CN107357997A - A kind of analogy method and system of hydraulic pressure process for machining - Google Patents
A kind of analogy method and system of hydraulic pressure process for machining Download PDFInfo
- Publication number
- CN107357997A CN107357997A CN201710586939.4A CN201710586939A CN107357997A CN 107357997 A CN107357997 A CN 107357997A CN 201710586939 A CN201710586939 A CN 201710586939A CN 107357997 A CN107357997 A CN 107357997A
- Authority
- CN
- China
- Prior art keywords
- neural network
- basis function
- network
- function neural
- output
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B30—PRESSES
- B30B—PRESSES IN GENERAL
- B30B15/00—Details of, or accessories for, presses; Auxiliary measures in connection with pressing
- B30B15/16—Control arrangements for fluid-driven presses
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B30—PRESSES
- B30B—PRESSES IN GENERAL
- B30B15/00—Details of, or accessories for, presses; Auxiliary measures in connection with pressing
- B30B15/26—Programme control arrangements
-
- 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
- G06F2119/18—Manufacturability analysis or optimisation for manufacturability
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Control Of Presses (AREA)
Abstract
The invention discloses a kind of analogy method and system of hydraulic pressure process for machining.Method includes:The radial basis function neural network for each technical process established in hydraulic pressure machining process;The input vector and output vector of hydraulic press process are gathered, obtains training sample, and initial parameter is set, the radial basis function neural network is trained.The method and system of the present invention regards hydraulic pressure machining process as a nonlinear and time-varying system, radial basis function neural network is established to each technical process in process, and the input and output vector by gathering actual condition obtains training sample, radial basis function neural network is trained by training sample, improves the simulation precision of the processing technology of hydraulic press.
Description
Technical field
The present invention relates to hydraulic press intelligent control field, the analogy method of more particularly to a kind of hydraulic pressure machining process and it is
System.
Background technology
Hydraulic press is a kind of important part machining tool, is widely used in each metalloid and nonmetallic shapes
In.Hydraulic press realizes the transmission of energy using liquid as working media, and shape is brought it about by applying powerful pressure to blank
Become to complete its forming technique.The control accuracy of hydraulic press will directly influence the quality and yield rate of part production and processing
Etc. index.At present, most of hydraulic press control mode is obtained according to a large amount of knowhow induction and conclusions, is not established special
Fixed system model, this method are difficult to be extended and promote, and the customization flexible production pattern that will be difficult in adapt to future.
The accurate model of hydraulic press is established, can be to optimize its control process to provide fundamental basis, it is significant.
The model of hydraulic press is established, a solution is the side using the identification of modelling by mechanism combination Linear Time-Invariant System
Method.Modelling by mechanism, i.e., the hydraulic press course of work is described using various math equations, specifically according to the intrinsic parameter of hydraulic press
(hydraulic cylinder sectional area, hydraulic pump discharge, charge oil pressure etc.), using mechanics principle carry out a series of complicated formulas derive and
Simplify, what is obtained is the second-order linearity model of system.However, because the work operating mode of hydraulic press is with respect to other machining equipments
It is more complicated and changeable, it is complicated nonlinear and time-varying system, passes through the system model of traditional Mathematical Modeling Methods acquisition and its
The course of work has very big error under actual condition.
The content of the invention
The object of the present invention is in order to improve the simulation precision of the processing technology of hydraulic press, there is provided a kind of hydraulic pressure machining
The analogy method and system of technique.
To achieve the above object, the invention provides following scheme:
A kind of analogy method of hydraulic pressure process for machining, comprises the following steps:
The radial basis function neural network for each technical process established in hydraulic pressure machining process;
The input vector and output vector of each technical process of hydraulic press process are gathered, obtains training sample;
According to the training sample, initial parameter is set, the radial basis function neural network is trained, until every
The radial basis function neural network of individual technical process and the actual work of the technical process corresponding to the radial basis function neural network
Condition error is less than setting value, obtains hydraulic press process model;
Vector is actually entered according to hydraulic pressure machining process, using the hydraulic press process model, described in progress
The simulation of hydraulic pressure process for machining.
Optionally, the radial basis function neural network includes input layer, hidden layer and output layer, described to establish hydraulic press
The radial basis function neural network of each technical process in process, is specifically included:
Using formula x=[y (k-1) ..., y (k-ny);u(k),u(k-1),…,u(k-nu);j;K] calculate input layer
Network inputs;Wherein, y (k), u (k) are respectively that kth time samples obtained output vector, input vector, nyAnd nuRespectively export
With the maximum delay of input;J represents hydraulic press system in j-th of technical process;
UsingCalculate the excitation function of hidden layer;Wherein,
cmFor the central point vector of m-th RBF, G is Gauss functions, σmFor the characteristic parameter of m-th of RBF;M is
Hidden layer neuron number;
Using formula b=[yM] and formula (k)Calculate the network output of output layer;Wherein,
wmThe weight between hidden layer and output layer unit.
Optionally, it is described according to the training sample, initial parameter is set, the radial basis function neural network is carried out
Training, until radial basis function neural network and the technique mistake corresponding to the radial basis function neural network of each technical process
The actual condition error of journey is less than setting value, specifically includes:
Step 1, the initial value of the parameter in the radial basis function neural network is set;The parameter include central point to
Measure cm, characteristic parameter σmThe weight w between hidden layer and output layer unitm;
Step 2, the input vector in the training sample and output vector determine the radial ba-sis function network
The network inputs of network;
Step 3, exported according to the network inputs calculating network;
Step 4, the output vector in the training sample and the error of network output are calculated, judges that the error is
It is no to be less than the setting value, if it is not, step 5 is then performed, if so, then terminating to train;
Step 5, based on learning algorithm, the parameter in radial basis function neural network, return to step 2 are adjusted.
Optionally, the parameter for being based on learning algorithm, adjusting in radial basis function neural network, is specifically included:
The distance between center according to each RBF determines the center of each RBF and other radial direction base letters
The minimum range at number center extends constant as corresponding to each RBF;
The mean square error of the error exported according to the output vector and the network, calculating network output and output vector
Cost function as error gradient;
According to the extension constant and the cost function, institute is adjusted with the learning rate set in the negative direction of error gradient
State the parameter in radial basis function neural network.
A kind of simulation system of hydraulic pressure process for machining, including radial basis function neural network establish module, data acquisition
Module, radial basis function neural network training module, hydraulic pressure process for machining analog module;
Institute's radial basis function neural network establishes module, for establishing each technical process in hydraulic pressure machining process
Radial basis function neural network;
The data acquisition module, input vector and output for each technical process for gathering hydraulic press process
Vector, obtain training sample;
The radial basis function neural network training module, for according to the training sample, initial parameter being set, to institute
State in radial basis function neural network and be trained, until radial basis function neural network and the radial direction base of each technical process
The actual condition error of technical process corresponding to Function Neural Network is less than setting value, obtains hydraulic press process model.
The hydraulic pressure process for machining analog module, for the vector that actually enters according to hydraulic pressure machining process, use
The hydraulic press process model, carry out the simulation of hydraulic pressure process for machining.
Optionally, the radial basis function neural network establishes module, specifically includes:
Input layer setting up submodule, for using formula x=[y (k-1) ..., y (k-ny);u(k),u(k-1),…,u
(k-nu);j;K] calculate input layer network inputs;Wherein, y (k), u (k) be respectively kth time sampling obtain output vector,
Input vector, nyAnd nuThe maximum delay for respectively exporting and inputting;J represents hydraulic press system in j-th of technical process;
Hidden layer setting up submodule, for usingCalculate hidden
Excitation function containing layer;Wherein, cmFor the central point vector of m-th RBF, G is Gauss functions, σmFor m-th of footpath
To the characteristic parameter of basic function;M is hidden layer neuron number;
Output layer setting up submodule, for using formula b=[yM] and formula (k)Calculate
The network output of output layer;Wherein, wmThe weight between hidden layer and output layer unit.
Optionally, the radial basis function neural network training module specifically includes:
Initial parameter sets submodule, for setting the initial value of the parameter in the radial basis function neural network;Institute
Stating parameter includes central point vector cm, characteristic parameter σmThe weight w between hidden layer and output layer unitm;
Network inputs determination sub-module, the footpath is determined for the input vector according to the training sample and output vector
To the network inputs of basis function neural network;
Network exports calculating sub module, for being exported according to the network inputs calculating network;
Error judgment submodule, for calculating the output vector of the training sample and the error of network output, sentence
Whether the error of breaking is less than setting value, if it is not, parameter adjustment is then carried out, if so, terminating training;
Parameter adjustment submodule, for based on learning algorithm, adjusting the parameter in radial basis function neural network.
Optionally, the parameter adjustment submodule specifically includes:
Constant determining unit is extended, each RBF is determined for the distance between center according to each RBF
The minimum range at center and other RBF centers extend constant as each RBF is corresponding;
Cost function calculation unit, it is defeated for the error according to the output vector and network output, calculating network
Go out and cost function of the mean square error of output vector as error gradient;
Parameter adjustment unit, for according to the extension constant and cost function, error gradient negative direction with
The learning rate adjusting parameter of setting.
According to specific embodiment provided by the invention, the invention discloses following technique effect:
The invention discloses a kind of analogy method and system of hydraulic pressure process for machining, regard hydraulic pressure machining process as one
Individual nonlinear and time-varying system, the radial basis function neural network for each technical process established in hydraulic pressure machining process;Collection
The input vector and output vector of hydraulic pressure machining process, training sample is obtained, and initial parameter is set, to the radial direction base letter
Number neutral net is trained, until the precision of RBF (Radical Basis Function, RBF) neutral net expires
Foot requires, using the method and system of the present invention, can improve the simulation precision of the processing technology of hydraulic press.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to institute in embodiment
The accompanying drawing needed to use is briefly described, it should be apparent that, drawings in the following description are only some implementations of the present invention
Example, for those of ordinary skill in the art, without having to pay creative labor, can also be according to these accompanying drawings
Obtain other accompanying drawings.
Fig. 1 is a kind of flow chart of the analogy method of hydraulic pressure process for machining provided by the invention;
Fig. 2 is a kind of schematic diagram of the analogy method of hydraulic pressure process for machining provided by the invention;
Fig. 3 is that a kind of radial basis function neural network of analogy method of hydraulic pressure process for machining provided by the invention is trained
Flow chart;
Fig. 4 is a kind of structured flowchart of the simulation system of hydraulic pressure process for machining provided by the invention.
Embodiment
It is an object of the invention to provide a kind of analogy method and system of hydraulic pressure process for machining, to improve adding for hydraulic press
The simulation precision of work technique.
In order to facilitate the understanding of the purposes, features and advantages of the present invention, it is below in conjunction with the accompanying drawings and specific real
Mode is applied to be described in further detail invention.
As shown in figure 1, a kind of analogy method of hydraulic pressure process for machining, comprises the following steps:
Step 101, the radial basis function neural network for each technical process established in hydraulic pressure machining process;
Step 102, the input vector and output vector of each technical process in hydraulic press process are gathered, is instructed
Practice sample;
Step 103, according to the training sample, initial parameter is set, the radial basis function neural network is instructed
Practice, until radial basis function neural network and the technical process corresponding to the radial basis function neural network of each technical process
Actual condition error be less than setting value, obtain hydraulic press process model;
Step 104, vector is actually entered according to hydraulic pressure machining process, using the hydraulic press process model,
Carry out the simulation of the hydraulic pressure process for machining.
Specifically, the schematic diagram of the analogy method of processing technology as shown in Figure 2, the principle of the analogy method are:It is first
First, a RBF neural is established, then, the input u (k) and output y (k) of simulated object is gathered, according to the defeated of simulated object
Enter u (k) and export y (k) obtain RBF neural network inputs x=[y (k-1) ..., y (k-ny);u(k),u(k-
1),…,u(k-nu);j;K], and using the network output y of RBF neural calculating RBF neuralm(k), network is exported
ym(k) compare with simulated object output y (k), error e (k) is drawn, according to error e (k) to the ginseng in the RBF neural
Number is adjusted, and the RBF neural network model is accurately simulated the simulated object.
Optionally, the radial basis function neural network includes input layer, hidden layer and output layer, described to establish hydraulic press
The radial basis function neural network of each technical process in process, is specifically included:
Using formula x=[y (k-1) ..., y (k-ny);u(k),u(k-1),…,u(k-nu);j;K] calculate input layer
Network inputs;Wherein, y (k), u (k) are respectively that kth time samples obtained output vector, input vector, nyAnd nuRespectively export
With the maximum delay of input;J represents hydraulic press system in j-th of technical process;
UsingCalculate the excitation function of hidden layer;Wherein,
cmFor the central point vector of m-th RBF, G is Gauss functions, σmFor the characteristic parameter of m-th of RBF;M is
Hidden layer neuron number;
Using formula b=[yM] and formula (k)Calculate the network output of output layer;Wherein,
wmThe weight between hidden layer and output layer unit.
Optionally, it is described according to the training sample, initial parameter is set, the radial basis function neural network is carried out
Training, until radial basis function neural network and the technique mistake corresponding to the radial basis function neural network of each technical process
The actual condition error of journey is less than setting value, as shown in figure 3, specifically including:
Step 1, the initial value of the parameter in the radial basis function neural network is set;The parameter include central point to
Measure cm, characteristic parameter σmThe weight w between hidden layer and output layer unitm;
Step 2, the input vector in the training sample and output vector determine the radial ba-sis function network
The network inputs of network;
Step 3, exported according to the network inputs calculating network;
Step 4, the output vector in the training sample and the error of network output are calculated, judges that the error is
It is no to be less than the setting value, if it is not, step 5 is then performed, if so, then terminating to train;
Step 5, based on learning algorithm, the parameter in radial basis function neural network, return to step 2 are adjusted.
Optionally, the parameter for being based on learning algorithm, adjusting in radial basis function neural network, is specifically included:
The distance between center according to each RBF determines the center of each RBF and other radial direction base letters
The minimum range at number center extends constant as corresponding to each RBF;
Concrete mode is:Using formulaAnd δm=λ dmIt is normal to calculate extension corresponding to each RBF
Number;Wherein, dmRepresent the minimum range at the center of m-th of RBF and the center of other RBFs, m=1,
2 ... ..., M;M1=1,2 ... ..., M, m1 ≠ m;δm1To extend constant corresponding to the m1 RBF, λ is overlap coefficient;
The mean square error of the error exported according to the output vector and the network, calculating network output and output vector
Cost function as error gradient;
Concrete mode is:The error exported according to the output vector and the network, utilizes formulaCalculate
Network exports and cost function of the mean square error of output vector as error gradient;Wherein, To input error signal during k-th of sample, k=1,2 ... ..., K;K is sample size;
According to the extension constant and the cost function, institute is adjusted with the learning rate set in the negative direction of error gradient
State the parameter in radial basis function neural network.
Concrete mode is:Equation below adjusting parameter is utilized with certain learning rate η in the negative direction of error gradient:
As shown in figure 4, the present invention also provides a kind of simulation system of hydraulic pressure process for machining, including Radial Basis Function neural
Network establishes module 401, data acquisition module 402, radial basis function neural network training module 403, hydraulic pressure process for machining
Analog module 404;
Institute's radial basis function neural network establishes module 401, for establishing each technique mistake in hydraulic pressure machining process
The radial basis function neural network of journey;
The data acquisition module 402, for gather hydraulic press process each technical process input vector and
Output vector, obtain training sample;
The radial basis function neural network training module 403, for according to the training sample, setting initial parameter,
To being trained in the radial basis function neural network, until radial basis function neural network and the footpath of each technical process
It is less than setting value to the actual condition error of the technical process corresponding to basis function neural network, obtains hydraulic pressure process for machining mould
Type.
The hydraulic pressure process for machining analog module 404, for actually entering vector according to hydraulic pressure machining process, is adopted
With the hydraulic press process model, the simulation of hydraulic pressure process for machining is carried out.
Optionally, the radial basis function neural network establishes module 401, specifically includes:
Input layer setting up submodule, for using formula x=[y (k-1) ..., y (k-ny);u(k),u(k-1),…,u
(k-nu);j;K] calculate input layer network inputs;Wherein, y (k), u (k) be respectively kth time sampling obtain output vector,
Input vector, nyAnd nuThe maximum delay for respectively exporting and inputting;J represents hydraulic press system in j-th of technical process;
Hidden layer setting up submodule, for usingCalculate hidden
Excitation function containing layer;Wherein, cmFor the central point vector of m-th RBF, G is Gauss functions, σmFor m-th of footpath
To the characteristic parameter of basic function;M is hidden layer neuron number;
Output layer setting up submodule, for using formula b=[yM] and formula (k)Calculate
The network output of output layer;Wherein, wmThe weight between hidden layer and output layer unit.
Optionally, the radial basis function neural network training module 403 specifically includes:
Initial parameter sets submodule, for setting the initial value of the parameter in the radial basis function neural network;Institute
Stating parameter includes central point vector cm, characteristic parameter σmThe weight w between hidden layer and output layer unitm;
Network inputs determination sub-module, the footpath is determined for the input vector according to the training sample and output vector
To the network inputs of basis function neural network;
Network exports calculating sub module, for being exported according to the network inputs calculating network;
Error judgment submodule, for calculating the output vector of the training sample and the error of network output, sentence
Whether the error of breaking is less than setting value, if it is not, parameter adjustment is then carried out, if so, terminating training;
Parameter adjustment submodule, for based on learning algorithm, adjusting the parameter in radial basis function neural network.
Optionally, the parameter adjustment submodule specifically includes:
Constant determining unit is extended, each RBF is determined for the distance between center according to each RBF
The minimum range at center and other RBF centers extend constant as each RBF is corresponding;
Cost function calculation unit, it is defeated for the error according to the output vector and network output, calculating network
Go out and cost function of the mean square error of output vector as error gradient;
Parameter adjustment unit, for according to the extension constant and cost function, error gradient negative direction with
The learning rate adjusting parameter of setting.
Optionally, analogy method and system provided by the invention can be applied to the simulation of the technical process of hydraulic press, but not
The simulation of the technical process of hydraulic press is only limitted to, during applied to hydraulic pressure, first, gathers each technique in hydraulic press process
The input vector and output vector of process, training sample is obtained, is specifically included:
Hydraulic press has " beginning → slide block down → deceleration and pressurization → pressurize delay → master cylinder in part process
Multiple technical process such as pressure release → sliding block backhaul → stopping ", each technical process have its distinctive operating mode feature;
The sampling of data is carried out to the hydraulic press system when each operation stage starts, and is finished in the operation stage
When terminate data sampling, so as to form the inputoutput data collection in the stage;
Hydraulic press system is described as follows in the output of different process process stage:
Y (k)=fj(y(k-1),…,y(k-ny);u(k),u(k-1),…,u(k-nu);θ(k))+v(k) (1)
Wherein, y (k), u (k) and v (k) are respectively system output vector, input vector and the noise that kth time sampling obtains
Vector;nyAnd nuThe maximum delay for respectively exporting and inputting;V (k) is that the Gauss that average is zero is independently distributed noise sequence;θ
(k) it is the system time-varying parameter vector of hydraulic press;fj() is unknown nonlinear vector function to be identified, represents hydraulic press system
System input and non-linear relation of outlet chamber in j-th of technical process;
The inputoutput data collection of each technical process is obtained, builds training sample;
Then, different RBF neurals is respectively adopted for each operation stage to be simulated, and uses training sample
The RBF neural of each operation stage is trained;Concretely comprise the following steps:
(1) slide block down operation stage hydraulic press system is simulated
1. initial period, simultaneously positive in hydraulic servo control system input pressurization signal u (t), hydraulic servo electric motor starting
Work, hydraulic cylinder start to pressurize, and hydraulic slide starts downward movement;
2. for first hydraulic processes stage " slide block down ", the pressure sensor being installed on slider of hydraulic press is utilized
Hydraulic pressure systemic effect is sampled in the pressure F on sliding block with time interval Δ t with range sensora(k) and slider displacement L (k), and
Exported as the stage system;
3. hydraulic system input is [u (k-nu),...,u(k);K], output be [Fa(k-ny),...,Fa(k);L(k-
ny) ..., L (k)], RBF neural Θ is obtained according to the input of hydraulic system and output1Input and output, wherein inputting
With output retardation nuAnd nyIt is set to fixed natural number.
(2) slow down and pressing technology stage hydraulic press system is simulated
1. second operation stage " slow down and pressurize " of system is triggered when hydraulic slide contacts with part base tool, now
Continue the system input maintained like;
2. hydraulic pressure is sampled with time interval Δ t with range sensor using the pressure sensor being installed on slider of hydraulic press
Systemic effect and the pressure F on sliding blocka(k) and slider displacement L (k), while the pressure being installed on hydraulic cushion of hydraulic press is utilized
Sensor is with the stress F of time interval Δ t sampling part base toolsb(k);And exported as the stage system;
3. hydraulic system input is [u (k-nu),...,u(k);K], output be [Fa(k-ny),...,Fa(k);L(k-
ny),...,L(k);Fb(k-ny),...,Fb(k) RBF neural Θ], is obtained according to the input of hydraulic system and output2It is defeated
Enter and export, wherein input and output retardation nuAnd nyIt is set to fixed natural number;
(3) pressurize delay and the simulation of master cylinder decompression technique stage hydraulic press system
1. when hydraulic slide completely in place after, hydraulic press enters " pressurize delay and master cylinder pressure release " operation stage, now liquid
Pressure servo-control system input u (t) is gradually decreased as zero, and completes hydraulic cylinder pressure relief;
2. hydraulic pressure systemic effect and cunning are sampled with time interval Δ t using the pressure sensor being installed on slider of hydraulic press
Pressure F on blocka(k), while using the pressure sensor being installed on hydraulic cushion of hydraulic press with time interval Δ t part is sampled
The stress F of base toolb(k);And exported as the stage system;
3. hydraulic system input is [u (k-nu),...,u(k);K], output be [Fa(k-ny),...,Fa(k);Fb(k-
ny),...,Fb(k) RBF neural Θ], is obtained according to hydraulic system input and output3Input and output, wherein input and
Export retardation nuAnd nyIt is set to fixed natural number;
(4) sliding block backhaul operation stage hydraulic press system is simulated
1. after the hydroforming having to part base terminates, returned signal u (t) is inputted in hydraulic servo control system,
Hydraulic servo electric motor starting and reverse operation, hydraulic cylinder start reversely to pressurize, and hydraulic slide starts drawback movement;
2. for last hydraulic processes stage " slide block down ", the pressure sensing being installed on slider of hydraulic press is utilized
Device is with range sensor with time interval Δ t sampling hydraulic pressure systemic effects and the buffer brake F on sliding blockaAnd slider displacement L (k)
(k), and as the stage system export;
3. hydraulic system input is [u (k-nu),...,u(k);K], output be [Fa(k-ny),...,Fa(k);L(k-
ny) ..., L (k)], RBF neural Θ is obtained according to the input of hydraulic system and output4Input and output, wherein inputting
With output retardation nuAnd nyIt is set to fixed natural number.
According to specific embodiment provided by the invention, the invention discloses following technique effect:
The invention discloses a kind of analogy method and system of hydraulic pressure process for machining, regard hydraulic pressure machining process as one
Individual nonlinear and time-varying system, the radial basis function neural network for each technical process established in hydraulic pressure machining process;Collection
The input vector and output data of hydraulic pressure machining process, training sample is obtained, and initial parameter is set, to RBF nerves
Network is trained, and until the precision satisfaction requirement of RBF neural, using the method and system of the present invention, can improve liquid
The simulation precision of the processing technology of press.
Each embodiment is described by the way of progressive in this specification, what each embodiment stressed be and other
The difference of embodiment, between each embodiment identical similar portion mutually referring to.For system disclosed in embodiment
For, because it is corresponded to the method disclosed in Example, so description is fairly simple, related part is said referring to method part
It is bright.
Specific case used herein is set forth to the principle and embodiment of invention, the explanation of above example
It is only intended to help the method and its core concept for understanding the present invention, described embodiment is only that the part of the present invention is real
Example, rather than whole embodiments are applied, based on the embodiment in the present invention, those of ordinary skill in the art are not making creation
Property work under the premise of the every other embodiment that is obtained, belong to the scope of protection of the invention.
Claims (8)
1. a kind of analogy method of hydraulic pressure process for machining, it is characterised in that comprise the following steps:
The radial basis function neural network for each technical process established in hydraulic pressure machining process;
The input vector and output vector of each technical process of hydraulic press process are gathered, obtains training sample;
According to the training sample, initial parameter is set, the radial basis function neural network is trained, until each work
The radial basis function neural network of skill process and the actual condition of the technical process corresponding to the radial basis function neural network miss
Difference is less than setting value, obtains hydraulic press process model;
Vector is actually entered according to hydraulic pressure machining process, using the hydraulic press process model, carries out the hydraulic pressure
The simulation of process for machining.
2. according to the method for claim 1, it is characterised in that the radial basis function neural network includes input layer, hidden
Containing layer and output layer, the radial basis function neural network of each technical process established in hydraulic pressure machining process, specifically
Including:
Using formula x=[y (k-1) ..., y (k-ny);u(k),u(k-1),…,u(k-nu);j;K] calculate input layer network
Input;Wherein, y (k), u (k) are respectively that kth time samples obtained output vector, input vector, nyAnd nuRespectively export and defeated
The maximum delay entered;J represents hydraulic press system in j-th of technical process;
UsingCalculate the excitation function of hidden layer;Wherein, cmFor
The central point vector of m-th RBF, G are Gauss functions, σmFor the characteristic parameter of m-th of RBF;M is hidden
Neuron number containing layer;
Using formula b=[yM] and formula (k)Calculate the network output of output layer;Wherein, wmFor
Weight between hidden layer and output layer unit.
3. according to the method for claim 1, it is characterised in that described right according to the training sample, setting initial parameter
The radial basis function neural network is trained, until radial basis function neural network and the radial direction base of each technical process
The actual condition error of technical process corresponding to Function Neural Network is less than setting value, obtains hydraulic press process model,
Specifically include:
Step 1, the initial value of the parameter in the radial basis function neural network is set;The parameter includes central point vector cm、
Characteristic parameter σmThe weight w between hidden layer and output layer unitm;
Step 2, the input vector in the training sample and output vector determine the radial basis function neural network
Network inputs;
Step 3, exported according to the network inputs calculating network;
Step 4, the output vector in the training sample and the error of network output are calculated, judges whether the error is small
In the setting value, if it is not, step 5 is then performed, if so, then terminating to train;
Step 5, based on learning algorithm, the parameter in radial basis function neural network, return to step 2 are adjusted.
4. according to the method for claim 3, it is characterised in that it is described to be based on learning algorithm, adjust Radial Basis Function neural
Parameter in network, is specifically included:
The distance between center according to each RBF is determined in center and the other RBFs of each RBF
The minimum range of the heart extends constant as corresponding to each RBF;
The error exported according to the output vector and the network, calculating network output and the mean square error conduct of output vector
The cost function of error gradient;
According to the extension constant and the cost function, the footpath is adjusted with the learning rate set in the negative direction of error gradient
Parameter into basis function neural network.
A kind of 5. simulation system of hydraulic pressure process for machining, it is characterised in that including radial basis function neural network establish module,
Data acquisition module, radial basis function neural network training module, hydraulic pressure process for machining analog module;
Institute's radial basis function neural network establishes module, for establishing the radial direction of each technical process in hydraulic pressure machining process
Basis function neural network;
The data acquisition module, for gather hydraulic press process each technical process input vector and export to
Amount, obtain training sample;
The radial basis function neural network training module, for according to the training sample, initial parameter being set, to the footpath
It is trained into basis function neural network, until radial basis function neural network and the RBF of each technical process
The actual condition error of technical process corresponding to neutral net is less than setting value, obtains hydraulic press process model;
The hydraulic pressure process for machining analog module, for actually entering vector according to hydraulic pressure machining process, using described
Hydraulic press process model, carry out the simulation of hydraulic pressure process for machining.
6. system according to claim 5, it is characterised in that the radial basis function neural network establishes module, specifically
Including:
Input layer setting up submodule, for using formula x=[y (k-1) ..., y (k-ny);u(k),u(k-1),…,u(k-nu);
j;K] calculate input layer network inputs;Wherein, y (k), u (k) be respectively kth time sampling obtain output vector, input to
Amount, nyAnd nuThe maximum delay for respectively exporting and inputting;J represents hydraulic press system in j-th of technical process;
Hidden layer setting up submodule, for usingCalculate hidden layer
Excitation function;Wherein, cmFor the central point vector of m-th RBF, G is Gauss functions, σmFor m-th of radial direction base letter
Several characteristic parameters;M is hidden layer neuron number;
Output layer setting up submodule, for using formula b=[yM] and formula (k)Calculate output
The network output of layer;Wherein, wmThe weight between hidden layer and output layer unit.
7. system according to claim 5, it is characterised in that the radial basis function neural network training module specifically wraps
Include:
Initial parameter sets submodule, for setting the initial value of the parameter in the radial basis function neural network;The ginseng
Number includes central point vector cm, characteristic parameter σmThe weight w between hidden layer and output layer unitm;
Network inputs determination sub-module, the radial direction base is determined for the input vector according to the training sample and output vector
The network inputs of Function Neural Network;
Network exports calculating sub module, for being exported according to the network inputs calculating network;
Error judgment submodule, for calculating the output vector of the training sample and the error of network output, judge institute
State whether error is less than setting value, if it is not, parameter adjustment is then carried out, if so, terminating training;
Parameter adjustment submodule, for based on learning algorithm, adjusting the parameter in radial basis function neural network.
8. system according to claim 7, it is characterised in that the parameter adjustment submodule specifically includes:
Constant determining unit is extended, is determined for the distance between center according to each RBF in each RBF
The minimum range at the heart and other RBF centers extends constant as each RBF is corresponding;
Cost function calculation unit, for according to the output vector and the network output error, calculating network output and
Cost function of the mean square error of output vector as error gradient;
Parameter adjustment unit, for according to the extension constant and cost function, in the negative direction of error gradient to set
Learning rate adjusting parameter.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710586939.4A CN107357997A (en) | 2017-07-18 | 2017-07-18 | A kind of analogy method and system of hydraulic pressure process for machining |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710586939.4A CN107357997A (en) | 2017-07-18 | 2017-07-18 | A kind of analogy method and system of hydraulic pressure process for machining |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107357997A true CN107357997A (en) | 2017-11-17 |
Family
ID=60284746
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710586939.4A Pending CN107357997A (en) | 2017-07-18 | 2017-07-18 | A kind of analogy method and system of hydraulic pressure process for machining |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107357997A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109558681A (en) * | 2018-11-30 | 2019-04-02 | 北京新能源汽车股份有限公司 | A kind of preparation method and device of the loss power of insulated gate bipolar transistor |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6606157B1 (en) * | 1996-09-14 | 2003-08-12 | University Of Hertfordshire | Detection of hazardous airborne fibres |
CN101930223A (en) * | 2010-09-07 | 2010-12-29 | 曾谊晖 | Intelligent screening system based on numerical control processing technology for difficult-to-machine metal |
US8346693B2 (en) * | 2009-11-24 | 2013-01-01 | King Fahd University Of Petroleum And Minerals | Method for hammerstein modeling of steam generator plant |
CN105290122A (en) * | 2015-08-20 | 2016-02-03 | 张守武 | AGC system thickness measurement device supported by RBF network |
CN105608295A (en) * | 2016-01-29 | 2016-05-25 | 杭州电子科技大学 | Multi-objective evolutionary algorithm (MOEA) and radial basis function (RBF) neural network optimization modeling method of coking furnace pressure |
-
2017
- 2017-07-18 CN CN201710586939.4A patent/CN107357997A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6606157B1 (en) * | 1996-09-14 | 2003-08-12 | University Of Hertfordshire | Detection of hazardous airborne fibres |
US8346693B2 (en) * | 2009-11-24 | 2013-01-01 | King Fahd University Of Petroleum And Minerals | Method for hammerstein modeling of steam generator plant |
CN101930223A (en) * | 2010-09-07 | 2010-12-29 | 曾谊晖 | Intelligent screening system based on numerical control processing technology for difficult-to-machine metal |
CN105290122A (en) * | 2015-08-20 | 2016-02-03 | 张守武 | AGC system thickness measurement device supported by RBF network |
CN105608295A (en) * | 2016-01-29 | 2016-05-25 | 杭州电子科技大学 | Multi-objective evolutionary algorithm (MOEA) and radial basis function (RBF) neural network optimization modeling method of coking furnace pressure |
Non-Patent Citations (9)
Title |
---|
HUAIZHONG CHEN: "Research of the Electro-hydraulic Servo System Based on RBF Fuzzy Neural Network Controller", 《JOURNAL OF SOFTWARE》 * |
ZHIHUAI XIAO等: "Identifying of Hydraulic Turbine Generating Unit Model Based on Neural Network", 《PROCEEDINGS OF THE SIXTH INTERNATIONAL CONFERENCE ON INTELLIGENT SYSTEMS DESIGN AND APPLICATIONS》 * |
庞中华等: "《系统辨识与自适应控制MATLAB仿真 修订版》", 31 August 2013, 北京航空航天大学出版社 * |
曹裕: "《复杂环境下我国企业财务困境模式及预警研究:基于企业生命周期的视角》", 30 June 2015 * |
朱学莉: "《智能建筑环境检测与控制技术》", 31 December 2012, 北京:中国电力出版社 * |
李艳聪等: "基于神经网络和遗传算法的液压机上梁轻量化和刚度优化设计", 《机械科学与技术》 * |
梁景凯等: "《智能控制技术》", 31 March 2016, 哈尔滨工业大学出版社 * |
莫玉梅: "基于神经网络的镁合金汽车车轮锻压工艺优化", 《轻合金加工技术》 * |
龙四营: "拉伸液压机预紧方案设计与压制性能优化仿真技术及其应用研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109558681A (en) * | 2018-11-30 | 2019-04-02 | 北京新能源汽车股份有限公司 | A kind of preparation method and device of the loss power of insulated gate bipolar transistor |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110400006A (en) | Oil well output prediction technique based on deep learning algorithm | |
CN104537415B (en) | A kind of non-linear process industry failure predication and recognition methodss based on compressed sensing and DROS ELM | |
CN105574587B (en) | A kind of online operating mode course monitoring method of plastic injection molding process | |
CN109685314B (en) | Non-intrusive load decomposition method and system based on long-term and short-term memory network | |
CN100507460C (en) | Dynamic soft measuring and form establishing method base pulse response formwork and parameter optumization | |
CN105834835A (en) | Method for monitoring tool wear on line based on multiscale principal component analysis | |
CN108760829A (en) | A kind of electronic nose recognition methods based on bionical olfactory bulb model and convolutional neural networks | |
CN109884892A (en) | Process industry system prediction model based on crosscorrelation time lag grey correlation analysis | |
CN110210687A (en) | A kind of Nonlinear Dynamic production process product quality prediction technique returned based on local weighted slow feature | |
CN107234495A (en) | Set up the method for average material removal rate forecast model and the method for predicted velocity | |
CN104615840B (en) | The modification method and system of a kind of digital simulation model | |
CN104915515A (en) | BP neural network based GFET modeling method | |
CN109634108A (en) | The different factor full format non-model control method of the MIMO of parameter self-tuning | |
Leng et al. | Manufactured solutions and the verification of three-dimensional Stokes ice-sheet models | |
CN105892296A (en) | Fractional order dynamic matrix control method for industrial heating furnace system | |
CN110955928A (en) | Constant-load stress testing system and method for concrete bridge and information data processing terminal | |
CN103279030B (en) | Dynamic soft measuring modeling method and device based on Bayesian frame | |
CN107357997A (en) | A kind of analogy method and system of hydraulic pressure process for machining | |
CN115186586A (en) | Compaction construction quality iterative inversion control system based on machine-path cooperation | |
CN109918704B (en) | Die forging die life prediction method based on finite element simulation | |
CN104634265A (en) | Soft measurement method for thickness of mineral floating foam layer based on multivariate image feature fusion | |
CN110222825B (en) | Cement product specific surface area prediction method and system | |
CN103136425A (en) | Rapid spacecraft quality characteristic analysis method | |
CN102929147B (en) | On-line control system and method for machining vacuum glass | |
CN112749508B (en) | Road feel simulation method based on GMM and BP neural network |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171117 |