CN108593191A - A kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods - Google Patents
A kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods Download PDFInfo
- Publication number
- CN108593191A CN108593191A CN201810178185.3A CN201810178185A CN108593191A CN 108593191 A CN108593191 A CN 108593191A CN 201810178185 A CN201810178185 A CN 201810178185A CN 108593191 A CN108593191 A CN 108593191A
- Authority
- CN
- China
- Prior art keywords
- gas
- liquid
- pressure
- neural network
- phase flow
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01L—MEASURING FORCE, STRESS, TORQUE, WORK, MECHANICAL POWER, MECHANICAL EFFICIENCY, OR FLUID PRESSURE
- G01L13/00—Devices or apparatus for measuring differences of two or more fluid pressure values
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B01—PHYSICAL OR CHEMICAL PROCESSES OR APPARATUS IN GENERAL
- B01D—SEPARATION
- B01D47/00—Separating dispersed particles from gases, air or vapours by liquid as separating agent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The present invention relates to a kind of cleaner gas liquid two-phase flow form identifying system and methods, neural network and time domain parsing, wavelet analysis are organically combined, form the acquisition of pressure signal, the analysis of pressure difference signal, a full set of complete identification process of gas-liquid mixture phase identification, the serious forgiveness of neural network is high, nonlinearity is high, it compensates for and the high drawback of error was identified using frequency spectrum analysis method in the past, means of identification is precisely, rapidly.The effect monitored in real time is played as this kind of wet-scrubbing equipment of cardinal principle to generating retention mass using gas shock liquid level, measuring recognition accuracy by experiment reaches 99.6%.The characteristic parameter that pressure difference signal is parsed realizes the rapid conversion measured between identification as input sample feature vector library by training neural network.
Description
Technical field
The present invention relates to a kind of cleaner gas liquid two-phase flow form identifying system and methods, can join to different operation
The gas-liquid two-phase flow pattern of (liquid level, wind speed etc.) is accurately differentiated under several, to make corresponding regulation and control in time, makes dedusting
Device gas liquid two-phase flow form is in a kind of with the generation of gas shock liquid level under the preferable operating mode of dust catching effect, being suitable for
Retention mass is the monitoring and optimization of the operating status of the wet-scrubbing equipment of cardinal principle.
Background technology
Retention mass is generated as a kind of wet-scrubbing equipment of cardinal principle using gas shock liquid level, can handle high humidity, height
Temperature and the exhaust gas containing sulfide, ammonia, flue dust etc. are suitable for Traditional mining, metallurgy, chemical industry, cement, the contour row of Industrial Stoves
Clearance industry gas cleaning supervises such cleaner operating status in the case where flue gas emission limit is increasingly stringenter
It surveys and optimization has important social effect.
Gas-liquid two-phase admixture is an important symbol for differentiating such cleaner to dust purification effect quality.It removes
The amount of liquid phase disclosure satisfy that in the water tank of dirt equipment when needing, and when dust-contained airflow is weaker to the impact of liquid phase, gas-liquid two-phase connects
Touch it is uneven, insufficient, the retention mass of generation it is less, the dust in air-flow cannot be retained down, dust is easily straight with air-flow
It connects and is drained into air, cause certain atmosphere pollution, dust removing effects not ideal enough;When dust-contained airflow is stronger to the impact of liquid phase
When, gas-liquid two-phase contact is more uniformly distributed, fully, generates a large amount of retention mass, the dust in air-flow is retained down, clean gas
Stream is drained into air, and dust removing effects are ideal.Conversely, when the amount of liquid phase cannot meet the needs in water tank, no matter gas phase is to gas
The impact of stream is strong and weak, not will produce more retention mass, meets the needs for catching dirt, and dust removing effects are poor, and can be further
Equipment operation is influenced, is resulted in greater loss.
But in existing equipment operational process, in closed rugged environment, staff is difficult to observe in real time for many places
The operational effect of equipment, moreover, currently, the instrument price of on-line monitoring efficiency of dust collection is expensive, unsuitable high concentrate dust discharges
Link.
The present invention develops a kind of system of the such cleaner gas-liquid two-phase flow pattern of monitoring, is run by collecting device
Pressure difference signal in journey under different operation parameter, and further parsing has been invented a kind of dedusting and has been set using the method for neural network
The new method of standby gas-liquid mixture phase reaches accurate differentiation and monitoring to equipment gas-liquid two-phase flow pattern, cleaner is made to run
State is clearly illustrated in face of staff, can be made corresponding regulation and control in time, be made cleaner long-time stable
Efficient operation, reaches preferable dust removing effects.
Invention content
One is provided the present invention be directed to labile gas-liquid two-phase admixture in above-mentioned cleaner operational process
Kind wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods.It is further to its by acquiring the Pressure Fluctuation Signal of cleaner
Parsing, obtains corresponding time and frequency parameter, builds the sample database of corresponding flow pattern, by neural network method, reach to dedusting
Equipment flow pattern accurately differentiates, reflects the quality of cleaner dust removing effects indirectly.
The technical solution adopted by the present invention is:A kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods, passes through pressure
Acquisition system acquires the Pressure Fluctuation Signal of cleaner, is further parsed to it, obtains corresponding time and frequency parameter, structure is corresponding
The sample database of flow pattern is reached and is accurately differentiated to cleaner flow pattern by neural network method, the specific steps are:
Step 1) gas-liquid mixture phase divides, and adjusts the liquid level and wind speed of cleaner, obtains the difference of deduster operation
Gas-liquid mixture phase is divided by operating status:Hydrostatic potential difference pattern, faint fluctuation model, low liquid level gas-liquid resonance mode are high
Liquid level gas-liquid resonance mode and volume air-breathing steep pattern;
Step 2) signal acquisition acquires the time series of the Pressure Fluctuation Signal under different operating modes using differential pressure pick-up,
Pressure signal in each case is not less than 500 samples;
Step 3) signal is analyzed and feature vector library is established, and is parsed to collected Differential Pressure Fluctuation signal of Gas, 1. first
Time domain parsing is carried out to Differential Pressure Fluctuation signal of Gas in each case, obtains the systems such as the pressure mean value, standard deviation, degree of skewness of pressure difference signal
Characteristic parameter is counted, gas liquid two-phase flow form pressure characteristics of mean vector library, standard difference vector library, degree of skewness vector library are established;
2. then, carrying out at least three layers of wavelet analysis to Differential Pressure Fluctuation signal of Gas in each case, extraction reflects the small of each nowed forming
Wave Decomposition coefficient energy value, by taking three layers of wavelet analysis as an example, including each band energy value EA3、ED3、ED2And ED1Four category features are joined
Number, and establish corresponding characteristic vector data library;3. statistical analysis feature, wavelet analysis feature in each case is placed into
Together, the specific characteristic vector data library of each gas liquid two-phase flow form is constituted;
Step 4) Establishment of Neural Model establishes corresponding neural network model, is carried out using BP neural network model
Neural metwork training;It is using the characteristic vector data library of above-mentioned analysis as input sample, the output quantity of different flow patterns is fixed respectively
Justice is:Faint fluctuation model [0,0,0,0,0,1], low liquid level gas-liquid resonance state [1,0,0,0,0,0], shearing liquid tentiform state
[0,0,1,0,0,0], hydrostatic potential difference state [0,0,0,1,0,0], high liquid level gas-liquid resonance state [0,1,0,0,0,0] and volume are inhaled
The suitably hiding number of plies is arranged as output sample database in bubble regime [0,0,0,0,1,0], and the part for choosing sample database is made
Neural network is then subjected to repetition training as verify data as test data, a part for training sample, a part,
The highest parameter setting of accuracy of identification is obtained, best neural network model is obtained, finally carries out trained neural network
It preserves;
Trained neural network model is connected to each other with pressure acquisition system, will adopt by the realization of step 5) Flow Regime Ecognition
The pressure data collected is parsed, then by calling the neural network model having had built up, to each characteristic parameter
Carry out analysis and distinguishing, the final accurately differentiation for realizing gas liquid two-phase flow form.
Beneficial effects of the present invention:
(1) present invention by neural network and time domain parsing, wavelet analysis organically combine, form pressure signal acquisition,
The analysis of pressure difference signal, a full set of complete identification process of gas-liquid mixture phase identification, the serious forgiveness of neural network is high, nonlinearity
Height compensates for and the high drawback of error was identified using frequency spectrum analysis method in the past, and means of identification is precisely, rapidly.
(2) present invention can rise to generating retention mass using gas shock liquid level as this kind of wet-scrubbing equipment of cardinal principle
To the effect monitored in real time, recognition accuracy is measured by experiment and reaches 99.6%.
(3) characteristic parameter that the present invention parses pressure difference signal passes through training as input sample feature vector library
Good neural network realizes the rapid conversion measured between identification.
Description of the drawings
Fig. 1 is the structural schematic diagram of pressure acquisition system.
In figure, 1. cleaners, 2. clean gas flow output ends, 3. pressure sensors, 4. data collecting cards, 5. computers.
Specific implementation mode
Pressure acquisition system is made of pressure sensor 3, data collecting card 4 and computer 5.
Pressure sensor 3 selects the sensor that market generally uses, range:0~3kPa, signal output:0~5V DC, essence
Degree is not less than 0.5%, and response time≤5ms, precision is less than ± 0.1%.Sensor mounting location is the clean of cleaner 1 to be measured
Net airflow output end 2 carrys out the Differential Pressure Fluctuation signal of Gas of measuring apparatus entirety.
Data collecting card 4 selects market routine high-speed collection card, sample frequency to be not less than 1kHz, and sampling number is not less than
5120。
Computer 5 selects market conventional computer, system requirements to be not less than windows Xp, memory 2GB.
The identification of 1 inside gas-liquid mixture phase of dust removal combined equipment, make specific elaboration the present invention.
(1) liquid level and wind speed for adjusting cleaner 1, obtain the different mode of deduster operation:Hydrostatic potential difference pattern, it is micro-
Weak fluctuation model, low liquid level gas-liquid resonance mode, high liquid level gas-liquid resonance mode, volume air-breathing steep pattern.It is sensed using differential pressure
Device is acquired the pressure signal under different flow patterns, sample frequency 1024, and sampled point is 5120 points, acquires within every five seconds
Once, each flow pattern extracts at least 500 groups of experimental datas;
(2) collected Differential Pressure Fluctuation signal of Gas is parsed, 1. Differential Pressure Fluctuation signal of Gas in each case is carried out first
Time domain parses, and obtains the statistical natures parameters such as the pressure mean value, standard deviation, degree of skewness of pressure difference signal, and establishes corresponding statistics
Analyze characteristic vector data library;2. then, carrying out at least three layers of wavelet analysis, extraction to Differential Pressure Fluctuation signal of Gas in each case
Reflect the coefficient of wavelet decomposition energy value of each nowed forming, including each band energy value EA3、ED3、ED2And ED1Four category feature parameters,
And establish corresponding wavelet analysis characteristic vector data library;3. by statistical analysis feature, wavelet analysis feature in each case
It is placed into together, constitutes seven kinds of specific characteristic vector data library (pressure mean value, standards of each gas liquid two-phase flow form
Difference, degree of skewness, EA3、ED3、ED2And ED1)。
(3) using seven feature vectors databases of above-mentioned analysis as input sample, the output quantity of different flow patterns is distinguished
It is defined as:Faint fluctuation model [0,0,0,0,0,1], low liquid level gas-liquid resonance state [1,0,0,0,0,0], shearing liquid tentiform state
[0,0,1,0,0,0], hydrostatic potential difference state [0,0,0,1,0,0], high liquid level gas-liquid resonance state [0,1,0,0,0,0] and volume are inhaled
Bubble regime [0,0,0,0,1,0], as output sample database.The 70% of selected characteristic vector library is instructed as training sample
Practice, then choose therein 15% data as the generalization ability of verification neural metwork training, it is last finally to choose data
15% is used as test data.It is 12 that the number of plies is hidden in setting, and neural network is carried out repetition training, its performance is made to reach best.
(4) trained neural network model is preserved, is then tested:By trained neural network model and pressure
Power acquisition system is connected to each other, and then by calling the neural network model having had built up, is divided each characteristic parameter
Analysis differentiates, finally realizes the accurately differentiation of gas liquid two-phase flow form.
(5) simulation and forecast is carried out to trained neural network model with the ten groups of data newly measured, obtained completely correct
Flow Regime Ecognition result (as shown in table 1).
Part sample, characteristic and the recognition result that 1 present invention of table is embodied.
Claims (6)
1. a kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods acquires the pressure of cleaner by pressure acquisition system
Fluctuation signal further parses it, obtains corresponding time and frequency parameter, build the sample database of corresponding flow pattern, pass through nerve
Network method reaches and accurately differentiates to cleaner flow pattern, characterized in that the specific steps are:
Step 1) gas-liquid mixture phase divides, and adjusts the liquid level and wind speed of cleaner, obtains the different operations of deduster operation
Gas-liquid mixture phase is divided by state:Hydrostatic potential difference pattern, faint fluctuation model, low liquid level gas-liquid resonance mode, high liquid level
Gas-liquid resonance mode and volume air-breathing steep pattern;
Step 2) signal acquisition acquires the time series of the Pressure Fluctuation Signal under different operating modes using differential pressure pick-up, each
In the case of pressure signal be not less than 500 samples;
Step 3) signal is analyzed and feature vector library is established, and is parsed to collected Differential Pressure Fluctuation signal of Gas, 1. first to every
Differential Pressure Fluctuation signal of Gas carries out time domain parsing in the case of kind, and it is special to obtain the statistics such as pressure mean value, standard deviation, degree of skewness of pressure difference signal
Parameter is levied, gas liquid two-phase flow form pressure characteristics of mean vector library, standard difference vector library, degree of skewness vector library are established;2. so
Afterwards, at least three layers of wavelet analysis are carried out to Differential Pressure Fluctuation signal of Gas in each case, extraction reflects the small wavelength-division of each nowed forming
Solve coefficient energy value;By taking three layers of wavelet analysis as an example, extraction includes each band energy value EA3、ED3、ED2And ED1Four category features are joined
Number, and establish corresponding characteristic vector data library;3. statistical analysis characteristic parameter in each case, wavelet analysis feature are joined
Number is placed into together, constitutes the specific characteristic vector data library of each gas liquid two-phase flow form;
Step 4) Establishment of Neural Model establishes corresponding neural network model, and nerve is carried out using BP neural network model
Network training;Using the characteristic vector data library of above-mentioned analysis as input sample, the output quantity of different flow patterns is respectively defined as:
Faint fluctuation model [0,0,0,0,0,1], low liquid level gas-liquid resonance state [1,0,0,0,0,0], shearing liquid tentiform state [0,0,1,
0,0,0], hydrostatic potential difference state [0,0,0,1,0,0], high liquid level gas-liquid resonance state [0,1,0,0,0,0] and volume air-breathing blister
The suitably hiding number of plies is arranged as output sample database in state [0,0,0,0,1,0], chooses a part for sample database as training
Sample, a part are used as verify data as test data, a part, then, neural network are carried out repetition training, obtains and knows
The other highest parameter setting of precision, obtains best neural network model, finally preserves trained neural network;
Trained neural network model is connected to each other with pressure acquisition system, will collect by the realization of step 5) Flow Regime Ecognition
Pressure data parsed, then by calling the neural network model that has had built up, to the progress of each characteristic parameter
Analysis and distinguishing, the final accurately differentiation for realizing gas liquid two-phase flow form.
2. a kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods according to claim 1, characterized in that establish phase
The feature answered is to vectorial library, including statistical nature vector and wavelet analysis feature vector:The statistical nature vector includes pressure
Power mean value, standard deviation and degree of skewness;The wavelet analysis feature vector includes each frequency range coefficient of wavelet decomposition energy value EA3、ED3、
ED2、ED1。
3. a kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods according to claim 3, characterized in that pressure is adopted
Collecting system is made of pressure sensor, data collecting card and computer.
4. a kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods according to claim 3, characterized in that pressure passes
Sensor range:0~3kPa, signal output:0~5V DC, precision be not less than 0.5%, the response time≤5ms, precision be less than ±
0.1%;Sensor mounting location is the clean gas flow output end of cleaner to be measured, the pressure-difference fluctuation letter of measuring apparatus entirety
Number.
5. a kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods according to claim 3, characterized in that data are adopted
4 sample frequency of truck is not less than 1kHz, and sampling number is not less than 5120.
6. a kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods according to claim 3, characterized in that computer
System requirements is not less than windows Xp, memory 2GB.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810178185.3A CN108593191A (en) | 2018-03-05 | 2018-03-05 | A kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810178185.3A CN108593191A (en) | 2018-03-05 | 2018-03-05 | A kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108593191A true CN108593191A (en) | 2018-09-28 |
Family
ID=63625689
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810178185.3A Pending CN108593191A (en) | 2018-03-05 | 2018-03-05 | A kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108593191A (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1554936A (en) * | 2003-12-19 | 2004-12-15 | 浙江大学 | Gas/liquid twe phase flow flow pattern identifying method and system based on soft measurement technology |
CN1664555A (en) * | 2005-03-17 | 2005-09-07 | 上海交通大学 | Two-phase fluid flow pattern identification method based on time sequence and neural net pattern identification |
CN104807589A (en) * | 2015-04-17 | 2015-07-29 | 西安交通大学 | Online identification method for gas-liquid two-phase-flow flow pattern in gathering and transportation-vertical pipe system |
CN105806643A (en) * | 2016-05-24 | 2016-07-27 | 中国矿业大学 | Identification method and device for gas-liquid mixing state of dust remover |
WO2017065684A1 (en) * | 2015-10-15 | 2017-04-20 | Fredrik Winquist | An efficient gas/liquid separator for identification and measurement of volatile compounds in a liquid |
-
2018
- 2018-03-05 CN CN201810178185.3A patent/CN108593191A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1554936A (en) * | 2003-12-19 | 2004-12-15 | 浙江大学 | Gas/liquid twe phase flow flow pattern identifying method and system based on soft measurement technology |
CN1664555A (en) * | 2005-03-17 | 2005-09-07 | 上海交通大学 | Two-phase fluid flow pattern identification method based on time sequence and neural net pattern identification |
CN104807589A (en) * | 2015-04-17 | 2015-07-29 | 西安交通大学 | Online identification method for gas-liquid two-phase-flow flow pattern in gathering and transportation-vertical pipe system |
WO2017065684A1 (en) * | 2015-10-15 | 2017-04-20 | Fredrik Winquist | An efficient gas/liquid separator for identification and measurement of volatile compounds in a liquid |
CN105806643A (en) * | 2016-05-24 | 2016-07-27 | 中国矿业大学 | Identification method and device for gas-liquid mixing state of dust remover |
Non-Patent Citations (4)
Title |
---|
XIAOCHUAN LI等: ""Study of gas–liquid two-phase flow patterns of self-excited dust scrubbers"", 《CHEMICAL ENGINEERING SCIENCE》 * |
李枫: ""两相流流型在线识别技术与应用软件设计"", 《中国优秀硕士学位论文全文数据库(电子期刊)》 * |
程旭丽: ""气液两相流流型在线识别系统"", 《中国优秀硕士学位论文全文数据库(电子期刊)》 * |
魏涛等: ""除尘器气液两相流动形态识别的特征参数挖掘"", 《机械工程学报》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108414923A (en) | A kind of analog-circuit fault diagnosis method based on the extraction of depth confidence network characterization | |
CN105044590B (en) | A kind of single measuring node analog-circuit fault diagnosis method | |
CN1194227C (en) | Fast non-destructive detection method and device of food smell based on gas sensor array technology | |
CN109100627A (en) | A kind of power equipment partial discharges fault diagnostic method based on end-to-end mode | |
CN105223240A (en) | A kind of method utilizing detection by electronic nose crab freshness | |
CN102879436A (en) | Method of using electronic nose for detecting freshness of river crucian carp | |
CN100507509C (en) | Oil gas water multiphase flow type identification method based on main component analysis and supporting vector machine | |
CN101226133B (en) | Method for specification and recognition of hemocyte pulse signal | |
CN111666169A (en) | Improved isolated forest algorithm and Gaussian distribution-based combined data anomaly detection method | |
CN104569062B (en) | Electronic nose air chamber based on molecular sieve filtration | |
CN112557894B (en) | Breaker fault diagnosis method | |
CN109187287A (en) | Source of atmospheric particulate matter analytic method based on stable element particle diameter distribution information | |
CN205898730U (en) | Multichannel polytypic sensor capability test system | |
CN103886405A (en) | Boiler combustion condition identification method based on information entropy characteristics and probability nerve network | |
CN106419911A (en) | Emotional detection method based on brain electric wave analysis | |
CN103389323B (en) | Method for evaluating ages of precious medicinal materials quickly and losslessly | |
CN115453064B (en) | Fine particulate matter air pollution cause analysis method and system | |
CN105738581B (en) | A kind of method for quick identification of the different freshness hickory nuts based on electronic nose | |
CN104111274A (en) | Method for identifying producing area of red bayberry juice by using gas sensor array type electronic nose fingerprint analysis system | |
CN113011660A (en) | Air quality prediction method, system and storage medium | |
CN108398533A (en) | Electric nasus system and its air source discriminating in storage and localization method | |
CN109142494A (en) | One kind being directed to resistor-type electrochemical sensor optimum operating voltage choosing method | |
CN108593191A (en) | A kind of wet-scrubbing equipment gas-liquid two-phase flow pattern recognition methods | |
CN102282932A (en) | Method for rapidly detecting germination rate of vegetable seeds | |
CN115169504B (en) | Equipment abnormity identification method in coal gas fine desulfurization process |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180928 |
|
RJ01 | Rejection of invention patent application after publication |