CN109598096A - A kind of online planing method of smithing technological parameter based on adaptive Dynamic Programming - Google Patents

A kind of online planing method of smithing technological parameter based on adaptive Dynamic Programming Download PDF

Info

Publication number
CN109598096A
CN109598096A CN201910038183.9A CN201910038183A CN109598096A CN 109598096 A CN109598096 A CN 109598096A CN 201910038183 A CN201910038183 A CN 201910038183A CN 109598096 A CN109598096 A CN 109598096A
Authority
CN
China
Prior art keywords
network
technological parameter
forging
smithing technological
microstructure
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
Application number
CN201910038183.9A
Other languages
Chinese (zh)
Other versions
CN109598096B (en
Inventor
蔺永诚
谌东东
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.)
Central South University
Original Assignee
Central South University
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 Central South University filed Critical Central South University
Priority to CN201910038183.9A priority Critical patent/CN109598096B/en
Publication of CN109598096A publication Critical patent/CN109598096A/en
Application granted granted Critical
Publication of CN109598096B publication Critical patent/CN109598096B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • B21—MECHANICAL METAL-WORKING WITHOUT ESSENTIALLY REMOVING MATERIAL; PUNCHING METAL
    • B21J—FORGING; HAMMERING; PRESSING METAL; RIVETING; FORGE FURNACES
    • B21J9/00—Forging presses
    • B21J9/10—Drives for forging presses
    • B21J9/20—Control devices specially adapted to forging presses not restricted to one of the preceding subgroups
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Geometry (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Mechanical Engineering (AREA)
  • Forging (AREA)

Abstract

The present invention provides a kind of online planing methods of the smithing technological parameter based on adaptive Dynamic Programming.This method comprises the following steps: (1) initializing the weight for executing network and evaluating network of adaptive dynamic programming method；(2) according to the microstructure state at forging system current time and last moment metal forging, the smithing technological parameter of subsequent time is provided by execution network；(3) smithing technological parameter is input to forging system, obtains the microstructure state of subsequent time metal forging；(4) according to the microstructure state of the smithing technological parameter at current time and metal forging, the estimated value of performance index function is provided by evaluation network, and the weight for executing network and evaluation network is updated；(5) step (2) are transferred to, the smithing technological parameter planning of subsequent time is carried out.The method of the present invention can rapidly cook up reasonable smithing technological parameter, and the microstructure for accuracy controlling metal forging provides technical support.

Description

A kind of online planing method of smithing technological parameter based on adaptive Dynamic Programming
Technical field:
The invention belongs to technical field of forging, it is related to a kind of smithing technological parameter based on adaptive Dynamic Programming in line gauge The method of drawing.
Background technique:
In complicated forging process, the microstructure state of forging is monitored in real time, and online to forging technology Parameter is planned and is adjusted, and is the key that guarantee forging quality.Therefore, how according to the microstructure target of forging quickly, Accurately determining following technological parameter is to realize microstructure regulation urgent need to solve the problem.
Adaptive dynamic programming method is the intelligent control method based on the theory of optimal control and intensified learning principle, is solution The certainly effective ways of large-scale complex nonlinear optimization control problem.The thought of adaptive Dynamic Programming is to utilize approximation to function knot Structure approaches performance index function and control strategy in Dynamic Programming Equation, to meet the principle of optimization to obtain optimal control System and optimal performance index function.Based on intensified learning principle, adaptive dynamic programming method is simulated people and is learnt from environmental feedback Thinking, realize intelligence learning.In view of adaptive, the self-learning capability of adaptive dynamic programming method, the method is extensive Applied to robot control, energy control and the various fields such as economic management and decision.It is difficult for metal forging microstructure The problem of with accuracy controlling, it may be considered that utilize adaptive dynamic programming method, the online smithing technological parameter and in real time planned Regulate and control the microstructure of forging.
The present invention proposes a kind of method of online planning smithing technological parameter on the basis of adaptive Dynamic Programming, It realizes the online regulation of forging microstructure, improves the quality and performance of forging to a certain extent.
Summary of the invention:
The purpose of the present invention is to provide a kind of online planing method of the smithing technological parameter based on adaptive Dynamic Programming, Solve the problems, such as that the microstructure of forging is difficult to accuracy controlling in forging process.
The scheme that the present invention solves above-mentioned problem is: a kind of smithing technological parameter based on adaptive Dynamic Programming is in line gauge The method of drawing, this method comprises the following steps:
Step 1: initializing the weight for executing network and evaluating network of adaptive dynamic programming method；
Step 2: according to the microstructure state at forging system current time and last moment metal forging, by execution network Provide the smithing technological parameter of subsequent time；
Step 3: smithing technological parameter being input to forging system, obtains the microstructure shape of subsequent time metal forging State；
Step 4: according to the microstructure state of the smithing technological parameter at current time and metal forging, being given by evaluation network The estimated value of performance index function out is updated the weight for executing network and evaluation network；
Step 5: being transferred to step 2, carry out the smithing technological parameter planning of subsequent time.
According to above scheme, the execution network of adaptive dynamic programming method described in step 1 and evaluation network can be retouched It states are as follows:
Executing network and evaluation network is constructed with BP neural network, can be expressed as respectively
U (k)=Wa2(k)·σ(Wa1(k)·x(k)) (1)
J (k)=Wc2(k)·σ(Wc1(k)·[x(k)；u(k)]) (2)
In formula: x and u, which is respectively indicated, executes outputting and inputting for neural network, and J is the output (performance for evaluating neural network The estimated value of target function), Wa1And Wa2It is the weight for executing neural network, Wc1And Wc2It is the weight for evaluating neural network, σ () is activation primitive.Performance index function J can be indicated are as follows:
In formula: γ is discount factor (0≤γ≤1), and U is utility function.
Utility function can be indicated for the purpose of controlling microstructure are as follows:
U (k)=a (f (k)-fexp)2+b(g(k)-gexp)2 (4)
In formula: f and g respectively indicates recrystallization score and average grain size, fexpAnd gexpIt respectively indicates and desired ties again Brilliant score and average grain size, a and b are respectively the weight for recrystallizing score and average grain size.
According to above scheme, the microstructure state of metal forging described in step 2 be can be described as:
In forging process, the microstructure state of metal forging refers to recrystallization score and average grain size, this two A quantity of state can not real-time measurement, can use microstructure model prediction.
According to above scheme, being updated described in step 4 to the weight for executing network and evaluation network be can be described as:
Execute the right value update formula of neural network are as follows:
Wa (k+1)=Wa (k)+Δ Wa (k) (6)
In formula: Wa is the weight for executing neural network,
Evaluate the right value update formula of neural network are as follows:
Wc (k+1)=Wc (k)+Δ Wc (k) (8)
In formula: Wc is the weight for evaluating neural network,
According to above scheme, microstructure state is predicted in real time using the microstructure model of metal forging, according to building The utility function of control microstructure smithing technological parameter is planned by adaptive dynamic programming method online, metal is forged The microstructure of part is regulated and controled.
Beneficial effects of the present invention: the present invention is difficult to essence for the microstructure of metal forging during actual industrial production Really the problem of regulation, according to the utility function towards microstructure, forging work is planned online using adaptive dynamic programming method Skill parameter, the microstructure of real-time monitoring metal forging are applicable in various metals die-forging of forge pieces and flat-die forging processing, New method is provided to obtain the microstructure of uniform, fine.The invention and popularization and application of this method are to the microcosmic of regulation metal forging Tissue has important engineering significance.
Detailed description of the invention:
The online planning flow chart of smithing technological parameter of the Fig. 1 based on adaptive Dynamic Programming；
Fig. 2 nickel-base alloy (GH4169) forges sample: before (a) forging；(b) after forging；
Fig. 3 executes network structure；
Fig. 4 evaluates network structure；
The smithing technological parameter of Fig. 5 planning；
Fig. 6 nickel-base alloy microstructure control result: score (a) is recrystallized；(b) average grain size；
The nickel-base alloy microstructure that Fig. 7 is finally obtained.
Specific embodiment:
The present invention will be described in detail with reference to the accompanying drawings and detailed description.
The present invention is a kind of online planing method of the smithing technological parameter based on adaptive Dynamic Programming, and flow chart is as schemed Shown in 1.Below by taking the forging process of nickel-base alloy (GH4169) sample (Fig. 2) as an example, forging of the present invention is discussed in detail The implementation detail that technological parameter is planned online, method include:
Step 1: initializing the weight for executing network and evaluating network of adaptive dynamic programming method；
In adaptive dynamic programming method, executing network and evaluation network is constructed with BP neural network.It is forging Before making experiment, what the execution network of adaptive dynamic programming method and the weight for evaluating network were all randomly generated.
Step 2: according to the microstructure state at forging system current time and last moment metal forging, by execution network Provide the smithing technological parameter of subsequent time；
Network structure is executed as shown in figure 3, its expression formula are as follows:
ha1(k)=Wa1(k)·[f(k)；g(k)；f(k-1)；g(k-1)] (9)
In formula: f and g respectively indicates recrystallization score and average grain size,For strain rate, Wa1And Wa2It is to execute mind Weight through network, ha1And ha2Respectively indicate outputting and inputting for hidden layer.According to forging system current time and last moment The microstructure state of nickel-base alloy forging finds out the smithing technological parameter of subsequent time using execution network.
Step 3: smithing technological parameter being input to forging system, obtains the microstructure of subsequent time nickel-base alloy forging State；
In forging process, nickel-base alloy Microstructure evolution is extremely complex, recrystallizes score and average grain size Can not real-time measurement, can use nickel-base alloy microstructure model and predicted according to smithing technological parameter, the prediction model Are as follows:
In formula: f is dynamic recrystallization score, and ε is strain, εcFor dynamic recrystallization critical strain, ε0.5Occur for recrystallization 50% strain, gdrx, g and g0Dynamic recrystallization crystallite dimension, average grain size and Initial Grain Size are respectively indicated,For Strain rate, T are forming temperature, and R is gas constant.
Step 4: according to the microstructure state of the smithing technological parameter at current time and nickel-base alloy forging, by evaluation net Network provides the estimated value of performance index function, is updated to the weight for executing network and evaluation network；
Network structure is evaluated as shown in figure 4, its expression formula are as follows:
J (k)=Wc2(k)·hc2(k) (15)
In formula: Wc1And Wc2It is the weight for evaluating neural network, hc1And hc2Respectively indicate outputting and inputting for hidden layer, J It is the output (estimated value of performance index function) for evaluating neural network.Performance index function J can be indicated are as follows:
In formula: γ is discount factor (0≤γ≤1), and U is utility function.
Utility function can be indicated for the purpose of controlling nickel-base alloy forging microstructure are as follows:
U (k)=0.4 (f (k) -1)2+0.6(g(k)-10)2 (17)
In formula: f and g respectively indicates recrystallization score and average grain size.Desired recrystallization score in this experiment It is respectively 100% and 10 μm with average grain size, it is therefore an objective to obtain most uniform tiny microstructure.
Evaluation network can estimate performance index function according to the smithing technological parameter that provide of network is executed, and feed back to holding Row neural network carries out right value update to improve its performance, executes network weight more new formula are as follows:
Wa2(k+1)=Wa2(k)+ΔWa2(k) (19)
Wa1(k+1)=Wa1(k)+ΔWa1(k) (21)
In formula: la is to execute neural network learning rate, generally takes 0.1.
Evaluation network needs to restrain performance index function sequence, right value update formula are as follows:
Wc2(k+1)=Wc2(k)+ΔWc2(k) (23)
Wc1(k+1)=Wc1(k)+ΔWc1(k) (25)
In formula: lc is evaluation neural network learning rate, generally takes 0.1.
Step 5: being transferred to step 2, carry out the smithing technological parameter planning of subsequent time.
According to the above-mentioned online planing method of smithing technological parameter, according to the utility function of the control microstructure of building, Nickel-base alloy smithing technological parameter is planned online, the smithing technological parameter of acquisition is as shown in figure 5, in strain less than 0.4 When, strain rate is about 0.08s-1；With the increase of strain, strain rate is linearly decreased to 0.02s-1, it is greater than in strain After 0.9, strain rate is 0.02s always-1.Nickel-base alloy microstructure control result as shown in fig. 6, when strain is 0.9, Recrystallization score reaches 100%, and the average grain size finally obtained is 11 μm.Nickel-based alloy sample after forging is carried out micro- Structure observation is seen, the microstructure finally obtained recrystallizes score as shown in fig. 7, obtained microstructure is highly uniform fine With average grain size be 99% and 12 μm, with propose method control effect it is consistent.This illustrates that proposition method can be advised effectively Smithing technological parameter is drawn, the microstructure of nickel base superalloy is regulated and controled.
From the above it can be found that method proposed by the present invention can plan smithing technological parameter online, realize Ni-based The Effective Regulation of alloy forged piece microstructure provides reliably to obtain the nickel-based high-temperature alloy forge piece microstructure of uniform, fine Approach.
Example of the invention is illustrated above in conjunction with attached drawing, but the present invention is not limited to above-mentioned specific embodiment party Formula, above-mentioned specific embodiment are merely exemplary.Any invention no more than the claims in the present invention, of the invention Within protection scope.

Claims (1)

1. a kind of online planing method of smithing technological parameter based on adaptive Dynamic Programming, it is characterised in that: forged for metal The problem of part microstructure is difficult to accuracy controlling proposes in view of the adaptive and self-learning capability of adaptive dynamic programming method A kind of quick, simple, the efficient online planing method of smithing technological parameter, this method comprises the following steps:
Step 1: initializing the weight for executing network and evaluating network of adaptive dynamic programming method；
Execute the expression formula of network are as follows:
ha1(k)=Wa1(k)·[f(k)；g(k)；f(k-1)；g(k-1)] (1)
In formula: f and g respectively indicates recrystallization score and average grain size,For strain rate, Wa1And Wa2It is to execute nerve net The weight of network, ha1And ha2Respectively indicate outputting and inputting for hidden layer；
Evaluate the expression formula of network are as follows:
J (k)=Wc2(k)·hc2(k) (6)
In formula: Wc1And Wc2It is the weight for evaluating neural network, hc1And hc2Outputting and inputting for hidden layer is respectively indicated, J is property The estimated value of energy target function；
Step 2: according to the microstructure state at forging system current time and last moment metal forging, being provided by execution network The smithing technological parameter of subsequent time；
Step 3: smithing technological parameter being input to forging system, obtains the microstructure state of subsequent time metal forging；
Step 4: according to the microstructure state of the smithing technological parameter at current time and metal forging, by providing property of evaluation network The estimated value of energy target function is updated the weight for executing network and evaluation network；
Execute the right value update formula of network are as follows:
Wa2(k+1)=Wa2(k)+ΔWa2(k) (8)
Wa1(k+1)=Wa1(k)+ΔWa1(k) (10)
In formula: la is to execute e-learning rate, generally takes 0.1；
Evaluate the right value update formula of network are as follows:
Wc2(k+1)=Wc2(k)+ΔWc2(k) (12)
Wc1(k+1)=Wc1(k)+ΔWc1(k) (14)
In formula: lc is evaluation e-learning rate, generally takes 0.1；
Step 5: being transferred to step 2, carry out the smithing technological parameter planning of subsequent time；
The microstructure state of metal forging described in step 2,3 and 4 is to recrystallize score and average grain size, step 2, 3, smithing technological parameter described in 4 and 5 is strain rate.
CN201910038183.9A 2019-01-16 2019-01-16 Forging process parameter online planning method based on self-adaptive dynamic planning Active CN109598096B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910038183.9A CN109598096B (en) 2019-01-16 2019-01-16 Forging process parameter online planning method based on self-adaptive dynamic planning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910038183.9A CN109598096B (en) 2019-01-16 2019-01-16 Forging process parameter online planning method based on self-adaptive dynamic planning

Publications (2)

Publication Number Publication Date
CN109598096A true CN109598096A (en) 2019-04-09
CN109598096B CN109598096B (en) 2023-07-25

Family

ID=65965200

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910038183.9A Active CN109598096B (en) 2019-01-16 2019-01-16 Forging process parameter online planning method based on self-adaptive dynamic planning

Country Status (1)

Country Link
CN (1) CN109598096B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1851281A (en) * 2006-05-25 2006-10-25 杭州友谐冷精锻技术有限公司 Cold fine forging process for automobile bearing innerand outer ring, and its forging die
CN106055844A (en) * 2016-07-06 2016-10-26 中南大学 Prediction and control method of nickel-base super alloy microstructure on the basis of BP (Back Propagation) neural network
CN106777672A (en) * 2016-12-14 2017-05-31 中南大学 Metal material forging microstructure flexible measurement method based on adaptive expert system
CN107273635A (en) * 2017-07-05 2017-10-20 中南大学 A kind of online planing method of contour forging technique for controlling nickel-base alloy microstructure to be distributed

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1851281A (en) * 2006-05-25 2006-10-25 杭州友谐冷精锻技术有限公司 Cold fine forging process for automobile bearing innerand outer ring, and its forging die
CN106055844A (en) * 2016-07-06 2016-10-26 中南大学 Prediction and control method of nickel-base super alloy microstructure on the basis of BP (Back Propagation) neural network
CN106777672A (en) * 2016-12-14 2017-05-31 中南大学 Metal material forging microstructure flexible measurement method based on adaptive expert system
CN107273635A (en) * 2017-07-05 2017-10-20 中南大学 A kind of online planing method of contour forging technique for controlling nickel-base alloy microstructure to be distributed

Also Published As

Publication number Publication date
CN109598096B (en) 2023-07-25

Similar Documents

Publication Publication Date Title
Chen et al. Laser cladding of nanoparticle TiC ceramic powder: Effects of process parameters on the quality characteristics of the coatings and its prediction model
CN106055844B (en) Nickel base superalloy microstructure predicting control method based on BP neural network
Chen et al. Modeling and simulation on dynamic recrystallization of 30Cr2Ni4MoV rotor steel using the cellular automaton method
Chen et al. A design framework for optimizing forming processing parameters based on matrix cellular automaton and neural network-based model predictive control methods
Zhang et al. Effect of cross wedge rolling on the microstructure of GH4169 alloy
EP2048553A1 (en) Control circuit for regulating a process, in particular a combustion process
Chen et al. Prediction of microstructural evolution during hot forging
Khan et al. Artificial intelligence model to predict surface roughness of Ti-15-3 alloy in EDM process
Quan et al. Modelling of the hot flow behaviors for Ti-13Nb-13Zr alloy by BP-ANN model and its application
CN109598096A (en) A kind of online planing method of smithing technological parameter based on adaptive Dynamic Programming
CN105868468B (en) Novel neighbor capturing method based on cellular automaton
Bambach et al. Hot working behavior of selective laser melted and laser metal deposited Inconel 718
Sen et al. Electro jet drilling using hybrid NNGA approach
Luo et al. Modeling of constitutive relationships and microstructural variables of Ti–6.62 Al–5.14 Sn–1.82 Zr alloy during high temperature deformation
CN106777672A (en) Metal material forging microstructure flexible measurement method based on adaptive expert system
Pradhan et al. Neuro-fuzzy model on material removal rate in electrical discharge machining in AISI D2 steel
Bingxiu et al. Study on the forecasting of the hot corrosion resistance of typical superalloys for aeroengines
Rodic et al. Application of ANFIS for modeling and predicting the EDM surface roughness
CN115198214B (en) Manufacturing method of marine diesel engine air valve capable of regulating and controlling mixed grain boundary
CN109783915A (en) Nickel-base alloy forging microstructure control method based on particle swarm optimization algorithm
Furrer et al. Computational modeling and simulation of alloy 718
Liu et al. Automatic reading and writing model of welding parameters predicted based on PSO-RFR
Li et al. Machine learning-based prediction for time series damage evolution of Ni-based superalloy microstructures
Das et al. Prediction of MRR in EDM of EN31 steel using artificial neural network
Thomas et al. Modeling of microstructure evolution during the thermomechanical processing of nickel-base superalloys

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant