the total power, the frequency value represented by the corresponding abscissa is the effective power distribution range R P ；
Step 3: constructing a feature vector;
the step 2 is used for calculating 5 characteristic parameters of the differential pressure signal, and the 5 characteristic parameters are combined into a one-dimensional array [ sigma S ] with the length of 5 k f max σ P R P ]And normalizing the array:
S＝σ+S k +f max +σ P +R P
wherein V represents any one of the feature vectors;
after normalization treatment, for the sample which needs to be identified and judged to be in a flow pattern, the obtained vector is the characteristic vector, and the vector is directly input into an SVM intelligent identification module to be in flow pattern judgment; for the feature vector as the training sample, an element is added at the end of the array, the element is used as a label to indicate the attribute of the feature vector, the element 1 represents the bubble flow, the element 2 represents the bullet flow, and the element 3 represents the annular flow; the feature vector required by the SVM intelligent recognition module can be obtained through the processing of the step 3;
step 4: judging a corresponding flow pattern at the moment by utilizing an SVM intelligent recognition module according to the input feature vector and a preset recognition model;
the preset recognition model in the SVM intelligent recognition module is a model which is obtained based on a large amount of experimental data and Support Vector Machine (SVM) technology training; when a feature vector is input, the recognition model firstly reads each group of data contained in the feature vector one by one, then calculates the distance between the feature vector and the support vector in the recognition model according to the support vector machine principle, and then judges the flow pattern corresponding to the feature vector according to the distance;
step 5: output flow pattern: and outputting the flow pattern recognition result of the gas-liquid two-phase flow in the porous medium pipeline through a screen or an electric signal.
The invention has the beneficial effects that: the intelligent recognition method for the gas-liquid two-phase flow patterns in the porous medium is provided, and pressure transmitters arranged at the inlet and outlet of the porous medium channel are used for collecting inlet and outlet pressure difference signals, so that the influence of signal collection on the flow in the pipe can be reduced to the greatest extent. Through the feature extraction technology, the time domain and frequency domain features of the differential pressure signals are extracted creatively, and the comprehensiveness of the features is guaranteed. And then the flow pattern corresponding to the differential pressure signal is rapidly and accurately judged according to the characteristic vector by the trained SVM intelligent recognition module. The system has the advantages of low cost, easy realization, high accuracy and good real-time performance.
Drawings
Fig. 1 is a schematic diagram of a porous medium gas-liquid two-phase flow pattern intelligent identification system provided by an embodiment of the invention.
Fig. 2 is a schematic diagram of differential pressure signals of different flow patterns of a porous medium gas-liquid two-phase flow measured by the invention, wherein (a) is a schematic diagram of differential pressure signals of bubble flow, (b) is a schematic diagram of differential pressure signals of elastic flow, and (c) is a schematic diagram of differential pressure signals of annular flow.
FIG. 3 is a schematic diagram of probability density functions corresponding to different flow type differential pressure signals obtained through calculation.
Fig. 4 is a standard deviation σ distribution of the differential pressure signal under experimental conditions.
FIG. 5 is a graph of the bias S of probability density curves under experimental conditions k Distribution.
FIG. 6 is a schematic diagram of a power spectral density function corresponding to a differential pressure signal of different flow patterns. Wherein (a) is a power spectral density diagram of the bubble flow, (b) is a power spectral density diagram of the bullet flow, and (c) is a power spectral density diagram of the annular flow.
FIG. 7 is a graph showing the maximum power versus frequency f under experimental conditions max 。
FIG. 8 is the standard deviation σ of the power spectral density under experimental conditions P 。
FIG. 9 is a graph of the effective power distribution range R under experimental conditions P 。
Detailed Description
The invention will be further described with reference to specific embodiments.
The invention provides a flow pattern intelligent identification method of gas-liquid two-phase flow in a porous medium, which comprises the steps of firstly measuring pressure difference signals of an inlet and an outlet of a porous medium pipeline through a signal acquisition module, then processing the measured pressure difference signals in a characteristic extraction module by using probability density functions (probability density function, PDF) and power spectrum densities (power spectral density, PSD), and quantifying time domain characteristics and frequency domain characteristics of the pressure difference signals through calculation to obtain characteristic parameters reflecting the time domain characteristics and the frequency domain characteristics of the signals. And constructing a feature vector containing the time-frequency features of the signals by utilizing the feature parameters. The feature vectors are then input into an SVM intelligent recognition module developed based on support vector machine (Support Vector Machine, SVM) technology and computer technology. The module is provided with a preset recognition model, so that the corresponding flow pattern can be rapidly and accurately recognized according to the input feature vector, as shown in fig. 1.
The method comprises the following steps:
step 1: obtaining measurement data
The measurement data comprise pressure difference signals of different flow patterns of the gas-liquid two-phase flow of the porous medium. In this example, the flow patterns of the gas-liquid two-phase flow include bubble flow, bullet flow, and annular flow. Specifically, the differential pressure signal is obtained by a pressure transmitter 1 and a pressure transmitter 2 which are arranged at the inlet and outlet positions of the experimental section, and the pressure transmitter is a Rossment-3051 pressure transmitter (the precision is 0.04%). The sampling frequency was 400Hz.
By the above measuring device, the present example measured the flow pattern of the gas-liquid two-phase flow in the porous medium formed by stacking glass particles having a diameter of 3 mm. Wherein the flow rate of water was 0.29 mm.s, respectively -1 、0.59mm·s -1 、1.17mm·s -1 . For each flow rate of water, the gas flow rate is gradually increased, and the gas flow rate ranges from 0.005 m.s to 0.44 m.s -1 . In this case, 270 sets of data were measured as samples. Including the bubble flow 120 group; a set of bullet flows 90; annular flow 60 sets. Fig. 2 shows waveforms of differential pressure signals of different flow patterns, which are measured in order of (a) bubble flow, (b) bullet flow, and (c) annular flow.
Step 2: and extracting characteristic parameters of the differential pressure signal.
First, calculating time domain characteristic parameters of the differential pressure signal.
Step 2.1: and calculating to obtain a probability density function of the differential pressure signal. For a set of time-varying differential pressure signals x i Let its cumulative distribution function be F x (x) If there is a measurable function f x (x) The method meets the following conditions:
then x i Is a continuous random variable, and f x (x) Is a function of its probability density. FIG. 3 shows probability density function curves for different flow type differential pressure signals.
Step 2.2: the standard deviation sigma of the differential pressure signal is calculated according to the following formula:
wherein x is i Representing the original differential pressure signal,and representing the average value of the original differential pressure signal in the whole time interval, wherein n is the number of data signal points.
Step 2.3: calculating the skewness S of the probability density curve of the differential pressure signal according to the following formula k ：
Wherein mu 3 Is x i The third-order center distance of the device is calculated by the following formula:
FIGS. 4 and 5 show the standard deviation σ of the differential pressure signal and the skewness S of the probability density curve under experimental conditions k Is a distribution of the (b).
Frequency domain characteristic parameters of the signal are then calculated.
Step 2.4: calculating to obtain the power spectrum density P of the differential pressure signal f 。
Given a time-varying differential pressure signal x (t), the power spectral density P of the signal f Is defined as:
wherein the method comprises the steps ofIs the square norm of the fourier transform of x (t) calculated using the fast fourier transform. Fig. 6 shows the power spectral density calculation results for the 3 flow patterns. Wherein (a) represents the power spectral density of the bubble flow, (b) represents the power spectral density of the bullet flow, and (c) represents the power spectral density of the annular flow.
Step 2.5: calculating to obtain frequency domain characteristic parameter of differential pressure signal, maximum power corresponding to frequency f max ：
The abscissa corresponding to the maximum value in the power spectrum density of the differential pressure signal is the maximum power corresponding to the frequency f max 。
Step 2.6: calculating standard deviation sigma of power spectrum density P ：
Wherein P is i Representing the corresponding power spectral density value for each frequency bin,representing a weighted average of the power spectral densities of all frequency bins;
step 2.7: calculating the effective distribution range R of power P
The power spectral densities of the signals are summed. Then starting from 0Hz, the power is accumulatedAnd when the summation power reaches 99% of the total power, the frequency value represented by the corresponding abscissa is the effective distribution range R of the power P 。
FIGS. 7, 8 and 9 show the maximum power corresponding to frequency f of the 3 flow pattern differential pressure signals under experimental conditions, respectively max Standard deviation sigma of power spectral density P And a power effective distribution range R P 。
Step 3: construction of feature vectors
Calculating 5 characteristic parameters (sigma, S) mentioned in step 2 corresponding to the differential pressure signal, wherein the 5 characteristic parameters comprise 3 time domain characteristic parameters k ) And 3 frequency domain characteristic parameters (f max 、σ P And R is P ). Combining the 5 characteristic parameters into a one-dimensional array [ sigma S ] with the length of 5 k f max σ P R P ]And normalizing the array to obtain the required feature vector. In particular, for a feature vector as a training sample, a label is added to surface the attribute of the feature vector (1 represents a bubble stream, 2 represents a bullet stream, and 3 represents a ring stream). For example, the feature vector of one bubble stream is:
table 1 gives examples of three flow pattern feature vectors.
TABLE 1
Normalizing the array:
S＝σ+S k +f max +σ P +R P
wherein V represents any one of the feature vectors.
Step 4: and taking the normalized feature vector as input, and carrying out flow pattern recognition by utilizing an SVM intelligent recognition module.
The preset recognition model in the SVM intelligent recognition module is a model which is obtained based on a large amount of experimental data and Support Vector Machine (SVM) technology training; when a feature vector is input, the recognition model firstly reads each group of data contained in the feature vector one by one, then calculates the distance between the feature vector and the support vector in the recognition model according to the support vector machine principle, and then judges the flow pattern corresponding to the feature vector according to the distance.
In this example, 270 sets of data were obtained as samples. Including the bubble flow 120 group; a set of bullet flows 90; annular flow 60 sets. 60% of the data (bubble flow 72 groups, elastic flow 54 groups and annular flow 36 groups) are selected as training samples, and are input into a support vector machine for training to obtain an SVM classification model. The remaining samples were used as test samples to detect the accuracy of the model. Table 2 shows the test results of the samples.
It can be seen that for the test data (bubble flow 48, bullet flow 36, annular flow 24) there were 1 set of bubble flows and 3 sets of bullet flows identified incorrectly, while the annular flows were all identified correctly. The overall recognition rate was 96.3%.
Step 5: output flow pattern: and outputting the flow pattern recognition result of the gas-liquid two-phase flow in the porous medium pipeline through a screen or an electric signal.
Claims (1)
1. The intelligent recognition method for the gas-liquid two-phase flow pattern in the porous medium is characterized by comprising the following steps of:
step 1: acquiring differential pressure signal data;
obtaining an inlet-outlet pressure difference signal of the porous medium pipeline through a pressure transmitter arranged at an inlet and an outlet of the porous medium pipeline;
step 2: extracting features of the differential pressure signals obtained through measurement, and generating feature vectors; the specific feature extraction steps are as follows:
step 2.1: calculating time domain characteristic parameters of the differential pressure signal, and obtaining probability density function f of the differential pressure signal firstly x (x) The method comprises the steps of carrying out a first treatment on the surface of the Then 2 required domain feature parameters are calculated: standard deviation sigma of differential pressure signal, deviation S of probability density function of differential pressure signal k ：
Step 2.1.1: calculating to obtain a probability density function of the differential pressure signal; for a set of time-varying differential pressure signals x i Let its cumulative distribution function be F x (x) If there is a measurable function f x (x) The method meets the following conditions:
then the differential pressure signal x i Is a continuous random variable, and f x (x) Is a differential pressure signal x i Probability density functions of (2);
step 2.1.2: the standard deviation sigma of the differential pressure signal is calculated according to the following formula:
wherein x is i Representing the original differential pressure signal,representing the average value of the original differential pressure signal in the whole time interval, wherein n is the number of data signal points;
step 2.1.3: calculating the bias S of the probability density function of the differential pressure signal according to the following formula k ：
Wherein mu 3 Is x i The third-order center distance of the device is calculated by the following formula:
step 2.2: calculating frequency domain characteristic parameters of the differential pressure signals; the power spectral density P of the signal is calculated first f And on the basis of the above, the required 3 characteristic parameters are obtained: maximum power corresponding to frequency f max Standard deviation sigma of power spectral density P Power efficient distribution range R P ；
Step 2.2.1: calculating to obtain the power spectrum density P of the differential pressure signal f
Given a time-varying differential pressure signal x (t), the power spectral density P of the signal f Is defined as:
wherein the method comprises the steps ofIs the square norm of the fourier transform of x (t) calculated using the fast fourier transform;
step 2.2.2: calculating to obtain frequency domain characteristic parameter of differential pressure signal, maximum power corresponding to frequency f max ：
The abscissa corresponding to the maximum value in the power spectrum density of the differential pressure signal is the maximum power corresponding to the frequency f max ；
Step 2.2.3: calculating standard deviation sigma of power spectrum density P ：
Wherein P is i Representing the corresponding power spectral density value for each frequency bin,representing a weighted average of the power spectral densities of all frequency bins;
step 2.2.4: calculating the effective distribution range R of power P
Solving the total power of the signals; and summing the power from 0Hz, and when the summed power reaches 99% of the total power, the frequency value represented by the corresponding abscissa is the effective power distribution range R P ；
Step 3: constructing a feature vector;
the step 2 is used for calculating 5 characteristic parameters of the differential pressure signal, and the 5 characteristic parameters are combined into a one-dimensional array [ sigma S ] with the length of 5 k f max σ P R P ]And normalizing the array:
S＝σ+S k +f max +σ P +R P
wherein V represents any one of the feature vectors;
after normalization treatment, for the sample which needs to be identified and judged to be in a flow pattern, the obtained vector is the characteristic vector, and the vector is directly input into an SVM intelligent identification module to be in flow pattern judgment; for the feature vector as the training sample, an element is added at the end of the array, the element is used as a label to indicate the attribute of the feature vector, the element 1 represents the bubble flow, the element 2 represents the bullet flow, and the element 3 represents the annular flow; the feature vector required by the SVM intelligent recognition module can be obtained through the processing of the step 3;
step 4: judging a corresponding flow pattern at the moment by utilizing an SVM intelligent recognition module according to the input feature vector and a preset recognition model;
the preset recognition model in the SVM intelligent recognition module is a model which is obtained based on a large amount of experimental data and Support Vector Machine (SVM) technology training; when a feature vector is input, the recognition model firstly reads each group of data contained in the feature vector one by one, then calculates the distance between the feature vector and the support vector in the recognition model according to the support vector machine principle, and then judges the flow pattern corresponding to the feature vector according to the distance;
step 5: output flow pattern: and outputting the flow pattern recognition result of the gas-liquid two-phase flow in the porous medium pipeline through a screen or an electric signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210102131.5A CN114510974B (en) | 2022-01-27 | 2022-01-27 | Intelligent recognition method for gas-liquid two-phase flow pattern in porous medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210102131.5A CN114510974B (en) | 2022-01-27 | 2022-01-27 | Intelligent recognition method for gas-liquid two-phase flow pattern in porous medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114510974A CN114510974A (en) | 2022-05-17 |
| CN114510974B true CN114510974B (en) | 2024-04-09 |
Family
ID=81549304
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210102131.5A Active CN114510974B (en) | 2022-01-27 | 2022-01-27 | Intelligent recognition method for gas-liquid two-phase flow pattern in porous medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114510974B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115165792B (en) * | 2022-06-24 | 2023-09-29 | 宝腾智能润滑技术(东莞)有限公司 | Method and device for detecting gas-liquid two-phase flow state in pipeline |
| CN117129054A (en) * | 2023-10-20 | 2023-11-28 | 中国核电工程有限公司 | Method, device and system for monitoring two-phase storage tank material pouring end point |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5545252A (en) * | 1995-03-01 | 1996-08-13 | The Perkin-Elmer Corporation | Flow regulation in gas chromatograph |
| CN1904580A (en) * | 2006-08-01 | 2007-01-31 | 东北电力大学 | Gas liquid two phase flow type identification method based on information coalescence and flow type signal collection device |
| CN104330336A (en) * | 2014-11-07 | 2015-02-04 | 中国石油大学(华东) | ICA and SVM-based gas-liquid two-phase flow pattern identification method |
| CN107991384A (en) * | 2017-12-21 | 2018-05-04 | 浙江启尔机电技术有限公司 | The detection device and method of flow pattern of gas-liquid two-phase flow in a kind of micro-pipe |
| WO2019062595A1 (en) * | 2017-09-30 | 2019-04-04 | 深圳市岩尚科技有限公司 | Method for quality evaluation of photoplethysmogram |
| CN111896616A (en) * | 2020-03-29 | 2020-11-06 | 中国石油大学(华东) | Gas-liquid two-phase flow pattern identification method based on acoustic emission-BP neural network |
| CN112507638A (en) * | 2020-12-18 | 2021-03-16 | 北京科技大学 | Deep condensate gas reservoir multiphase and multi-flow state discrimination method and device |
-
2022
- 2022-01-27 CN CN202210102131.5A patent/CN114510974B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5545252A (en) * | 1995-03-01 | 1996-08-13 | The Perkin-Elmer Corporation | Flow regulation in gas chromatograph |
| CN1904580A (en) * | 2006-08-01 | 2007-01-31 | 东北电力大学 | Gas liquid two phase flow type identification method based on information coalescence and flow type signal collection device |
| CN104330336A (en) * | 2014-11-07 | 2015-02-04 | 中国石油大学(华东) | ICA and SVM-based gas-liquid two-phase flow pattern identification method |
| WO2019062595A1 (en) * | 2017-09-30 | 2019-04-04 | 深圳市岩尚科技有限公司 | Method for quality evaluation of photoplethysmogram |
| CN107991384A (en) * | 2017-12-21 | 2018-05-04 | 浙江启尔机电技术有限公司 | The detection device and method of flow pattern of gas-liquid two-phase flow in a kind of micro-pipe |
| CN111896616A (en) * | 2020-03-29 | 2020-11-06 | 中国石油大学(华东) | Gas-liquid two-phase flow pattern identification method based on acoustic emission-BP neural network |
| CN112507638A (en) * | 2020-12-18 | 2021-03-16 | 北京科技大学 | Deep condensate gas reservoir multiphase and multi-flow state discrimination method and device |
Non-Patent Citations (2)
| Title |
|---|
| 王强 ; 周云龙 ; 程思勇 ; 王俊霞 ; .基于小波和Elman神经网络的气液两相流流型识别方法.热能动力工程.2007,(02),58-61+65+116. * |
| 陈宏,匡波,杨扬,徐济.空气-水两相水平流动空泡率测量及流型识别研究.核动力工程.(S2),106-109. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114510974A (en) | 2022-05-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114510974B (en) | Intelligent recognition method for gas-liquid two-phase flow pattern in porous medium | |
| CN102109451B (en) | Non-contact conductive gas/liquid two-phase flow pattern identifying device and method | |
| Bin et al. | Time-frequency signal processing for gas-liquid two phase flow through a horizontal venturi based on adaptive optimal-kernel theory | |
| CN101984340B (en) | Method for identifying flow type of soft grain two-phase turbulence based on artificial intelligence | |
| CN103033567B (en) | Pipeline defect signal identification method based on guided wave | |
| CN107480705A (en) | A kind of running water pipe leakage detection method | |
| CN111896616B (en) | Gas-liquid two-phase flow pattern identification method based on acoustic emission-BP neural network | |
| CN105134619A (en) | Failure diagnosis and health evaluation method based on wavelet power, manifold dimension reduction and dynamic time warping | |
| Zheng et al. | Gas–liquid two phase flow measurement method based on combination instrument of turbine flowmeter and conductance sensor | |
| CN104807589A (en) | Online identification method for gas-liquid two-phase-flow flow pattern in gathering and transportation-vertical pipe system | |
| CN110953488B (en) | Gas-liquid two-phase flow pipeline leakage acoustic emission detection method based on stack self-coding | |
| Sun et al. | Neural networks approach for prediction of gas–liquid two-phase flow pattern based on frequency domain analysis of vortex flowmeter signals | |
| Khan et al. | Flow Regime Identification in Gas-Liquid Two-Phase Flow in Horizontal Pipe by Deep Learning | |
| CN113108853B (en) | Method for improving flow measurement accuracy of low-flow-velocity fluid | |
| Ying et al. | Applications of machine learning in pipeline monitoring | |
| CN104330336B (en) | Gas-liquid two-phase flow pattern recognition methods based on ICA and SVM | |
| CN106951845A (en) | Collect defeated riser systems flow pattern of gas-liquid two-phase flow method of discrimination and system | |
| CN1554936A (en) | Gas/liquid twe phase flow flow pattern identifying method and system based on soft measurement technology | |
| CN106338722A (en) | High-resolution radar one-dimension range profile target identification method based on multi-order samples | |
| Rojik et al. | Water pipe leak detection using the k-nearest neighbor method | |
| CN107402116B (en) | A kind of Method for Discriminating Gas-liquid Two Phase Flow and detection device | |
| CN105139025B (en) | Gas-solid fluidized bed flow pattern online intelligent recognition method based on nonlinear analysis method | |
| Gao et al. | Leakage detection of pipeline based on weighted-permutation entropy | |
| CN109806815B (en) | Method for detecting gas-liquid dispersion state of jet bubbling reactor based on liquid velocity fluctuation | |
| Wang et al. | Multiple blockage identification of drainage pipeline based on VMD feature fusion and support vector machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |