CN106706215A - Thermodynamic system valve inner leakage monitoring method - Google Patents
Thermodynamic system valve inner leakage monitoring method Download PDFInfo
- Publication number
- CN106706215A CN106706215A CN201611019372.4A CN201611019372A CN106706215A CN 106706215 A CN106706215 A CN 106706215A CN 201611019372 A CN201611019372 A CN 201611019372A CN 106706215 A CN106706215 A CN 106706215A
- Authority
- CN
- China
- Prior art keywords
- leakage
- valve
- sample
- neural network
- valves leakage
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M3/00—Investigating fluid-tightness of structures
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
Abstract
The invention provides a thermodynamic system valve inner leakage monitoring method. The monitoring method comprises the steps that (1) valve inner leakage modeling is carried out to acquire sample large data between the valve inner leakage and a characteristic parameter; (2) data preprocessing is carried out to improve the learning precision and efficiency of a recognition model, and validity verification is carried out on sample large data between the valve inner leakage and the characteristic parameter in the step (1) to remove error data and normalize the data; and (3) based on an improved SVM nonlinear combination recognition model, a BP neural network recognition model, an RBF radial recognition model and a GRNN neural network recognition model are combined to calculate the thermodynamic system valve inner leakage under the specific characteristic parameter to monitor the valve inner leakage. According to the invention, a scientific professional basis is provided for an enterprise to arrange overhauling and change a planned overhauling mode according to the valve leakage condition, and the method plays an important role in reducing the valve leakage of the enterprise and improving the economy of production operation.
Description
Technical field
The present invention relates to the technical field of valves leakage amount monitoring, more particularly to therrmodynamic system valves leakage amount monitoring side
Method.
Background technology
Valve is wide in industrial production and measures a kind of big equipment, is extremely important control unit in fluid delivery system
Part, can be used to control water, air, steam, various Korrosionsmediums, mud, oil product, liquid metal and radiating medium etc. various
The flowing of type of fluid.Its basic function is turned off or connects the circulation of pipeline medium, changes the flow direction of medium, and control is situated between
The pressure and flow of matter, the normal operation of protection pipeline and equipment.Valve is widely used in power plant, and it is responsible in connection power plant
Each subsystem, is to ensure that the important accessory of power plant's safety work, security performance and maintenance of its reliability level to system
Cost impact is especially protruded.
Because power plant valve is chronically in the environment of HTHP, internal leakage fault frequently occurs.Valves leakage shadow
Ring power plant safety production:Equipment cannot isolate defect elimination during valves leakage will make operation, and safety measure cannot be performed in place, serious prestige
Coerce the life security of service work personnel;Valves leakage may cause to wash away to elbow of pipeline or flash vessel etc., when serious very
To the outer pipeline booster of machine is caused, the field personnel to equipment brings hazard to person, causes unit unplanned outage;With high pressure
As a example by bypass valve leakage, desuperheating water serious leak will cause Unit Commitment stage cold reheaing steam pipe ponding, trigger water hammer
Even produce the major accident of turbine water induction.Meanwhile, valves leakage will have a strong impact on unit economy, be leaked with hydrophobic valve
As a example by, according to statistics:40% interior leakage influences unit heat consumption rate 1% or so;60% interior leakage influence unit heat consumption rate 1% with
On;Leakage influence unit heat consumption rate is 4% or so in indivedual unit hydrophobic valves.Valves leakage can also reduce unit operating efficiency,
Increase maintenance of equipment expense.Because different source of leaks have different Leak Mechanisms, it causes fault progression trend and serious journey
Degree is different, and the required maintenance policy for taking also should be different.Some can be solved in production run by taking corresponding measure,
And some then need maintenance down or change valve.According to statistics, in overhaul, more than 50% valve is all to be overhauled
Disintegrate, in the case of not clear interior leakage reason, valve is carried out to overhaul disintegration, not only wasted human and material resources, it is also possible to make
Into some artificial damages.
It is desirable to thering is a kind of therrmodynamic system valves leakage quantity monitoring method to overcome or at least mitigating prior art
Drawbacks described above.
The content of the invention
Deposited in the prior art to overcome it is an object of the invention to provide a kind of therrmodynamic system valves leakage quantity monitoring method
Above mentioned problem.
To achieve the above object, the present invention provides a kind of therrmodynamic system valves leakage quantity monitoring method, the monitoring method
Comprise the following steps:
(1) sample big data between valves leakage amount and characteristic parameter is obtained by valves leakage modeling;
(2) data prediction, to improve the study precision and efficiency of identification model, to valves leakage described in step (1)
Sample big data carries out validation verification between amount and characteristic parameter, and rejects wrong data data are normalized;
(3) based on the method for improving SVM nonlinear combination identification models, BP neural network identification model, RBF are radially known
Other model and GRNN neural network recognization models are combined, and calculate the therrmodynamic system valves leakage amount under special characteristic parameter,
So as to realize the monitoring to valves leakage amount.
Preferably, the step (1) includes:Therrmodynamic system duct length, internal diameter, external diameter and insulation are obtained by measuring
Thickness degree, then sets up pipeline infinitesimal section radial direction temperature field;Pipeline infinitesimal section axial direction temperature field is set up, along pipeliner
Matter flow direction, calculates each infinitesimal section paragraph by paragraph, finally gives valve preceding pipeline temperature field, each leakage rate correspondence one
Tube wall temperature.
Preferably, the therrmodynamic system pipeline is divided into multiple infinitesimal pipeline sections, according to heat transfer principle to the infinitesimal
Pipeline section carries out radial and axial modeling, and is solved along Working fluid flow direction using iterative method;According to N sections of infinitesimal pipeline section
Friction loss and heat exchange amount calculate the N sections of infinitesimal pipeline section outlet parameter, and in this, as the N+1 sections of infinitesimal pipe
The intake condition of section, until obtaining the wall surface temperature of last the infinitesimal pipeline section before valve, it is established that join with feature in temperature field
The large sample model of several.
Preferably, it is described based on improve SVM nonlinear combination identification models method the step of (3) including walking in detail below
Suddenly:
Step A:With the BP neural network identification model, RBF radial direction base identification models and GRNN neural network recognization moulds
Type is identified respectively to sample big data between the middle valves leakage amount for pre-processing of step (2) and characteristic parameter, by what is acquired
The discre value of various recognition methods forms new test sample and training sample with actual value, and as improving, SVM is non-linear
The sample of built-up pattern;
Step B:Initialization SVM models, to Lagrange multiplier αi、And threshold value b carries out random assignment;
Step C:New training sample is established as meeting the object function of SVM algorithm, it is entered using LIBSVM algorithms
Row is solved, and obtains Lagrange multiplier αi、And the value of threshold value b;
Step D:The Lagrange multiplier α that will be calculated by step Ci、And threshold value b values are brought into object function, use
Test sample is calculated based on the valve leakage automatic inspection for improving SVM nonlinear combinations identification model under special characteristic parameter;
Step E:The valve leakage automatic inspection calculated by step D and the actual leakage rate of valve are compared, calculation error,
When error is less than the precision for determining, then terminate learning process, if not reaching determination precision, return to step B proceeds to learn
Practise;
Step F:Calculated after special characteristic parameter is input into using the combination identification model after the completion of study, drawn heat
Force system valves leakage amount, realizes the monitoring to valves leakage amount.
Preferably, the BP neural network identification model in the step (3) is comprising 1 layer of three-layer network of hidden layer
Structure, input layer number is 3, and the input layer is respectively therrmodynamic system entrance power pressure, entrance
Tube wall temperature before Temperature of Working and valve;Output layer neuron is valve leakage automatic inspection;The quantity of intermediate layer neuron is using progressively
Experimental method, is determined with the minimum target of training sample root-mean-square error, and the quantity of the intermediate layer neuron is 12;It is implicit
Layer neural transferring function uses S type tans, output layer neural transferring function to use S type logarithmic functions, training algorithm
Using Levenberg-Marquardt algorithms.
Preferably, the RBF radial directions identification model in the step (3) uses input layer, hidden layer, the three of output layer
Rotating fields, input layer quantity is 3, and input layer is respectively therrmodynamic system entrance power pressure, entrance
Tube wall temperature before Temperature of Working and valve;Output layer neuron is valve leakage automatic inspection;For hidden layer Gaussian function center, use
Orthogonal Least Square is chosen, and network output weights are trained using least square method.
Preferably, the connection between the structure and neuron of GRNN neural network recognization models described in the step (3)
Sample big data determines weights between valves leakage amount and characteristic parameter described in the step (1), the GRNN neutral nets
The smoothing parameter of identification model is determined by test method(s), and according to sample characteristics, selection smoothing parameter is 0.1-0.9, and with step
A length of 0.05 change.
For the inapplicable or precision present situation not high that current therrmodynamic system valves leakage quantity monitoring method both at home and abroad is present,
The invention discloses a kind of therrmodynamic system valves leakage quantity monitoring method, the therrmodynamic system valves leakage quantity monitoring method is realized
Valve leak is monitored, the accuracy rate of valves leakage amount monitoring is improve.
Brief description of the drawings
Fig. 1 is the flow chart of therrmodynamic system valves leakage quantity monitoring method.
Fig. 2 is the basic functional principle schematic diagram of therrmodynamic system pipeline and valve.
Specific embodiment
To make the purpose, technical scheme and advantage of present invention implementation clearer, below in conjunction with the embodiment of the present invention
Accompanying drawing, the technical scheme in the embodiment of the present invention is further described in more detail.In the accompanying drawings, identical from start to finish or class
As label represent same or similar element or the element with same or like function.Described embodiment is the present invention
A part of embodiment, rather than whole embodiments.Embodiment below with reference to Description of Drawings is exemplary, it is intended to used
It is of the invention in explaining, and be not considered as limiting the invention.Based on the embodiment in the present invention, ordinary skill people
The every other embodiment that member is obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.Under
Face is described in detail with reference to accompanying drawing to embodiments of the invention.
Therrmodynamic system valves leakage quantity monitoring method of the invention is comprised the following steps:
(1) sample big data between valves leakage amount and characteristic parameter is obtained by valves leakage modeling;
(2) data prediction, to improve the study precision and efficiency of identification model, to valves leakage described in step (1)
Sample big data carries out validation verification between amount and characteristic parameter, and rejects wrong data data are normalized;
(3) based on the method for improving SVM nonlinear combination identification models, BP neural network identification model, RBF are radially known
Other model and GRNN neural network recognization models are combined, and calculate the therrmodynamic system valves leakage amount under special characteristic parameter,
So as to realize the monitoring to valves leakage amount.
It is as shown in Figure 1 the overview flow chart of present invention offer, mainly includes three steps:
(1) valves leakage modeling.Radially valves leakage is obtained by setting up pipeline infinitesimal section with the temperature field of axial direction
Sample big data between amount and characteristic parameter.
(2) data prediction.To improve the study precision and efficiency of identification model, data are carried out with validation verification, and
Reject wrong data and data normalization.
(3) interior leakage quantity combination identification modeling.Using based on BP neural network identification model, RBF radial direction bases identification model,
GRNN neural network recognization models, in combination with the recognition methods for improving SVM, obtain nonlinear combination identification model, by this
Combination identification model is processed pretreated data, finally realizes the monitoring to valves leakage amount.
It is as shown in Figure 2 heat distribution pipeline of the invention and valve basic functional principle, it is known that therrmodynamic system entrance work
The temperature t of matter0, pressure P0, tube wall temperature t before actually measured valve can calculate valve leakage automatic inspection G.Following function is obtained to close
It is formula:
G=f (P0,t0,L,D,D1,H,t)
In formula:P0:Heat distribution pipeline entrance power pressure;t0:Heat distribution pipeline entrance Temperature of Working;t:Tube wall temperature before valve;
L:Duct length;D:Internal diameter of the pipeline;:Outer diameter tube;H:Insulation layer thickness.
After heat distribution pipeline determines, only it is to be understood that P0,t0, t, it is possible to determine valve leakage automatic inspection.Therefore, setting up thermal pipe
Channel temp, under temperature, the pressure condition of known entrance working medium, is calculated the tube wall before different leakage rate lower valves
Temperature.Obtain following functional relation:
T=g (P0,t0,L,D,D1,H,G)
By the leakage rate and the data such as working medium and valve preceding pipeline wall temperature that obtain, set up training sample set, with input
The hybrid intelligent training pattern for building, obtains nonlinear function model result.So that it is determined that valve leakage automatic inspection, realizes in valve
The monitoring of leakage quantity.
As Fig. 1 therrmodynamic system valves leakage quantity monitoring method specific implementation process is:
Step 1:Modeled by valves leakage and obtain big-sample data between valves leakage amount and characteristic parameter.
In order to calculate thermo parameters method of the pipe-line system under different leakage operating modes, pipe-line system is divided into several micro-
First section, each infinitesimal section can be considered simple cylinder, regard each infinitesimal segment pipe as a control volume, sets up and radially passes
Thermal model.Because working medium is in continuous heat release, so working medium is gradually reduced along flow direction temperature, by setting up between infinitesimal section
Temperature field relation, i.e. axial direction models for temperature field can paragraph by paragraph calculate to obtain tube wall temperature before valve from entrance.
Step 101:Set up pipeline infinitesimal section radial direction temperature field.
Intraductal working medium in the heat exchange mode heat convection successively between working medium and inside pipe wall of radial direction, lead by tube wall and heat-insulation layer
Heat heat exchange, and the heat convection between heat-insulation layer outer wall and nature.Because longitudinal heat conduction heat very little of heat-insulation layer and tube wall,
So the heat of these four radial direction heat exchange modes can approximately regard as it is equal.Due to pipeline longitudinal temperature gradient less, then will can manage
Wall regards multi-layer cylinder wall Heat Conduction Problems as with the radiating of heat-insulation layer.
Heat conduction transmission heat Calculation formula be:
In formula, k is the thermal conductivity factor of tube wall or heat-insulation layer, W/ (mK);Δ t is inside and outside wall temperature difference, DEG C;d1、d2, L point
Wei not the internal diameter of tube wall or heat-insulation layer, external diameter and length, m.
Heat convection transmits heat Calculation formula:
In formula, k is the thermal conductivity factor of fluid, W/ (mK);deqFor the equivalent of pipeline is directly the equivalent diameter of pipeline, m;F
It is heat exchange area, m2;Δ t is heat transfer temperature difference, DEG C;Nu is nusselt number.
(1) Nu between working medium and inside pipe wall during heat convection
Heat transfer of the pipeline to inner-walls of duct is forced convection heat release in pipe, and its nusselt number relational expression is:
During laminar flow:
Formula requirementIf being less than 2, Nu=3.66.
During turbulent flow:
When working medium is steam, in the case of valve leakage automatic inspection is smaller, Temperature of Working is gradually decreased to saturation temperature, and steam starts
Condensed in pipe, when the steam in pipe does not condense completely:
In formula (3)~formula (6), Pr is Prandtl number, and Re is Reynolds number, ηfWith ηwWork at respectively average working medium and wall
The power viscosity coefficient of matter, pl、vlThe respectively density of saturation water, dynamic viscosity coefficient, pv、wvFor saturated vapor density and
Speed, x1、x2To calculate the steam quality that infinitesimal section is imported and exported.
(2) Nu between heat-insulation layer outer wall and nature during heat convection
In general, heat pipeline heat insulation layer outer wall is mainly free convection with the heat exchange method of surrounding environment,
Its level is with the criterion equation of vertical cylinder heat transfer free convection:
In formula, Ra=GrPr, referred to as Rayleigh number.
Step 102:Set up pipeline infinitesimal section axial direction temperature field.
Using cell cube import fluid properties as the cell cube working medium qualitative parameter, cell cube export working medium pressure
PoutWith temperature outlet working medium ToutObtained by formula (8)-formula (10), so that it is determined that the qualitative ginseng of next segment unit body import working medium
Number.
In formula:QfrThermal discharge of the working medium in cell cube, G valve leakage automatic inspections, cpThe specific heat capacity at constant pressure of the cell cube working medium,
The Δ t cell cubes import and export the temperature difference of working medium, and the density of the ρ cell cube import working medium, D valve preceding pipeline internal diameters, v working medium flows through this
The average speed of cell cube, η frictional resistant coefficients, the pressure drop that Δ P working medium is produced by the cell cube, PinThe cell cube import work
The pressure of matter, PoutThe cell cube exports the pressure of working medium, L unit body lengths.
Solved using alternative manner, whole pipeline is logically separated into N sections, calculated paragraph by paragraph along Working fluid flow direction, directly
To valve.
Step 2 data prediction.Including Validation of Data, pick out wrong data and data normalization.
In order to improve the study precision and efficiency of identification model, it is necessary to the valves leakage amount and feature of acquisition in previous step
Big-sample data is pre-processed between parameter, carries out Validation of Data, by experiment or measured data, to the valve in sample
Leakage quantity is verified and adjusted with the mapping relations of its characteristic value in door, and picks out wrong data, mainly includes entrance work
Valves leakage amount that data and its correspondence during matter measuring instrumentss failure are obtained etc..Meanwhile, because neuron training is present satisfying
With problem, it is necessary to be normalized to sample data, the present embodiment is interval to [0.1,0.9] by data normalization, passes through
Equation below is realized:
In formula:X is data, the valve leakage automatic inspection for such as calculating, X in sampleminIt is the data minimum value in sample, XmaxFor
Data maximums in sample, Y is that sample data normalizes result.
Step 3:The combination identification modeling of valves leakage amount
It is by BP neural network identification model, RBF radial direction bases using based on the nonlinear combination identification model for improving SVM
Identification model, GRNN neural network recognization models couplings get up, and the advantage of these three recognition methods are comprehensively utilized, to improve system
Identification accuracy.Its course of work includes two steps:(1) modeled respectively for three kinds of neural network recognization models;With
BP neural network identification model, RBF radial direction bases identification model, GRNN neural network recognizations model are to pretreated valves leakage
Sample big data is modeled respectively between amount and characteristic parameter, the predicted value and its actual value of the various recognition methods for obtaining;(2)
Using three kinds of neural network recognization results, new test sample and training sample is formed, be improved SVM combination identification models
Modeling.
Step 301:Three kinds of neural network recognization models are modeled respectively.
With BP neural network identification model, RBF radial direction bases identification model, GRNN neural network recognization models to pretreatment after
Valves leakage amount and characteristic parameter between sample big data be modeled respectively, the predicted value of the various recognition methods for obtaining and its
Actual value;
(1) BP neural network identification model modeling
BP neural network refers to the multilayer feedforward neural network based on error backpropagation algorithm, and it can be forced with arbitrary accuracy
Nearly any Nonlinear Mapping, it is special with certain fault-tolerance and preferable robust with distributed information storage and processing structure
Property, it is obtained in many prediction fields and is widely applied.
In this method, BP neural network is using only comprising 1 layer of Three Tiered Network Architecture of hidden layer, input layer number is
3, according to characteristic parameter in sample, respectively tube wall temperature before entrance power pressure, entrance Temperature of Working, valve
Degree.Output layer neuron number is 1, i.e. valve leakage automatic inspection.Intermediate layer neuron number is determined using progressively experimental method, to instruct
Practice the minimum target of sample root-mean-square error to determine, be 12.Hidden layer neuron transmission function uses S type tans,
Output layer neural transferring function uses S type logarithmic functions, training algorithm to use Levenberg-Marquardt algorithms.
(2) RBF radial directions base identification model
Since the possibility of local convergence is had in the learning process of BP neural network, and convergence has very high point with initial value
System.It is a kind of good feed forward type identification model and RBF radial direction bases identification model has very strong non-linear mapping capability.Institute
With, comprehensive utilization RBF radial direction base identification models come in the learning process of Optimized BP Neural Network to restrain to initial value it is excessive according to
The convergent problem of locality for relying and occurring.
In this method, RBF radial direction identification models are divided into three layers, are respectively input layer, hidden layer, output layer, input layer nerve
First number is 3, according to characteristic parameter in sample, is managed respectively before entrance power pressure, entrance Temperature of Working, valve
Wall temperature.Output layer neuron number is 1, i.e. valve leakage automatic inspection.For hidden layer Gaussian function center, using an orthogonal most young waiter in a wineshop or an inn
Multiplication is chosen, and network output weights are trained using least square method, and the target of its learning training is overall error
Reach minimum.
(3) GRNN neural network recognizations model
When BP neural network and RBF radial direction bases are identified, it may appear that be absorbed in that local minimum, training effectiveness be low and convergence
Speed waits not enough slowly.In view of these, introduce GRNN neural network recognization models and optimize, it is approached can carry out the overall situation
While also there is best Property of Approximation, contrast BP neural network and RBF radial direction bases, its approximation capability, classification capacity, study
Speed ability is all very prominent.
After learning sample determines, the connection weight between structure and neuron in GRNN neutral nets is also just determined down
Come, the training process of identification model is the process for determining smoothing parameter, and smoothing parameter Selection experiment method is determined, due to sample
Originally employ normalized, therefore selection smoothing parameter is 0.1-0.9, and be 0.05 to change with step-length.
Step 302:Improve the nonlinear combination identification model modeling of SVM
The biggest advantage that SVMs (SVM) has is exactly that can solve dimension disaster, local extremum, mistake
The situation that study and common predicting means result gap are crossed.LIBSVM algorithms, are directed to a kind of improvement of SVM algorithm, this
The advantage that improved SVM algorithm has drawn other algorithms is planted, optimization problem is turned into a typical double optimization problem, and make
It has analytic solutions, and the calculating speed for this kind of algorithm of method of common problem application numerical solution is more rapid, calculates
Precision it is higher and computing resource that take is less.
Very big, the core in the influence of the selection to the precision that predicts the outcome of valve pipe leakage rate of LIBSVM algorithm Kernel Functions
The selection result of function directly affects the result of calculation of algorithm.Kernel function includes linear kernel function, Polynomial kernel function, radial direction base
Kernel function.The present invention has carried out tentative calculation for these four kernel functions respectively, by the comparative analysis to result of calculation, final choice
Kernel function of the gaussian radial basis function as LIBSVM algorithms, its formula is:
After selected kernel function, the determination of the regular constant C in LIBSVM algorithms is also critically important, has weight to the realization of algorithm
The influence wanted, under normal circumstances, the value of C between 10 and 100, after the value of C exceeds 100, will also result in and owe showing for study
As the value that C is taken in the present invention is 10.
Linear regression is carried out using LIBSVM algorithms, using following algorithm to Lagrange multiplier αi、And threshold value b is carried out
It is determined that:
(1) with BP neural network identification model, RBF radial direction bases identification model, GRNN neural network recognizations model to pre- place
Sample big data is identified respectively between the valves leakage amount and characteristic parameter after reason, the predicted value of the various recognition methods for obtaining
And its actual value forms new test sample and training sample, as the sample of LIBSVM built-up patterns;
(2) SVM models are initialized, to Lagrange multiplier αi、And threshold value b carries out random assignment;
(3) new training sample is established as meeting the object function of SVM algorithm, it is asked using LIBSVM algorithm
Solution, obtains Lagrange multiplier αi、And the value of threshold value b;
(4) value of three parameters for obtaining is brought into object function, is calculated with test sample and identification is combined based on SVM
Valve leakage automatic inspection of the model under special characteristic parameter;
(5) compare with the actual leakage rate of valve, calculate error.When error is less than the precision for determining, then tie
Beam learning process, if not reaching determination precision, return to step (2) proceeds study.
After the completion of study, duct length, internal diameter, external diameter, ambient temperature and insulation are being obtained by accurate measurement
In the case of thickness degree, tube wall temperature before input channel entrance power pressure, entrance Temperature of Working, valve, using this
Identification model can just be identified and monitoring to the valve leakage automatic inspection under a range of characteristic parameter.
The present invention in the case of known heat distribution pipeline length, internal diameter of the pipeline, outer diameter tube, insulation layer thickness are isoparametric,
The parameter such as tube wall temperature is used as input before using entrance power pressure, entrance Temperature of Working, valve, by based on improvement
The nonlinear combination identification model of SVM is processed, and obtains valves leakage amount, and the method can let out according to special parameter to valve
Leakage quantity carries out rational judgment, and the field diagnostic to valve security and valves leakage is significant.
It is last it is to be noted that:The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations.To the greatest extent
Pipe has been described in detail to the present invention with reference to the foregoing embodiments, it will be understood by those within the art that:It is still
Technical scheme described in foregoing embodiments can be modified, or which part technical characteristic is equally replaced
Change;And these modifications or replacement, do not make the essence of the essence disengaging various embodiments of the present invention technical scheme of appropriate technical solution
God and scope.
Claims (7)
1. a kind of therrmodynamic system valves leakage quantity monitoring method, it is characterised in that the monitoring method is comprised the following steps:
(1) sample big data between valves leakage amount and characteristic parameter is obtained by valves leakage modeling;
(2) data prediction, be improve identification model study precision and efficiency, to valves leakage amount described in step (1) with
Sample big data carries out validation verification between characteristic parameter, and rejects wrong data data are normalized;
(3) based on the method for improving SVM nonlinear combination identification models, BP neural network identification model, RBF are radially recognized into mould
Type and GRNN neural network recognization models are combined, and calculate the therrmodynamic system valves leakage amount under special characteristic parameter, so that
Realize the monitoring to valves leakage amount.
2. therrmodynamic system valves leakage quantity monitoring method as claimed in claim 1, it is characterised in that:The step (1) includes:
Therrmodynamic system duct length, internal diameter, external diameter and insulation layer thickness are obtained by measuring, pipeline infinitesimal section radial direction side is then set up
To temperature field;Pipeline infinitesimal section axial direction temperature field is set up, along pipeline Working fluid flow direction, each infinitesimal section is counted paragraph by paragraph
Calculate, finally give valve preceding pipeline temperature field, each leakage rate one tube wall temperature of correspondence.
3. therrmodynamic system valves leakage quantity monitoring method as claimed in claim 2, it is characterised in that:By the therrmodynamic system pipe
Road is divided into multiple infinitesimal pipeline sections, and radial and axial modeling is carried out to the infinitesimal pipeline section according to heat transfer principle, and utilization changes
Solved along Working fluid flow direction for method;Friction loss and heat exchange amount according to N sections of infinitesimal pipeline section calculate N sections it is described micro-
First pipeline section outlet parameter, and in this, as the N+1 sections of intake condition of the infinitesimal pipeline section, until obtaining last before valve
The wall surface temperature of individual infinitesimal pipeline section, it is established that the large sample model between temperature field and characteristic parameter.
4. therrmodynamic system valves leakage quantity monitoring method as claimed in claim 1, it is characterised in that:It is described based on improve SVM
The step of method of nonlinear combination identification model (3), includes step in detail below:
Step A:With the BP neural network identification model, RBF radial direction base identification models and GRNN neural network recognization models pair
Sample big data is identified respectively between the valves leakage amount and characteristic parameter of pretreatment in step (2), various by what is acquired
The discre value of recognition methods forms new test sample and training sample with actual value, as improvement SVM nonlinear combinations
The sample of model;
Step B:Initialization SVM models, to Lagrange multiplier αi、And threshold value b carries out random assignment;
Step C:New training sample is established as meeting the object function of SVM algorithm, it is asked using LIBSVM algorithms
Solution, obtains Lagrange multiplier αi、And the value of threshold value b;
Step D:The Lagrange multiplier α that will be calculated by step Ci、And threshold value b values are brought into object function, use test specimens
This calculating is based on improving valve leakage automatic inspection of the SVM nonlinear combinations identification model under special characteristic parameter;
Step E:The valve leakage automatic inspection calculated by step D and the actual leakage rate of valve are compared, calculation error, when by mistake
When difference is less than the precision for determining, then terminate learning process, if not reaching determination precision, return to step B proceeds study;
Step F:Calculated after special characteristic parameter is input into using the combination identification model after the completion of study, drawn heating power system
System valves leakage amount, realizes the monitoring to valves leakage amount.
5. therrmodynamic system valves leakage quantity monitoring method as claimed in claim 1, it is characterised in that:In the step (3)
The BP neural network identification model is that, comprising 1 layer of Three Tiered Network Architecture of hidden layer, input layer number is 3, described
Input layer is respectively therrmodynamic system entrance power pressure, entrance Temperature of Working and tube wall temperature before valve;Output
Layer neuron is valve leakage automatic inspection;The quantity of intermediate layer neuron uses progressively experimental method, with training sample root-mean-square error most
Small to determine for target, the quantity of the intermediate layer neuron is 12;Hidden layer neuron transmission function uses S type tangent letters
Number, output layer neural transferring function uses S type logarithmic functions, training algorithm to use Levenberg-Marquardt algorithms.
6. therrmodynamic system valves leakage quantity monitoring method as claimed in claim 1, it is characterised in that:In the step (3)
The RBF radial directions identification model uses input layer, hidden layer, the three-decker of output layer, and input layer quantity is 3,
Input layer is respectively therrmodynamic system entrance power pressure, entrance Temperature of Working and tube wall temperature before valve;Output
Layer neuron is valve leakage automatic inspection;For hidden layer Gaussian function center, chosen using Orthogonal Least Square, and apply
Least square method is trained to network output weights.
7. therrmodynamic system valves leakage quantity monitoring method as claimed in claim 1, it is characterised in that:Institute in the step (3)
State the valves leakage described in the step (1) of the connection weight between the structure and neuron of GRNN neural network recognization models
Sample big data determines between amount and characteristic parameter, and the smoothing parameter of the GRNN neural network recognizations model is carried out really by test method(s)
Fixed, according to sample characteristics, selection smoothing parameter is 0.1-0.9, and is 0.05 to change with step-length.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611019372.4A CN106706215B (en) | 2016-11-17 | 2016-11-17 | A kind of therrmodynamic system valves leakage quantity monitoring method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611019372.4A CN106706215B (en) | 2016-11-17 | 2016-11-17 | A kind of therrmodynamic system valves leakage quantity monitoring method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106706215A true CN106706215A (en) | 2017-05-24 |
CN106706215B CN106706215B (en) | 2019-02-05 |
Family
ID=58939983
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611019372.4A Expired - Fee Related CN106706215B (en) | 2016-11-17 | 2016-11-17 | A kind of therrmodynamic system valves leakage quantity monitoring method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106706215B (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107420628A (en) * | 2017-06-21 | 2017-12-01 | 中国华能集团清洁能源技术研究院有限公司 | Leakage monitoring device and monitoring method in a kind of steam valve |
CN108120570A (en) * | 2017-12-05 | 2018-06-05 | 浙江工业大学 | A kind of pneumatic control valve air-tightness method for diagnosing faults based on SVM |
CN108388685A (en) * | 2017-12-28 | 2018-08-10 | 中国石油天然气股份有限公司 | The prediction technique and device of leakage current amount in oil-gas pipeline |
CN109538944A (en) * | 2018-12-03 | 2019-03-29 | 北京无线电计量测试研究所 | A kind of pipeline leakage detection method |
CN109900436A (en) * | 2017-12-08 | 2019-06-18 | 中国石油化工股份有限公司 | Valves leakage in-circuit diagnostic system and method based on cloud computing |
CN110145695A (en) * | 2019-06-03 | 2019-08-20 | 大连理工大学 | A kind of hot duct leakage detection method based on the fusion of the depth confidence network information |
CN110705797A (en) * | 2019-10-09 | 2020-01-17 | 浙江海洋大学 | Ship oil consumption data prediction method based on ship sensor network |
CN111412959A (en) * | 2020-04-29 | 2020-07-14 | 长江水利委员会水文局 | Flow online monitoring calculation method, monitor and monitoring system |
CN111992026A (en) * | 2020-07-20 | 2020-11-27 | 张家港市锦明机械有限公司 | Denitration double-reactor arrangement method and system for realizing online switching |
CN112113719A (en) * | 2020-09-21 | 2020-12-22 | 中国人民解放军海军工程大学 | Hydraulic slide valve internal leakage detection method based on acoustic emission technology |
CN112924115A (en) * | 2021-03-16 | 2021-06-08 | 中电华创(苏州)电力技术研究有限公司 | Device and method for monitoring internal leakage of high-temperature and high-pressure pipeline valve |
CN113090959A (en) * | 2021-03-31 | 2021-07-09 | 杨大松 | Wisdom gas monitoring system |
CN113280983A (en) * | 2021-04-21 | 2021-08-20 | 浙江工业大学 | On-line diagnosis method for internal leakage of pneumatic regulating valve |
CN114414175A (en) * | 2022-01-17 | 2022-04-29 | 山东电力工程咨询院有限公司 | Thermodynamic system drain valve inner leakage detection method and system |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3938544A (en) * | 1974-05-28 | 1976-02-17 | Bernaerts Henry J | Valve with heat transport monitor for leakage determining rate |
CN101008992A (en) * | 2006-12-30 | 2007-08-01 | 北京市劳动保护科学研究所 | Method for detecting leakage of pipeline based on artificial neural network |
CN101592288A (en) * | 2009-05-22 | 2009-12-02 | 重庆大学 | A kind of method for identifying pipeline leakage |
JP2013257256A (en) * | 2012-06-14 | 2013-12-26 | Chugoku Electric Power Co Inc:The | Air leakage confirmation device of compressed air generator |
CN104061445A (en) * | 2014-07-09 | 2014-09-24 | 中国石油大学(华东) | Pipeline leakage detection method based on neural network |
CN104121014A (en) * | 2014-06-16 | 2014-10-29 | 西南石油大学 | Method for diagnosing type of leakage of drilled well based on neural network fusion technique |
CN104502024A (en) * | 2014-12-19 | 2015-04-08 | 中国石油天然气股份有限公司 | Method and device for quantization regression prediction of valve internal leakage flow rate |
CN104654024A (en) * | 2015-02-12 | 2015-05-27 | 常州大学 | Method for locating and analyzing leakage of city gas pipeline based on GRNN (Generalized Regression Neural Network) |
CN105864643A (en) * | 2016-03-24 | 2016-08-17 | 华北电力大学 | Gas pipeline leakage positioning experimental device and method based on RBF neural network |
-
2016
- 2016-11-17 CN CN201611019372.4A patent/CN106706215B/en not_active Expired - Fee Related
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3938544A (en) * | 1974-05-28 | 1976-02-17 | Bernaerts Henry J | Valve with heat transport monitor for leakage determining rate |
CN101008992A (en) * | 2006-12-30 | 2007-08-01 | 北京市劳动保护科学研究所 | Method for detecting leakage of pipeline based on artificial neural network |
CN101592288A (en) * | 2009-05-22 | 2009-12-02 | 重庆大学 | A kind of method for identifying pipeline leakage |
JP2013257256A (en) * | 2012-06-14 | 2013-12-26 | Chugoku Electric Power Co Inc:The | Air leakage confirmation device of compressed air generator |
CN104121014A (en) * | 2014-06-16 | 2014-10-29 | 西南石油大学 | Method for diagnosing type of leakage of drilled well based on neural network fusion technique |
CN104061445A (en) * | 2014-07-09 | 2014-09-24 | 中国石油大学(华东) | Pipeline leakage detection method based on neural network |
CN104502024A (en) * | 2014-12-19 | 2015-04-08 | 中国石油天然气股份有限公司 | Method and device for quantization regression prediction of valve internal leakage flow rate |
CN104654024A (en) * | 2015-02-12 | 2015-05-27 | 常州大学 | Method for locating and analyzing leakage of city gas pipeline based on GRNN (Generalized Regression Neural Network) |
CN105864643A (en) * | 2016-03-24 | 2016-08-17 | 华北电力大学 | Gas pipeline leakage positioning experimental device and method based on RBF neural network |
Non-Patent Citations (2)
Title |
---|
吴丁飞 等: ""电厂阀门内漏在线监测研究"", 《能源研究与信息》 * |
薛明华: ""热力系统阀门内漏在线监测研究"", 《第十一届长三角科技论坛电机电力科技分论坛》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107420628A (en) * | 2017-06-21 | 2017-12-01 | 中国华能集团清洁能源技术研究院有限公司 | Leakage monitoring device and monitoring method in a kind of steam valve |
CN108120570A (en) * | 2017-12-05 | 2018-06-05 | 浙江工业大学 | A kind of pneumatic control valve air-tightness method for diagnosing faults based on SVM |
CN109900436A (en) * | 2017-12-08 | 2019-06-18 | 中国石油化工股份有限公司 | Valves leakage in-circuit diagnostic system and method based on cloud computing |
CN108388685A (en) * | 2017-12-28 | 2018-08-10 | 中国石油天然气股份有限公司 | The prediction technique and device of leakage current amount in oil-gas pipeline |
CN109538944A (en) * | 2018-12-03 | 2019-03-29 | 北京无线电计量测试研究所 | A kind of pipeline leakage detection method |
CN109538944B (en) * | 2018-12-03 | 2020-07-07 | 北京无线电计量测试研究所 | Pipeline leakage detection method |
CN110145695A (en) * | 2019-06-03 | 2019-08-20 | 大连理工大学 | A kind of hot duct leakage detection method based on the fusion of the depth confidence network information |
CN110705797B (en) * | 2019-10-09 | 2023-09-22 | 浙江海洋大学 | Ship fuel consumption data prediction method based on ship sensing network |
CN110705797A (en) * | 2019-10-09 | 2020-01-17 | 浙江海洋大学 | Ship oil consumption data prediction method based on ship sensor network |
CN111412959A (en) * | 2020-04-29 | 2020-07-14 | 长江水利委员会水文局 | Flow online monitoring calculation method, monitor and monitoring system |
CN111992026A (en) * | 2020-07-20 | 2020-11-27 | 张家港市锦明机械有限公司 | Denitration double-reactor arrangement method and system for realizing online switching |
CN112113719A (en) * | 2020-09-21 | 2020-12-22 | 中国人民解放军海军工程大学 | Hydraulic slide valve internal leakage detection method based on acoustic emission technology |
CN112924115A (en) * | 2021-03-16 | 2021-06-08 | 中电华创(苏州)电力技术研究有限公司 | Device and method for monitoring internal leakage of high-temperature and high-pressure pipeline valve |
CN113090959A (en) * | 2021-03-31 | 2021-07-09 | 杨大松 | Wisdom gas monitoring system |
CN113280983A (en) * | 2021-04-21 | 2021-08-20 | 浙江工业大学 | On-line diagnosis method for internal leakage of pneumatic regulating valve |
CN114414175A (en) * | 2022-01-17 | 2022-04-29 | 山东电力工程咨询院有限公司 | Thermodynamic system drain valve inner leakage detection method and system |
Also Published As
Publication number | Publication date |
---|---|
CN106706215B (en) | 2019-02-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106706215B (en) | A kind of therrmodynamic system valves leakage quantity monitoring method | |
Liu et al. | Numerical simulation and simplified calculation method for heat exchange performance of dry air cooler in natural gas pipeline compressor station | |
Esen et al. | Performance prediction of a ground-coupled heat pump system using artificial neural networks | |
CN110285532A (en) | Method for controlling machine room air conditioner, apparatus and system based on artificial intelligence | |
CN102419827B (en) | Radial basis function (RBF) neural network-based boiling heat exchanging prediction method | |
CN109578816B (en) | Water supply pipe network pipe explosion detection method and device, control method and device | |
Zendehboudi et al. | Utilization of ANN and ANFIS models to predict variable speed scroll compressor with vapor injection | |
CN103994334B (en) | Pipeline Leak flow estimation unit based on KPCA-RBF curve matching and method | |
CN113919243B (en) | Supercritical fluid heat transfer associated agent model construction method based on machine learning | |
CN105184395A (en) | Initial parameter determination method for thermal power generating unit comprising afterheat utilizing system | |
Li et al. | Function chain neural network prediction on heat transfer performance of oscillating heat pipe based on grey relational analysis | |
CN103198210A (en) | Method and equipment for detecting and evaluating inner drag-reducing coating of natural gas pipeline | |
CN107491840B (en) | Flow wear characteristic prediction and service life evaluation method based on ELM neural network model | |
CN110362893A (en) | A kind of heat supply pipeline heat preservation property and hydrophobic monitoring diagnosis system based on Internet of Things | |
CN104500979A (en) | Online diagnosis and elimination system for steam flowing retention section in heating pipe network and working method | |
CN106936627A (en) | A kind of thermal power generating equipment performance monitoring method based on big data analysis mining | |
CN108021742A (en) | A kind of steam heating pipeline steady-state operating condition method of estimation for considering hydrophobic model | |
CN112884012A (en) | Building energy consumption prediction method based on support vector machine principle | |
CN114202111A (en) | Electronic expansion valve flow characteristic prediction based on particle swarm optimization BP neural network | |
CN113486950B (en) | Intelligent pipe network water leakage detection method and system | |
Guo et al. | A fault detection method for heat loss in a tyre vulcanization workshop using a dynamic energy consumption model and predictive baselines | |
CN114065450A (en) | Dispatching optimization method of central heating system based on 5G technology | |
CN111079920B (en) | Prediction method for flow non-uniformity coefficient of turbine gas collection cavity outlet | |
CN110852615A (en) | Comprehensive reliability evaluation model for intelligent electric energy meter in typical environment | |
CN111412391A (en) | Pipe network leakage detection method and system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190205 Termination date: 20201117 |
|
CF01 | Termination of patent right due to non-payment of annual fee |