CN112765883A - Method for determining valve closing process based on genetic algorithm and neural network - Google Patents

Method for determining valve closing process based on genetic algorithm and neural network Download PDF

Info

Publication number
CN112765883A
CN112765883A CN202110061921.9A CN202110061921A CN112765883A CN 112765883 A CN112765883 A CN 112765883A CN 202110061921 A CN202110061921 A CN 202110061921A CN 112765883 A CN112765883 A CN 112765883A
Authority
CN
China
Prior art keywords
pipeline
valve
closing
neural network
closing process
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
Application number
CN202110061921.9A
Other languages
Chinese (zh)
Inventor
蒋丹
刘保生
刘渊铭
颜信
周兵源
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN202110061921.9A priority Critical patent/CN112765883A/en
Publication of CN112765883A publication Critical patent/CN112765883A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/08—Fluids
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/14—Pipes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Evolutionary Biology (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Fluid Mechanics (AREA)
  • Algebra (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Physiology (AREA)
  • Genetics & Genomics (AREA)
  • Pipeline Systems (AREA)

Abstract

The invention discloses a method for determining a valve closing process based on a genetic algorithm and a neural network, and belongs to the technical field of hydraulic pipeline transient flow. In order to comprehensively consider parameters such as the closing time of the valve, the number of closing stages, the closing amount of each stage, the closing time corresponding to each stage and the like, the method adopts the genetic algorithm to optimize the closing process of the valve, so that the influence of each parameter on the closing process of the valve can be comprehensively considered, and the optimal solution under the current condition can be obtained more favorably. Finally, training a neural network on the basis of data obtained by a genetic algorithm, wherein the trained neural network can obtain the optimal valve closing process under given conditions without directly depending on the genetic algorithm; the optimal closing process of the valve can be rapidly calculated under various environments, and the water hammer effect in the closing process of the valve is reduced.

Description

Method for determining valve closing process based on genetic algorithm and neural network
Technical Field
The invention belongs to the technical field of hydraulic pipeline transient flow, and relates to application of a genetic algorithm and an artificial neural network in seeking an optimal valve closing process in a hydraulic pipeline.
Background
When the fluid in the hydraulic pipeline has a rapid change in flow rate due to the closing of a valve or the like, the fluid pressure in the pipeline is rapidly changed due to the inertia of the fluid, which is called hydraulic transient flow and also called water hammer. Transient flow can generate huge pressure impact on the valve and the inner wall of the pipeline, the pressure can reach dozens of times of the normal operation of the pipeline sometimes, and if the transient flow is not controlled, serious damage can be caused to the pipeline. Studies have shown that the pressure surge created by the transient flow can be reduced to a large extent with an appropriate valve closing procedure. However, it is not easy to select a proper valve closing process because the determination of the valve closing process involves many parameters, such as the closing time of the valve, the closing stages, the amount of closing in each stage, and the corresponding closing time in each stage. The current research determines the valve closing process on the basis of fixed valve closing time, and the valve closing process obtained by the method is not the optimal valve closing process because the closing time is not treated as a variable. It is therefore necessary to select a method which is efficient and which allows a combined consideration of the individual parameters in order to determine an optimum closing process of the valve.
Disclosure of Invention
The invention solves the technical problem that the water hammer effect cannot be further reduced when the valve is closed because the closing time of the valve is not considered in the prior art, and improves and designs a method for calculating the closing process of the valve.
The technical scheme of the invention is a method for determining the valve closing process based on a genetic algorithm and a neural network, which comprises the following steps:
step 1, establishing a one-dimensional transient flow mathematical model of a hydraulic pipeline;
the one-dimensional transient flow mathematical model of the hydraulic pipeline consists of a continuity equation and a motion equation, and respectively reflects the flow velocity of unstable water flow and the change rule of a water head in the transient flow process;
the continuity equation is:
Figure BDA0002903015000000011
the equation of motion is:
Figure BDA0002903015000000012
in the formula: v is the flow velocity of fluid inside the pipeline, H is the water head of the pipeline, g is the gravity acceleration, f is the Darcy-Weisbaha friction coefficient, D is the diameter of the pipeline, a is the propagation velocity of the water hammer wave, alpha is the included angle between the pipeline and the horizontal plane, x is the axial distance along the pipeline, and t is the time;
step 2, establishing a characteristic line equation of the transient flow of the hydraulic pipeline according to the model obtained in the step 1;
Figure BDA0002903015000000021
Figure BDA0002903015000000022
step 3, designing a genetic algorithm by taking the total closing time of the valve, the number of closing stages, the closing amount of each stage and the closing time corresponding to each stage as parameters needing optimization and aiming at minimizing the maximum water hammer pressure;
step 4, obtaining the optimal valve closing process under the current condition by using a genetic algorithm;
changing pipeline parameters and optimizing by using a genetic algorithm again;
step 6, training an artificial neural network by using the obtained data;
step 7, inputting corresponding parameters in the neural network during actual calculation to obtain the optimal valve closing process under corresponding conditions;
and 8, verifying the valve closing rule on a pipeline test bed.
Further, the specific method of step 3 is as follows: obtaining the closing process of the valve by considering the total closing time, the number of closing stages, the closing amount of each stage and the corresponding closing time of each stage of the valve;
dividing the whole closing process of the valve into N stages, and taking the time when each stage is ended as t0，t1，t2，…，tNAt the same time, the corresponding opening degree of the valve at the end of each stage is s0，s1，s2，…，sNSince the valve is fully open at the start time and fully closed at the end time, there are:
t0＝0，s0＝1
tN＝Tc，sN＝0
Tctotal valve closure time;
the specific design method of the genetic algorithm is as follows:
step 3.1: determining decision variables and value ranges of the variables; 1<N≤9，2L/a≤Tc≤30L/a，0<t1<t2<t3…<tN-1<Tc，0<si<1, wherein N is an integer, L represents the length of the pipeline, a represents the water hammer wave propagation speed, i is 1,2,3, …, N-1;
step 3.2: determining a target function by using the characteristic line equation of the transient flow of the hydraulic pipeline in the step 2; under the condition that the initial flow and pressure are determined, the flow and pressure of any point in the pipeline when the water hammer occurs can be obtained by utilizing a characteristic line equation, and the maximum value of the pressure in the pipeline, namely the maximum water hammer pressure is takenpmaxIs an objective function;
step 3.3: determining encoding and decoding methods of the chromosome;
step 3.4: determining a specific operation method of a genetic operator;
step 3.5: determining a fitness function Ft；Ft＝K/pmaxWherein K is a constant, pmaxIs the maximum water hammer pressure in the pipeline;
step 3.6: and determining the initial population size, the cross probability, the mutation probability and the maximum iteration number.
Further, the specific method of step 4 is as follows: obtaining the optimal valve closing process under the current condition by utilizing the genetic algorithm determined in the step 3;
the basic process of the genetic algorithm is as follows:
step 4.1, randomly generating M individuals as an initial population P (t), coding the initial population, setting the maximum evolution algebra as I, and simultaneously resetting a counter I of the evolution algebra to zero;
4.2, calculating the function fitness of each individual in the initial population;
4.3, selecting individuals in P (t) by using a selection operator;
4.4, performing cross operation on the selected group by using a cross operator;
step 4.5, acting the mutation operator on the crossed population to obtain the next generation of population P (t + 1);
and 4.6, judging the termination condition, if I < ═ I, assigning the value of I to I +1, and transferring to the step 4.2, if I > I: the iteration is terminated and the optimal solution under the current conditions is output.
Further, the step 6 trains an artificial neural network by using the obtained data to obtain an optimal valve closing process under corresponding conditions;
the specific process of predicting parameters using neural networks is as follows:
step 6.1: collecting training data, and utilizing the optimal valve closing process data under different pipeline conditions obtained in the step 5, wherein the density of the fluid, the viscosity of the fluid, the bulk modulus of the fluid, the length of the pipeline, the diameter of the pipeline and the initial flow rate of the pipeline are input of a network, and the closing time, the closing stage and the closing amount of each stage of the valve and the closing time corresponding to each stage are corresponding labels; dividing the data into two parts, wherein 80% of the data are used as a training set for training a neural network, and 20% of the data are used as a test set for testing the accuracy of the neural network;
step 6.2: determining a network structure, wherein the network comprises an input layer, two hidden layers and an output layer, all the layers are connected, the number of nodes of the input layer is 6, the number of nodes of the two hidden layers is 1000, the number of nodes of the output layer is 6, and an activation function of the hidden layers is Relu;
step 6.3: initializing network weight, wherein a He initialization method is selected for weight initialization because an activation function of a hidden layer is a Relu function;
step 6.4: training the network, inputting input into the network by using a training set in the collected training data, outputting the input by forward propagation calculation, comparing the input with label to calculate a loss function, updating the weight by backward propagation, and repeating the process until the loss function value reaches an expected value;
step 6.5: testing the accuracy of the neural network, using the neural network on a training data test set, testing whether the output of the network meets expectations, if not, retraining until meeting, and storing the network parameters;
step 6.6: and inputting the density of the relevant fluid, the viscosity of the fluid, the bulk modulus of elasticity of the fluid, the length of the pipeline, the diameter of the pipeline and the initial flow rate parameter of the pipeline by using the trained neural network, predicting the closing time, the closing stages, the closing amount of each stage and the closing time corresponding to each stage of the valve, and obtaining the optimal valve closing rule under any condition.
The step 6 describes a process of obtaining an optimal valve closing process under any condition by using a neural network, wherein the neural network is an important component of the patent and is responsible for obtaining an optimal valve closing rule by using condition parameters of fluid and pipelines.
In order to comprehensively consider parameters such as the closing time of the valve, the number of closing stages, the closing amount of each stage, the closing time corresponding to each stage and the like, the method adopts the genetic algorithm to optimize the closing process of the valve, so that the influence of each parameter on the closing process of the valve can be comprehensively considered, and the optimal solution under the current condition can be obtained more favorably. Finally, training a neural network on the basis of data obtained by a genetic algorithm, wherein the trained neural network can obtain the optimal valve closing process under given conditions without directly depending on the genetic algorithm; the optimal closing process of the valve can be rapidly calculated under various environments, and the water hammer effect in the closing process of the valve is reduced.
Drawings
FIG. 1 is a flow chart of the present invention.
FIG. 2 is a schematic diagram of a characteristic line method according to the present invention.
Fig. 3 is a simple hydraulic circuit model of the present invention.
Fig. 4 is a graph of pressure fluctuations as the valve of the present invention is momentarily closed.
Fig. 5 illustrates the optimal valve closing process of the present invention.
FIG. 6 is a comparison of pressure fluctuation curves for the present invention.
FIG. 7 is a diagram of a neural network according to the present invention.
FIG. 8 is a schematic view of a pipeline testing platform according to the present invention.
Detailed Description
The process of finding the optimal valve closing process in the hydraulic line using a genetic algorithm is described in further detail below with reference to the accompanying drawings.
First, a simple hydraulic line model as shown in fig. 3 was created, and the specific parameters of the model are shown in table 1.
TABLE 1 concrete parameters of simple hydraulic pipeline model
Length L of pipeline 120m
Fluid density ρ 1000kg/m3
Initial flow velocity v0 0.65m/s
Head pressure H0 38.75m
Coefficient of friction f 0.25
Velocity a of propagation of pressure wave0 1200m/s
Secondly, a mathematical model of transient flow of the hydraulic pipeline is established, and then the hydraulic transient process of the hydraulic pipeline is simulated on the basis of a characteristic line equation, so that a pressure fluctuation curve when the valve is instantaneously closed can be obtained as shown in fig. 4.
And finally, designing a genetic algorithm by comprehensively considering parameters such as the closing time of the valve, the number of closing stages, the closing amount of each stage, the closing time corresponding to each stage and the like. As shown in table 2, relevant parameters and value ranges thereof need to be considered when designing a genetic algorithm for the hydraulic pipeline model shown in fig. 3.
TABLE 2 parameters relating to valve closure procedure
Parameter(s) Value range
N
2≤N≤8,N∈Z
Tc 0.2s≤Tc≤3s
t1，t2，…，t N-1 0<t1<t2<t3…<tN-1<Tc
s1，s2，…，sN-1 0.9<s1<1,0<sN-1<0.1,0<si<1(i＝2,3,…,N-2)
Therefore, the genetic algorithm is used to find the optimal valve closing process, which takes into account a total of 2N parameters, N, Tc, t1，t2，…，tN-1And s1，s2，…，sN-1。
In addition, the fitness function of the invention is Ft＝8×107/pmaxWherein p ismaxIs the maximum water hammer pressure in the pipeline and is also the objective function. The number of initial populations is 80, and the maximum evolution generation number is 100.
And (3) operating a genetic algorithm to finally obtain related parameter values of the optimal valve closing process, wherein Tc＝1.7067s，N＝5，t1＝0.4072s，s1＝0.9983，t2＝0.7010s，s2＝0.5363，t3＝0.8812s，s3＝0.7671，t4＝1.4637s，s4＝0.099. In addition, it should be noted that the genetic algorithm is used to obtain only specific point values, and the point values need to be converted into corresponding closing curves, and at this time, MATLAB is used to process the data; by using interpolation, these points are connected to finally obtain the optimal valve closing process and the corresponding pressure fluctuation curve under the current conditions, as shown in fig. 5 and fig. 6, respectively.
The optimal valve closing process under one pipeline parameter is obtained, the pipeline parameter needs to be changed in order to obtain the optimal valve closing process under different pipeline parameters, and then the simulation model and the genetic algorithm are operated again to carry out optimization. In this way, an optimal valve closing process can be obtained for different pipeline parameters. For different pipelines, in order to obtain the optimal valve closing process, a simulation and optimization model needs to be operated once, and because the operation of the genetic algorithm is time-consuming, the optimal valve closing process of the pipeline is further obtained by training an artificial neural network, as shown in fig. 7; and then training and specifically verifying the neural network according to the obtained data.
The invention carries out the verification of the valve closing process on a pipeline test bed:
in order to verify the effectiveness of the valve closing process, the invention builds a simple pipeline test platform and carries out relevant tests on the platform to carry out actual verification. As shown in fig. 8, the whole test platform can be divided into three parts, namely a pipeline system, a control system and a data acquisition system according to functions.
The pipeline system comprises a water tank, a water hammer pipeline, valves 1 and 2 and a water pump. A large amount of water is filled in the water tank, so that stable initial pressure can be provided for the pipeline; the water hammer pipeline is a red copper pipe which is not easy to rust and can bear large water hammer pressure; the valve 1 is a research object of the invention, and the aim of controlling the water hammer pressure in the pipeline can be achieved by controlling the valve 1 to be closed according to a certain rule.
And the control system comprises an air source, a swing air cylinder and a pneumatic servo valve. The air source provides power for the swing cylinder, the swing cylinder controls the valve 1 to move according to a certain rule, and the turning-on and turning-off of the pneumatic servo valve can control the rotating direction and speed of the swing cylinder so as to control the opening degree of the valve 1.
And the data acquisition system comprises a pressure sensor, a data acquisition board and a computer. The pressure sensor is used for collecting pressure fluctuation at the position of the valve 1, then the data collecting plate collects analog voltage signals on the pressure sensor, and finally a pressure fluctuation curve in a pipeline is presented on a computer.
When an actual experiment is carried out on a test platform, the valve 1 and the valve 2 are completely opened, at the moment, water in the water tank flows to the water pump from the water tank and then returns to the water tank through the hose, and a loop is formed. After the water flow in the pipeline is stable, if the valve 1 is controlled to be closed according to a certain rule by using the swing cylinder, the pressure fluctuation condition in the pipeline in the current closing process can be obtained through the pressure sensor and the data acquisition board.
Thus, an optimal valve closing process under the current conditions can be obtained.
Finally, comparing the pressure fluctuation curve in the optimal valve closing process with the pressure fluctuation curve in the valve instantaneous closing process, the optimal valve closing process obtained based on the genetic algorithm and the neural network can be seen, the water hammer pressure in the pipeline can be restrained to a great extent, and the method is feasible and practical.

Claims (4)

1. A method of determining a valve closing process based on a genetic algorithm and a neural network, the method comprising:
step 1, establishing a one-dimensional transient flow mathematical model of a hydraulic pipeline;
the one-dimensional transient flow mathematical model of the hydraulic pipeline consists of a continuity equation and a motion equation, and respectively reflects the flow velocity of unstable water flow and the change rule of a water head in the transient flow process;
the continuity equation is:
Figure FDA0002903014990000011
the equation of motion is:
Figure FDA0002903014990000012
in the formula: v is the flow velocity of fluid inside the pipeline, H is the water head of the pipeline, g is the gravity acceleration, f is the Darcy-Weisbaha friction coefficient, D is the diameter of the pipeline, a is the propagation velocity of the water hammer wave, alpha is the included angle between the pipeline and the horizontal plane, x is the axial distance along the pipeline, and t is the time;
step 2, establishing a characteristic line equation of the transient flow of the hydraulic pipeline according to the model obtained in the step 1;
Figure FDA0002903014990000013
Figure FDA0002903014990000014
step 3, designing a genetic algorithm by taking the total closing time of the valve, the number of closing stages, the closing amount of each stage and the closing time corresponding to each stage as parameters needing optimization and aiming at minimizing the maximum water hammer pressure;
step 4, obtaining the optimal valve closing process under the current condition by using a genetic algorithm;
changing pipeline parameters and optimizing by using a genetic algorithm again;
step 6, training an artificial neural network by using the obtained data;
step 7, inputting corresponding parameters in the neural network during actual calculation to obtain the optimal valve closing process under corresponding conditions;
and 8, verifying the valve closing rule on a pipeline test bed.
2. The method for determining the valve closing process based on the genetic algorithm and the neural network as claimed in claim 1, wherein the specific method of the step 3 is as follows: obtaining the closing process of the valve by considering the total closing time, the number of closing stages, the closing amount of each stage and the corresponding closing time of each stage of the valve;
dividing the whole closing process of the valve into N stages, and taking the time when each stage is ended as t0，t1，t2，…，tNAt the same time, the corresponding opening degree of the valve at the end of each stage is s0，s1，s2，…，sNSince the valve is fully open at the start time and fully closed at the end time, there are:
t0＝0，s0＝1
tN＝Tc，sN＝0
Tctotal valve closure time;
the specific design method of the genetic algorithm is as follows:
step 3.1: determining decision variables and value ranges of the variables; 1<N≤9，2L/a≤Tc≤30L/a，0<t1<t2<t3…<tN-1<Tc，0<si<1, wherein N is an integer, L represents the length of the pipeline, a represents the water hammer wave propagation speed, i is 1,2,3, …, N-1;
step 3.2: determining a target function by using the characteristic line equation of the transient flow of the hydraulic pipeline in the step 2; under the condition that the initial flow and pressure are determined, the flow and pressure of any point in the pipeline when the water hammer occurs can be obtained by utilizing a characteristic line equation, and the maximum value of the pressure in the pipeline, namely the maximum water hammer pressure p is takenmaxIs an objective function;
step 3.3: determining encoding and decoding methods of the chromosome;
step 3.4: determining a specific operation method of a genetic operator;
step 3.5: determining a fitness function Ft；Ft＝K/pmaxWherein K is a constant, pmaxIs the maximum water hammer pressure in the pipeline;
step 3.6: and determining the initial population size, the cross probability, the mutation probability and the maximum iteration number.
3. The method for determining the valve closing process based on the genetic algorithm and the neural network as claimed in claim 1, wherein the specific method of the step 4 is as follows: obtaining the optimal valve closing process under the current condition by utilizing the genetic algorithm determined in the step 3;
the basic process of the genetic algorithm is as follows:
step 4.1, randomly generating M individuals as an initial population P (t), coding the initial population, setting the maximum evolution algebra as I, and simultaneously resetting a counter I of the evolution algebra to zero;
4.2, calculating the function fitness of each individual in the initial population;
4.3, selecting individuals in P (t) by using a selection operator;
4.4, performing cross operation on the selected group by using a cross operator;
step 4.5, acting the mutation operator on the crossed population to obtain the next generation of population P (t + 1);
and 4.6, judging the termination condition, if I < ═ I, assigning the value of I to I +1, and transferring to the step 4.2, if I > I: the iteration is terminated and the optimal solution under the current conditions is output.
4. The method for determining the valve closing process based on the genetic algorithm and the neural network as claimed in claim 1, wherein the step 6 trains the artificial neural network by using the obtained data to obtain the optimal valve closing process under the corresponding conditions;
the specific process of predicting parameters using neural networks is as follows:
step 6.1: collecting training data, and utilizing the optimal valve closing process data under different pipeline conditions obtained in the step 5, wherein the density of the fluid, the viscosity of the fluid, the bulk modulus of the fluid, the length of the pipeline, the diameter of the pipeline and the initial flow rate of the pipeline are input of a network, and the closing time, the closing stage and the closing amount of each stage of the valve and the closing time corresponding to each stage are corresponding labels; dividing the data into two parts, wherein 80% of the data are used as a training set for training a neural network, and 20% of the data are used as a test set for testing the accuracy of the neural network;
step 6.2: determining a network structure, wherein the network comprises an input layer, two hidden layers and an output layer, all the layers are connected, the number of nodes of the input layer is 6, the number of nodes of the two hidden layers is 1000, the number of nodes of the output layer is 6, and an activation function of the hidden layers is Relu;
step 6.3: initializing network weight, wherein a He initialization method is selected for weight initialization because an activation function of a hidden layer is a Relu function;
step 6.4: training the network, inputting input into the network by using a training set in the collected training data, outputting the input by forward propagation calculation, comparing the input with label to calculate a loss function, updating the weight by backward propagation, and repeating the process until the loss function value reaches an expected value;
step 6.5: testing the accuracy of the neural network, using the neural network on a training data test set, testing whether the output of the network meets expectations, if not, retraining until meeting, and storing the network parameters;
step 6.6: and inputting the density of the relevant fluid, the viscosity of the fluid, the bulk modulus of elasticity of the fluid, the length of the pipeline, the diameter of the pipeline and the initial flow rate parameter of the pipeline by using the trained neural network, predicting the closing time, the closing stage, the closing amount of each stage and the closing time corresponding to each stage of the valve, and obtaining the optimal valve closing process under any condition.
CN202110061921.9A 2021-01-18 2021-01-18 Method for determining valve closing process based on genetic algorithm and neural network Pending CN112765883A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110061921.9A CN112765883A (en) 2021-01-18 2021-01-18 Method for determining valve closing process based on genetic algorithm and neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110061921.9A CN112765883A (en) 2021-01-18 2021-01-18 Method for determining valve closing process based on genetic algorithm and neural network

Publications (1)

Publication Number Publication Date
CN112765883A true CN112765883A (en) 2021-05-07

Family

ID=75702758

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110061921.9A Pending CN112765883A (en) 2021-01-18 2021-01-18 Method for determining valve closing process based on genetic algorithm and neural network

Country Status (1)

Country Link
CN (1) CN112765883A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113656985A (en) * 2021-09-01 2021-11-16 上海仁童电子科技有限公司 Modeling method, fault diagnosis method and device for valve elements in brake system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009157878A (en) * 2007-12-28 2009-07-16 Hitachi Ltd Fuzzy controller, plant system, and fuzzy control method
CN103971162A (en) * 2014-04-04 2014-08-06 华南理工大学 Method for improving BP (back propagation) neutral network and based on genetic algorithm
CN105302997A (en) * 2015-11-25 2016-02-03 河海大学 Liquid column separation-bridged water hammer simulation method based on three-dimensional CFD (Computational Fluid Dynamics)
CN109034388A (en) * 2018-07-27 2018-12-18 湖北中烟工业有限责任公司 A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network
CN111442799A (en) * 2020-03-26 2020-07-24 西安普特流体控制有限公司 Transient current monitoring method and multi-channel water hammer detector
CN111695306A (en) * 2020-05-21 2020-09-22 河海大学 Valve transient characteristic simulation method based on three-dimensional water hammer CFD

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009157878A (en) * 2007-12-28 2009-07-16 Hitachi Ltd Fuzzy controller, plant system, and fuzzy control method
CN103971162A (en) * 2014-04-04 2014-08-06 华南理工大学 Method for improving BP (back propagation) neutral network and based on genetic algorithm
CN105302997A (en) * 2015-11-25 2016-02-03 河海大学 Liquid column separation-bridged water hammer simulation method based on three-dimensional CFD (Computational Fluid Dynamics)
CN109034388A (en) * 2018-07-27 2018-12-18 湖北中烟工业有限责任公司 A kind of prediction model of cigarette material and mainstream smoke constituents based on Genetic Algorithm Optimized Neural Network
CN111442799A (en) * 2020-03-26 2020-07-24 西安普特流体控制有限公司 Transient current monitoring method and multi-channel water hammer detector
CN111695306A (en) * 2020-05-21 2020-09-22 河海大学 Valve transient characteristic simulation method based on three-dimensional water hammer CFD

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
DHANDAYUDHAPANI RAMALINGAM: "DESIGN AIDS FOR AIR VESSELS FOR TRANSIENT PROTECTION OF LARGE PIPE NETWORKS - A FRAMEWORK BASED ON PARAMETERIZATION OF KNOWLEDGEBASE DERIVED FROM OPTIMIZED NETWORK MODELS", 《UNIVERSITY OF KENTUCKY DOCTORAL DISSERTATIONS》, 31 December 2007 (2007-12-31), pages 24 *
HUI GAO等: "THE THEORETICAL STUDY AND PARAMETER OPTIMIZATION OF TWO-STAGE AIR VALVE FOR PREVENTION OF WATER HAMMER IN THE LONG SUPPLY PIPE", 《2014 ISFMFE - 6TH INTERNATIONAL SYMPOSIUM ON FLUID MACHINERY AND FLUID ENGINEERING》, 18 June 2015 (2015-06-18), pages 1 - 7 *
何相慧等: "不同体型调压室的瞬态水力特性对比分析", 《HTTPS://KNS.CNKI.NET/KCMS/DETAIL/11.1882.TV.20201127.1102.001.HTML》, 27 November 2020 (2020-11-27), pages 1 - 9 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113656985A (en) * 2021-09-01 2021-11-16 上海仁童电子科技有限公司 Modeling method, fault diagnosis method and device for valve elements in brake system
CN113656985B (en) * 2021-09-01 2024-03-15 上海仁童电子科技有限公司 Modeling method, fault diagnosis method and device for valve element in braking system

Similar Documents

Publication Publication Date Title
CN107725283B (en) A kind of fan trouble detection method based on deepness belief network model
US20210383041A1 (en) In-situ thermodynamic model training
CN105298734B (en) A kind of parameter identification method of Adaptive System of Water-Turbine Engine
Wang et al. A compact constraint incremental method for random weight networks and its application
CN110059345B (en) Method and system for optimizing successive load shedding shutdown rule of pumped storage unit
CN102135021B (en) Method for predicting shaft power of industrial extraction condensing steam turbine
Li et al. A hybrid assembly sequence planning approach based on discrete particle swarm optimization and evolutionary direction operation
CN112084727A (en) Transition prediction method based on neural network
CN114202111A (en) Electronic expansion valve flow characteristic prediction based on particle swarm optimization BP neural network
CN105138717A (en) Transformer state evaluation method by optimizing neural network with dynamic mutation particle swarm
CN112115544A (en) Rocket track optimization method based on improved genetic algorithm
CN106650934B (en) The method that thermal process state-space model parameter is accurately recognized using improved genetic Optimization Algorithm
CN103942434A (en) Hydropower station plant dam structure vibration response predication method based on SSPSO-GRNN
CN112765883A (en) Method for determining valve closing process based on genetic algorithm and neural network
CN111859746A (en) Method for predicting variable working condition performance of turbomachinery based on flow field reconstruction
CN114330119A (en) Deep learning-based pumped storage unit adjusting system identification method
JP5582487B2 (en) Process state prediction method
CN110033118A (en) Elastomeric network modeling and the blower multiobjective optimization control method based on genetic algorithm
CN111369072A (en) Nuclear minimum mean square time sequence online prediction model based on sparsification method
JP2001160043A (en) Simulation apparatus for water pipe network
Cui et al. Chaotic time series prediction based on binary particle swarm optimization
CN106126808B (en) A kind of pure water pipeline dynamic friction item parameter identification method based on genetic algorithm
CN111125867A (en) Establishment and calculation method of chemical production pipeline real-time transient model based on chaotic particle swarm
Gao et al. Parameter Identification of hydro generation system with fluid transients based on improved genetic algorithm
CN114167717B (en) Thermal power unit DEH rotating speed control method based on improved PSO-fuzzy PID

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: 20210507

RJ01 Rejection of invention patent application after publication