CN115495853B - Particle blocking variable stiffness module parameter optimization method and device - Google Patents

Particle blocking variable stiffness module parameter optimization method and device Download PDF

Info

Publication number
CN115495853B
CN115495853B CN202211442382.4A CN202211442382A CN115495853B CN 115495853 B CN115495853 B CN 115495853B CN 202211442382 A CN202211442382 A CN 202211442382A CN 115495853 B CN115495853 B CN 115495853B
Authority
CN
China
Prior art keywords
variable stiffness
mechanical property
particle blocking
stiffness module
module
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.)
Active
Application number
CN202211442382.4A
Other languages
Chinese (zh)
Other versions
CN115495853A (en
Inventor
史震云
黄皓
郑唯一
傅帅
李瑶
卜琛玥
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.)
Beihang University
Original Assignee
Beihang 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 Beihang University filed Critical Beihang University
Priority to CN202211442382.4A priority Critical patent/CN115495853B/en
Publication of CN115495853A publication Critical patent/CN115495853A/en
Application granted granted Critical
Publication of CN115495853B publication Critical patent/CN115495853B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/17—Mechanical parametric or variational design
    • 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
    • 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
    • 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
    • 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
    • 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
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computing Systems (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Molecular Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention belongs to the technical field of soft robots, and discloses a particle blocking variable stiffness module parameter optimization method and device, wherein the method comprises the following steps: constructing a mathematical model for deducing a mechanical property curve; respectively carrying out mechanical property test on the modules with different parameters and obtaining a mechanical property curve; obtaining a mathematical model characteristic vector corresponding to the parameter of the module according to the mechanical property curve; constructing and training a shallow neural network with module parameters as input and mathematical model feature vectors as output; inputting module parameters to a trained shallow neural network, and obtaining a predicted mechanical property curve; and screening the predicted values of the mechanical property curve according to the design requirements, and obtaining the module parameters corresponding to the selected predicted values. According to the method, the gray-box model is constructed to derive the mechanical property curve, the shallow neural network is used for predicting the mechanical property curve of the variable stiffness module, the parameter optimization design is realized, and the design and manufacturing cost of the variable stiffness module is effectively reduced.

Description

Particle blocking variable stiffness module parameter optimization method and device
Technical Field
The invention belongs to the technical field of soft robots, and particularly relates to a particle blocking variable stiffness module parameter optimization method and device.
Background
The robot actuator hand grip part often needs to carry out heavy load man-machine interaction, and the problems of man-machine interaction safety and load capacity are respectively solved by a traditional rigid hand grip and a flexible hand grip. If the large load capacity and the man-machine interaction safety of the robot gripper are achieved at the same time, a better feasible direction is to achieve the man-machine interaction part by adopting a variable stiffness technology. The rigidity changing technology of the existing robot gripper can be divided into particle blocking, layer interference, magnetic fluid rigidity changing, low-melting-point phase change materials, thermal activity intelligent materials, antagonistic rigidity changing and the like. Particle blocking is a good variable-rigidity solution, and the solutions are applied to the aspects of endoscope mechanical arms, variable-rigidity soft mechanical arms, rigidity-adjustable line driving operation arms and the like. However, the existing particle blocking variable stiffness technology is not mature enough, and particularly, the control model is not complete enough. Some researches on particle blocking modeling are to model the particle blocking variable stiffness module as an elastomer, neglect the elastic potential energy change of the membrane, have insufficient fineness of the model from the particle perspective, mostly make many simplified assumptions on the model, and therefore the variable stiffness performance of the module can only be qualitatively predicted, and the variable stiffness quantization continuously adjustable control mechanism cannot be quantitatively realized. Therefore, how to provide a mathematical model for parameter optimization design of a particle blocking variable stiffness module capable of being quantified is an urgent technical problem to be solved.
Disclosure of Invention
Aiming at the problems, the invention provides a particle blocking variable stiffness module parameter optimization method and a device, and the specific technical scheme is as follows:
in a first aspect, a method for optimizing parameters of a particle blocking variable stiffness module is provided, where the particle blocking variable stiffness module includes: the device comprises an upper flange cover, a silica gel film, a filter screen and a lower flange cover; the silica gel film and the filter screen are positioned between the upper flange cover and the lower flange cover, and the upper flange cover and the lower flange cover are connected through a screw and nut structure; the silica gel film is of a hollow structure, hard spherical particles are filled in the silica gel film, and the filter screen can prevent the spherical particles from falling into an air passage of the lower flange cover; the air passage of the lower flange cover is connected with a vacuum pump, so that air in the particle blocking variable stiffness module can be extracted; the parameter optimization method of the particle blocking variable stiffness module comprises the following steps:
s1: constructing a mathematical model for representing the relation between the pressure borne by the particle blocking variable stiffness module and the height of the particle blocking variable stiffness module and used for deducing a mechanical property curve; the mathematical model has a feature vector;
s2: respectively carrying out mechanical property tests on the particle blocking variable stiffness modules with different parameters and obtaining corresponding mechanical property curves;
s3: obtaining the feature vector of the mathematical model corresponding to the parameter of the particle blocking variable stiffness module according to the mechanical property curve;
s4: constructing a shallow neural network by taking the parameters of the particle blocking variable stiffness module as input and the characteristic vectors of the mathematical model as output, and training the shallow neural network, wherein the shallow neural network is used for predicting the mechanical property curve of the particle blocking variable stiffness module;
s5: inputting the parameters of the particle blocking variable stiffness module to be optimized into the trained shallow neural network to obtain a predicted mechanical property curve;
s6: and screening predicted values of the mechanical property curve according to design requirements, and obtaining the parameters of the particle blocking variable stiffness module corresponding to the selected predicted values.
Further, the particles block the parameters of the variable stiffness module
Figure 912652DEST_PATH_IMAGE001
(ii) a Wherein it is present>
Figure 999557DEST_PATH_IMAGE002
The thickness of the silica gel film is the thickness of the silica gel film,H r is the hardness of the silica gel film,ris the diameter of the hard spherical particles,μis the coefficient of friction of the hard spherical particles,ρis the mass density of the hard spherical particles,p negative is a vacuum negative pressure value; and/or the presence of a gas in the gas,
the mathematical model is an analyzable gray-box model, i.e.
Figure 489576DEST_PATH_IMAGE003
(ii) a Wherein the content of the first and second substances,Fin order for the particles to block the pressure experienced by the variable stiffness module,hto block the height of the variable stiffness module by the particles,
Figure 661931DEST_PATH_IMAGE004
is a feature vector of the mathematical model; the feature vector
Figure 879286DEST_PATH_IMAGE004
The method comprises 4 characteristic values, and is used for fitting and compensating irregular deformation of the silica gel film and particle friction dissipation deviation in the compression process.
Further, the process of respectively performing mechanical property tests on the particle blocking variable stiffness modules with different parameters and obtaining corresponding mechanical property curves comprises the following steps:
setting parameters
Figure 86276DEST_PATH_IMAGE005
And obtaining a corresponding first parameter matrixA Input (ii) a Said parameter
Figure 730884DEST_PATH_IMAGE005
Comprises the wall thickness of the silica gel film is->
Figure 390535DEST_PATH_IMAGE002
Hardness of the silica gel filmH r Diameter of said hard spherical particlesrOr the vacuum negative pressure valuep negative Set different values to form a parameter->
Figure 411581DEST_PATH_IMAGE005
A parameter value of (d);
for the first parameter matrixA Input The manufactured particle blocking variable stiffness module is subjected to mechanical property test;
generating a matrix corresponding to the first parameter according to the mechanical property test resultA Input Different parameters of
Figure 20548DEST_PATH_IMAGE005
And forming a first set of graphs of the mechanical properties.
Further, the process of obtaining the feature vector of the mathematical model corresponding to the parameter of the particle blocking variable stiffness module according to the mechanical property curve includes:
fitting the mechanical property curves in the first set of mechanical property curves by using a least square method to obtain the eigenvectors of the mathematical model corresponding to the parameters of the particle blocking variable stiffness module
Figure 39320DEST_PATH_IMAGE004
；
Obtaining the first parameter matrixA Input Corresponding first eigenvector matrixB Eigenvalues 。
Further, the shallow neural network comprises a neuron number ofnA hidden layer of (a); and training the shallow neural network by adopting a Bayesian regularization method.
Further, the process of inputting the parameters of the particle occlusion variable stiffness module to be optimized into the trained shallow neural network to obtain a predicted mechanical property curve includes:
setting parameters of a particle blocking variable stiffness module to be optimized
Figure 248584DEST_PATH_IMAGE006
And a corresponding second parameter matrix is obtained>
Figure 745425DEST_PATH_IMAGE007
；
The second parameter matrix
Figure 723745DEST_PATH_IMAGE007
Input into the trained shallow neural network and obtain the matrix &'s with the second parameter>
Figure 913418DEST_PATH_IMAGE007
Corresponding second eigenvector matrix->
Figure 609978DEST_PATH_IMAGE008
；
The second parameter matrix
Figure 910510DEST_PATH_IMAGE007
Is greater than or equal to>
Figure 494069DEST_PATH_IMAGE006
And the second eigenvector matrix
Figure 854643DEST_PATH_IMAGE008
The feature vector->
Figure 38500DEST_PATH_IMAGE009
The input to the mathematical model deduces the mechanical property curve and forms a second set of graphs of the mechanical property.
Further, the process of screening the predicted mechanical property curve according to the design requirement and obtaining the parameter of the particle blocking variable stiffness module corresponding to the selected mechanical property curve comprises the following steps:
selecting a mechanical property curve meeting the performance requirement from the second set of mechanical property curves;
obtaining parameters of the particle blocking variable stiffness module corresponding to the selected mechanical property curve
Figure 877143DEST_PATH_IMAGE006
。
In a second aspect, a particle blocking variable stiffness module parameter optimization device is provided, where the particle blocking variable stiffness module includes: the device comprises an upper flange cover, a silica gel film, a filter screen and a lower flange cover; the silica gel film and the filter screen are positioned between the upper flange cover and the lower flange cover, and the upper flange cover and the lower flange cover are connected through a screw and nut structure; the silica gel film is of a hollow structure, hard spherical particles are filled in the silica gel film, and the filter screen can prevent the spherical particles from falling into an air passage of the lower flange cover; the air passage of the lower flange cover is connected with a vacuum pump, so that air in the particle blocking variable stiffness module can be extracted; the parameter optimization device of the particle blocking variable stiffness module comprises:
the first unit can construct a mathematical model for deducing a mechanical property curve, wherein the mathematical model represents the relationship between the pressure borne by the particle blocking variable stiffness module and the height of the particle blocking variable stiffness module; the mathematical model has feature vectors;
the second unit can respectively perform mechanical property tests on the particle blocking variable stiffness modules with different parameters and obtain corresponding mechanical property curves;
the third unit can obtain a characteristic vector of a mathematical model corresponding to the parameter of the particle blocking variable stiffness module according to the mechanical property curve;
the fourth unit is capable of constructing and training a shallow neural network which takes the parameters of the particle occlusion variable stiffness module as input and the characteristic vectors of the mathematical model as output, wherein the shallow neural network is used for predicting the mechanical property curve of the particle occlusion variable stiffness module;
the fifth unit can input parameters of the particle blocking variable stiffness module to be optimized to the trained shallow neural network and obtain a predicted value of a mechanical property curve;
and the sixth unit can screen the predicted value of the mechanical property curve according to the design requirement and obtain the parameter of the particle blocking variable stiffness module corresponding to the selected predicted value.
Compared with the prior art, the invention can at least achieve the following technical effects:
according to the method, the ash box model is constructed to derive the mechanical property curve, the shallow neural network is used for predicting the mechanical property curve of the particle blocking variable stiffness module under different design parameters, the optimal design of the variable stiffness module parameters is realized according to actual requirements, and the economic cost and the time cost of designing and manufacturing the particle blocking variable stiffness module are effectively reduced.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic structural diagram of a particle blocking variable stiffness module according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of a parameter optimization method for a particle blocking variable stiffness module according to an embodiment of the present invention.
Fig. 3 is a second schematic diagram of a parameter optimization method of a particle blocking variable stiffness module according to an embodiment of the present invention.
Fig. 4 is a third schematic diagram of a parameter optimization method for a particle blocking variable stiffness module according to an embodiment of the present invention.
Fig. 5 is a fourth schematic diagram of a parameter optimization method for a particle blocking variable stiffness module according to an embodiment of the present invention.
Fig. 6 is a fifth schematic view of a parameter optimization method for a particle blocking variable stiffness module according to an embodiment of the present invention.
Fig. 7 is a sixth schematic view of a parameter optimization method for a particle blocking variable stiffness module according to an embodiment of the present invention.
Fig. 8 is a schematic diagram of a parameter optimization device of a particle blocking variable stiffness module according to an embodiment of the present invention.
In the figure: 1-upper flange cover, 2-silica gel film, 3-filter screen, 4-lower flange cover.
Detailed Description
In order to make those skilled in the art better understand the technical solution of the present invention, the technical solution in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only a part of the embodiments of the present specification, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments in the present description, shall fall within the scope of protection of the present invention.
The following describes a particle blocking variable stiffness parameter optimization scheme according to the present invention by way of specific examples.
Example 1
The invention aims to overcome the defects of the existing variable stiffness module parameter optimization method and the like, and provides a quantifiable particle blocking variable stiffness parameter optimization scheme. As shown in fig. 1, the particle blocking variable stiffness module in the embodiment of the present invention includes: an upper flange cover 1, a silica gel film 2, a filter screen 3 and a lower flange cover 4. The upper flange cover 1 and the lower flange cover 4 can compress the silica gel film 2 through a screw and nut structure to realize the sealing of the module; the silica gel film 2 is of a hollow structure, and hard spherical particles are filled in the silica gel film; the filter screen 3 is used for preventing spherical particles from falling into an air passage of the lower flange cover 4, and the air passage of the lower flange cover 4 is connected with a vacuum pump for pumping air in the module so that the module is in a particle blocking state and becomes hard.
Based on the mechanical structure design of the particle blocking variable stiffness module, the technical idea of the parameter optimization scheme of the particle blocking variable stiffness module is as follows: firstly, establishing an ash box model by analyzing the free energy of the silica gel film 2 and the dissipation energy of friction of spherical particles, wherein the ash box model is used for deducing a mechanical property curve of a particle blocking variable stiffness module; secondly, performing mechanical property experiments by obtaining parameters of the variable stiffness module, including the film thickness, the film hardness, the particle diameter, the particle mass density, the particle friction coefficient and the applied negative pressure value to obtain mechanical property experiment data, performing least square fitting on the mechanical property experiment data and the ash box model, and obtaining a feature vector of the ash box model, wherein the feature vector comprises 4 feature values; thirdly, parameters including the film thickness, the film hardness, the particle diameter, the particle mass density, the particle friction coefficient and the applied negative pressure value of different modules are used as input quantities, the feature vector of the ash box model containing 4 feature values is used as an output quantity, shallow neural network fitting is carried out, and the mechanical property curves of the modules under different parameters are predicted, so that the actual variable stiffness module parameter optimization design can be guided, and the mechanical property required by the actual engineering can be achieved.
Therefore, referring to fig. 2, a schematic diagram of a parameter optimization method of a particle blocking variable stiffness module according to an embodiment of the present invention is shown. Corresponding to the particle blocking variable stiffness module shown in fig. 1, the particle blocking variable stiffness module parameter optimization method comprises the following steps:
s1: constructing a mathematical model for representing the relation between the pressure borne by the particle blocking variable stiffness module and the height of the particle blocking variable stiffness module and used for deducing a mechanical property curve; the mathematical model has feature vectors.
Optionally, the particles block parameters of the variable stiffness module
Figure 830055DEST_PATH_IMAGE010
(ii) a Wherein it is present>
Figure 361531DEST_PATH_IMAGE002
The thickness of the silica gel film is the thickness of the silica gel film,H r is the hardness of the silica gel film,ris the diameter of the hard spherical particles,μis the coefficient of friction of the hard spherical particles,ρis the mass density of the hard spherical particles,p negative is the vacuum negative pressure value.
For the pressure value corresponding to the stiffness changing module in FIG. 1FHeight of the modulehAccording to the principle of virtual work, the gravity potential energy with smaller influence is simplified, and then the pressure does workW F =E Filim +W ParticleFriction Wherein, in the step (A),W F work is done for the pressure and the pressure is obtained,E Filim the elastic potential energy of the silica gel film 2 is changed,W ParticleFriction is the friction dissipation of the hard spherical particles.
Optionally based on the above principleA mathematical model of the mechanical property curve is represented, the mathematical model represents the relation between the pressure borne by the particle blocking variable stiffness module and the height of the particle blocking variable stiffness module and is used for deducing the mechanical property curve, and the mathematical model adopts an analyzable ash box model
Figure 767104DEST_PATH_IMAGE011
(ii) a Wherein the content of the first and second substances,Fblocking the pressure on the variable stiffness module for the particles;hblocking the height of the variable stiffness module for the particles; />
Figure 675017DEST_PATH_IMAGE004
The eigenvector for the mathematical model contains 4 eigenvalues for fitting to compensate for the particle friction dissipation bias due to irregular deformation of the film during compression and due to vacuum.
S2: and respectively carrying out mechanical property test on the particle blocking variable stiffness modules with different parameters and obtaining corresponding mechanical property curves.
Optionally, the S2 is shown in fig. 3, and the process of respectively performing mechanical property tests on the particle blocking variable stiffness modules with different parameters and obtaining corresponding mechanical property curves includes:
s211: setting the parameters
Figure 230239DEST_PATH_IMAGE005
And obtaining a corresponding first parameter matrixA Input (ii) a The parameter->
Figure 932616DEST_PATH_IMAGE005
Comprises the wall thickness of the silica gel film is->
Figure 559906DEST_PATH_IMAGE002
Hardness of the silica gel filmH r Diameter of said hard spherical particlesrAnd/or the vacuum negative pressure valuep negative Setting the parameters to different values
Figure 802669DEST_PATH_IMAGE005
A parameter value of (d); />
S212: for the first parameter matrixA Input The manufactured particle blocking variable stiffness module is subjected to mechanical property test;
s213: generating a matrix corresponding to the first parameter according to the mechanical property test resultA Input Different from said parameter
Figure 667857DEST_PATH_IMAGE005
And forming a first set of graphs of the mechanical properties.
It should be noted that, in consideration of effectively reducing the design and manufacturing cost, when S2 is used for respectively performing mechanical property tests on the particle blocking variable stiffness modules with different parameters, only a small number of particle blocking variable stiffness modules are manufactured, and corresponding parameters are obtained by the particle blocking variable stiffness modules
Figure 603452DEST_PATH_IMAGE005
. Thus, it can be understood that the first parameter matrix in this stepA Input And the parameters corresponding to the variable stiffness module are blocked by a small amount of particles.
S3: and obtaining the characteristic vector of the mathematical model corresponding to the parameter of the particle blocking variable stiffness module according to the mechanical property curve.
Optionally, the step S3 of obtaining the feature vector of the mathematical model corresponding to the parameter of the particle blocking variable stiffness module according to the mechanical property curve is shown in fig. 4, and includes:
s311: fitting the mechanical property curves in the first set of mechanical property curves by using a least square method to obtain the eigenvectors of the mathematical model corresponding to the parameters of the particle blocking variable stiffness module
Figure 921300DEST_PATH_IMAGE004
；
S312: obtaining the first parameter matrixA Input Corresponding first eigenvector matrixB Eigenvalues 。
S4: and constructing a shallow neural network by taking the parameters of the particle blocking variable stiffness module as input and the feature vectors of the mathematical model as output, and training the shallow neural network, wherein the shallow neural network is used for predicting the mechanical property curve of the particle blocking variable stiffness module.
Optionally, the shallow neural network comprises a neuron number ofnA hidden layer of (a); and/or training the shallow neural network by adopting a Bayesian regularization method. Specifically, as shown in fig. 5, the number of eigenvalues of the parameter as the input layer of the shallow neural network is 6, and the number of neurons in the hidden layer is 6n=10, the number of eigenvalues of the eigenvector of the mathematical model as the output layer of the shallow neural network is 4. The working principle of the shallow neural network is not described in this embodiment. It should be noted that the number of hidden layer neurons in the shallow neural network should be larger than that of the hidden layer neuronsmax(p,q) WhereinpIn order to input the size of the layer,qis the output layer size. The number of hidden layer neurons is not excessive, unnecessary computational power loss is increased due to excessive neuron number, and the fitting precision and the generalization capability of the shallow layer neural network are not improved.
S5: inputting the parameters of the particle occlusion variable stiffness module to the trained shallow neural network, and obtaining the predicted mechanical property curve.
Optionally, S5, inputting the parameters of the particle occlusion variable stiffness module to the trained shallow neural network and obtaining the predicted mechanical property curve as shown in fig. 6, includes:
s511: setting the parameters
Figure 718486DEST_PATH_IMAGE006
And obtaining a corresponding second parameter valueParameter matrix
Figure 438181DEST_PATH_IMAGE007
；
S512: the second parameter matrix
Figure 544677DEST_PATH_IMAGE007
Input into the trained shallow neural network and obtain the matrix ^ and the second parameter>
Figure 349822DEST_PATH_IMAGE007
Corresponding second eigenvector matrix->
Figure 199966DEST_PATH_IMAGE008
；
S513: the second parameter matrix
Figure 774167DEST_PATH_IMAGE007
Is greater than or equal to>
Figure 51565DEST_PATH_IMAGE006
And the second eigenvector matrix corresponding to the parameter @>
Figure 78426DEST_PATH_IMAGE008
The feature vector in (a)>
Figure 482994DEST_PATH_IMAGE009
Input into the mathematical model to derive the mechanical property curve->
Figure 911701DEST_PATH_IMAGE012
And a second set of plots of mechanical properties is formed.
It should be noted that, in view of the fact that after the steps S1 to S4 are performed, when S5 inputs the parameters of the particle occlusion variable stiffness module to the trained shallow neural network and obtains the predicted mechanical property curve, the method is expanded to manufacture a large number of particle occlusion variable stiffness modules and obtain corresponding parametersNumber of
Figure 360000DEST_PATH_IMAGE006
. It can thus be appreciated that the second parameter matrix ≦ in this step>
Figure 936475DEST_PATH_IMAGE007
And the parameters corresponding to a plurality of particle blocking variable stiffness modules.
S6: and screening the predicted value of the mechanical property curve according to the design requirement, and obtaining the parameter of the particle blocking variable stiffness module corresponding to the selected predicted value.
Optionally, the step S6 of screening the predicted mechanical property curve according to the design requirement and obtaining the parameter of the particle blocking variable stiffness module corresponding to the selected mechanical property curve is shown in fig. 7, and includes:
s611: selecting the mechanical property curve satisfying a performance requirement from the second set of mechanical property curves;
s612: obtaining the parameters of the particle blocking variable stiffness module corresponding to the selected mechanical property curve
Figure 66105DEST_PATH_IMAGE006
。
On the basis of structural design of the particle blocking variable stiffness module, a mathematical model of a gray box model is constructed based on a virtual work principle; the variable stiffness module is designed and manufactured in a small amount, parameter values of the variable stiffness module are obtained for experimental testing, and a least square method is used for comparing a gray box model with actualF-hFitting the mechanical property curve to obtain a fitting characteristic value vector of the model; a shallow neural network is trained by designing and manufacturing variable stiffness modules in large quantity and acquiring corresponding parameters as input and fitting eigenvalue vectors as output, and the mechanical property of the modules under design parameters is predicted by the network. And (4) inputting a vector matrix based on the practically selectable model to obtain a corresponding mechanical property curve, screening out required design parameters according to practical requirements, and finishing parameter optimization design.
Example 2
Fig. 8 is a schematic structural diagram of a parameter optimization device of a particle blocking variable stiffness module according to an embodiment of the present invention. Referring to fig. 8, in an embodiment, the particle blocking variable stiffness module parameter optimization apparatus includes:
the particle blocking variable stiffness module comprising: the device comprises an upper flange cover 1, a silica gel film 2, a filter screen 3 and a lower flange cover 4; the silica gel film 2 and the filter screen 3 are positioned between the upper flange cover 1 and the lower flange cover 4, and the upper flange cover 1 and the lower flange cover 4 are connected through a screw and nut structure; the silica gel film 2 is of a hollow structure, hard spherical particles are filled in the silica gel film, and the filter screen 3 can prevent the spherical particles from falling into an air passage of the lower flange cover 4; the air passage of the lower flange cover 4 is connected with a vacuum pump, so that air in the particle blocking variable stiffness module can be extracted; the parameter optimization device corresponding to the particle blocking variable stiffness module comprises:
the first unit can construct a mathematical model for deducing a mechanical property curve, which represents the relationship between the pressure borne by the particle blocking variable stiffness module and the height of the particle blocking variable stiffness module; the mathematical model has feature vectors;
the second unit can respectively perform mechanical property tests on the particle blocking variable stiffness modules with different parameters and obtain corresponding mechanical property curves;
a third unit capable of obtaining the feature vector of the mathematical model corresponding to the parameter of the particle blocking variable stiffness module according to the mechanical property curve;
a fourth unit, configured to construct and train a shallow neural network that takes the parameter of the particle occlusion variable stiffness module as an input and the feature vector of the mathematical model as an output, where the shallow neural network is used to predict the mechanical property curve of the particle occlusion variable stiffness module;
the fifth unit can input parameters of the particle blocking variable stiffness module to be optimized to the trained shallow neural network and obtain a predicted value of a mechanical property curve;
and the sixth unit can screen the predicted value of the mechanical property curve according to the design requirement and obtain the parameter of the particle blocking variable stiffness module corresponding to the selected predicted value.
It should be understood that, the particle blocking variable stiffness module parameter optimization device according to the embodiment of the present invention may also execute the method executed by the particle blocking variable stiffness module parameter optimization device (or apparatus) in fig. 1 to 7, and implement the functions of the particle blocking variable stiffness module parameter optimization device (or apparatus) in the examples shown in fig. 1 to 7, which are not described herein again.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of other like elements in a process, method, article, or apparatus comprising the element.
The embodiments of the present invention are described in a progressive manner, and the same and similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

Claims (4)

1. A particle blocking variable stiffness module parameter optimization method, the particle blocking variable stiffness module comprising: the device comprises an upper flange cover (1), a silica gel film (2), a filter screen (3) and a lower flange cover (4); the silica gel film (2) and the filter screen (3) are positioned between the upper flange cover (1) and the lower flange cover (4), and the upper flange cover (1) and the lower flange cover (4) are connected through a screw and nut structure; the silica gel film (2) is of a hollow structure, hard spherical particles are filled in the silica gel film, and the filter screen (3) can prevent the spherical particles from falling into an air passage of the lower flange cover (4); the air flue of the lower flange cover (4) is connected with a vacuum pump and can extract air in the particle blocking variable stiffness module, and the method for optimizing the parameters of the particle blocking variable stiffness module is characterized by comprising the following steps:
s1: constructing a mathematical model for representing the relation between the pressure borne by the particle blocking variable stiffness module and the height of the particle blocking variable stiffness module and used for deducing a mechanical property curve; the mathematical model has feature vectors;
s2: respectively carrying out mechanical property test on the particle blocking variable stiffness modules with different parameters and obtaining corresponding mechanical property curves;
s3: obtaining the characteristic vector of the mathematical model corresponding to the parameter of the particle blocking variable stiffness module according to the mechanical property curve;
s4: constructing a shallow neural network by taking the parameters of the particle occlusion variable stiffness module as input and the characteristic vectors of the mathematical model as output, and training the shallow neural network, wherein the shallow neural network is used for predicting a mechanical property curve of the particle occlusion variable stiffness module;
s5: inputting the parameters of the particle blocking variable stiffness module to be optimized into the trained shallow neural network to obtain a predicted mechanical property curve;
s6: screening a predicted value of a mechanical property curve according to design requirements and obtaining parameters of a particle blocking variable stiffness module corresponding to the selected predicted value;
the particles block the parameters of the variable stiffness module
Figure 600425DEST_PATH_IMAGE001
(ii) a Wherein the content of the first and second substances,
Figure 16975DEST_PATH_IMAGE002
is the wall of the silica gel filmThe thickness of the composite material is thick,H r is the hardness of the silica gel film,ris the diameter of the hard spherical particles,μis the coefficient of friction of the hard spherical particles,ρis the mass density of the hard spherical particles,p negative is a vacuum negative pressure value; and/or the presence of a gas in the gas,
the mathematical model is an analyzable gray box model, i.e.
Figure 368322DEST_PATH_IMAGE003
(ii) a Wherein the content of the first and second substances,Fin order for the particles to block the pressure experienced by the variable stiffness module,hto block the height of the variable stiffness module by the particles,
Figure 455227DEST_PATH_IMAGE004
is a feature vector of the mathematical model; the feature vector
Figure 132196DEST_PATH_IMAGE004
The method comprises 4 characteristic values, wherein the 4 characteristic values are used for fitting and compensating irregular deformation of the silica gel film (2) and particle friction dissipation deviation in the compression process;
the process of respectively carrying out mechanical property test on the particle blocking variable stiffness modules with different parameters and obtaining corresponding mechanical property curves comprises the following steps:
setting parameters
Figure 304551DEST_PATH_IMAGE005
And obtaining a corresponding first parameter matrixA Input (ii) a Said parameter
Figure 194010DEST_PATH_IMAGE005
The different parameter values of (A) include the thickness of the silica gel film wall
Figure 401000DEST_PATH_IMAGE002
Hardness of the silica gel filmH r Diameter of said hard spherical particlesrOr the vacuum is negativePressure valuep negative Setting parameters formed by different values
Figure 248870DEST_PATH_IMAGE005
A parameter value of (d);
for the first parameter matrixA Input The manufactured particle blocking variable stiffness module is subjected to mechanical property test;
generating a matrix corresponding to the first parameter according to the mechanical property test resultA Input Different parameters of
Figure 642942DEST_PATH_IMAGE005
And forming a first set of graphs of mechanical properties;
the process of obtaining the feature vector of the mathematical model corresponding to the parameter of the particle blocking variable stiffness module from the mechanical property curve comprises:
fitting the mechanical property curves in the first set of mechanical property curves by using a least square method to obtain the eigenvectors of the mathematical model corresponding to the parameters of the particle blocking variable stiffness module
Figure 601671DEST_PATH_IMAGE004
；
Obtaining the first parameter matrixA Input Corresponding first eigenvector matrixB Eigenvalues ；
Inputting the parameters of the particle blocking variable stiffness module to be optimized into the trained shallow neural network to obtain a predicted mechanical property curve, wherein the process comprises the following steps:
setting parameters of a particle blocking variable stiffness module to be optimized
Figure 397589DEST_PATH_IMAGE006
And obtaining a corresponding second parameter matrix
Figure 150781DEST_PATH_IMAGE007
；
The second parameter matrix
Figure 297729DEST_PATH_IMAGE007
Inputting to the trained shallow neural network and obtaining the second parameter matrix
Figure 60148DEST_PATH_IMAGE007
Corresponding second eigenvector matrix
Figure 710573DEST_PATH_IMAGE008
；
The second parameter matrix
Figure 634666DEST_PATH_IMAGE007
Parameter (2) of
Figure 3331DEST_PATH_IMAGE006
And the second eigenvector matrix
Figure 38283DEST_PATH_IMAGE008
The feature vector of
Figure 74372DEST_PATH_IMAGE009
And inputting the data into the mathematical model to derive a mechanical property curve, and forming a second set of graphs of the mechanical property curve.
2. The particle blocking variable stiffness module parameter optimization method of claim 1, wherein the shallow neural network comprises a neuron number ofnA hidden layer of (a); and training the shallow neural network by adopting a Bayesian regularization method.
3. The parameter optimization method of the particle blocking variable stiffness module according to claim 1, wherein the process of screening a predicted mechanical property curve according to design requirements and obtaining parameters of the particle blocking variable stiffness module corresponding to the selected mechanical property curve comprises:
selecting a mechanical property curve meeting the performance requirement from the second set of mechanical property curves;
obtaining parameters of the particle blocking variable stiffness module corresponding to the selected mechanical property curve
Figure 169367DEST_PATH_IMAGE006
。
4. An apparatus for implementing the particle blocking variable stiffness module parameter optimization method of claim 1, the particle blocking variable stiffness module comprising: the device comprises an upper flange cover (1), a silica gel film (2), a filter screen (3) and a lower flange cover (4); the silica gel film (2) and the filter screen (3) are positioned between the upper flange cover (1) and the lower flange cover (4), and the upper flange cover (1) is connected with the lower flange cover (4) through a screw and nut structure; the silica gel film (2) is of a hollow structure, hard spherical particles are filled in the silica gel film, and the filter screen (3) can prevent the spherical particles from falling into an air passage of the lower flange cover (4); the air flue of lower flange lid (4) is connected the vacuum pump and can is extracted the air in the granule blocks the variable rigidity module, its characterized in that, the parameter optimization device of granule blocks variable rigidity module includes:
the first unit can construct a mathematical model for deducing a mechanical property curve, wherein the mathematical model represents the relationship between the pressure borne by the particle blocking variable stiffness module and the height of the particle blocking variable stiffness module; the mathematical model has feature vectors;
the second unit can respectively perform mechanical property tests on the particle blocking variable stiffness modules with different parameters and obtain corresponding mechanical property curves;
the third unit can obtain a characteristic vector of a mathematical model corresponding to the parameter of the particle blocking variable stiffness module according to the mechanical property curve;
the fourth unit is capable of constructing and training a shallow neural network which takes the parameters of the particle blocking variable stiffness module as input and the characteristic vectors of the mathematical model as output, wherein the shallow neural network is used for predicting the mechanical property curve of the particle blocking variable stiffness module;
the fifth unit can input parameters of the particle blocking variable stiffness module to be optimized to the trained shallow neural network and obtain a predicted value of a mechanical property curve;
and the sixth unit can screen the predicted value of the mechanical property curve according to the design requirement and obtain the parameter of the particle blocking variable stiffness module corresponding to the selected predicted value.
CN202211442382.4A 2022-11-18 2022-11-18 Particle blocking variable stiffness module parameter optimization method and device Active CN115495853B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211442382.4A CN115495853B (en) 2022-11-18 2022-11-18 Particle blocking variable stiffness module parameter optimization method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211442382.4A CN115495853B (en) 2022-11-18 2022-11-18 Particle blocking variable stiffness module parameter optimization method and device

Publications (2)

Publication Number Publication Date
CN115495853A CN115495853A (en) 2022-12-20
CN115495853B true CN115495853B (en) 2023-03-24

Family

ID=85116073

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211442382.4A Active CN115495853B (en) 2022-11-18 2022-11-18 Particle blocking variable stiffness module parameter optimization method and device

Country Status (1)

Country Link
CN (1) CN115495853B (en)

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108481318B (en) * 2018-03-07 2021-07-06 河南工业大学 Variable-rigidity driving rod based on granular body blocking theory and working system thereof
CN108527372B (en) * 2018-06-22 2020-10-27 中国科学技术大学 Robot joint self-adaptive control method of variable-rigidity series elastic driver
CN112182933A (en) * 2020-09-28 2021-01-05 西北工业大学 Modeling and optimal design method of variable-rigidity composite material cylindrical shell
CN114264440B (en) * 2021-11-17 2024-04-16 南京邮电大学 Variable-rigidity flexible touch sensor system and control method thereof

Also Published As

Publication number Publication date
CN115495853A (en) 2022-12-20

Similar Documents

Publication Publication Date Title
Peng et al. Process parameter optimization for fused deposition modeling using response surface methodology combined with fuzzy inference system
Yıldız A novel particle swarm optimization approach for product design and manufacturing
US7219087B2 (en) Soft computing optimizer of intelligent control system structures
Wang A unit cell approach for lightweight structure and compliant mechanism
Lu et al. Design and optimization of interval type-2 fuzzy logic controller for delta parallel robot trajectory control
Kitayama et al. Optimization of variable blank holder force trajectory by sequential approximate optimization with RBF network
Pham et al. Optimisation of a fuzzy logic controller for a flexible single-link robot arm using the Bees Algorithm
Shen et al. Vibration control of flexible structures with PZT sensors and actuators
Lilly et al. Dynamic simulation and neural network compliance control of an intelligent forging center
CN115495853B (en) Particle blocking variable stiffness module parameter optimization method and device
Kayacan et al. Velocity control of a spherical rolling robot using a grey-PID type fuzzy controller with an adaptive step size
Endelt et al. Iterative learning and feedback control applied on a deep drawing process
Ata et al. Half car suspension system integrated with PID controller
Du et al. Static output feedback control for electrohydraulic active suspensions via T–S fuzzy model approach
Wikman et al. Estimation of constitutive parameters for powder pressing by inverse modelling
Sun et al. Hybrid model predictive control of damping multi-mode switching damper for vehicle suspensions
Liu et al. Active suspension control design using a combination of LQR and backstepping
Kharola A PID based ANFIS & fuzzy control of inverted pendulum on inclined plane (IPIP)
Tinkir et al. Investigation of the effect of hydromechanical deep drawing process parameters on formability of AA5754 sheets metals by using neuro-fuzzy forecasting approach
Ömürlü et al. Application of fuzzy PID control to cluster control of viaduct road vibrations
Darsivan et al. Active engine mounting control algorithm using neural network
Rajak et al. On the influence of porosity and pore size on AlSi17 alloy foam using artificial neural network
Xinliang et al. Sliding mode force control of an electrohydraulic servo system with RBF neural network compensation
Matthews et al. Hierarchical Design of Composite Materials With Negative Stiffness Inclusions Using a Bayesian Network Classifier
Al-awad Genetic Algorithm Control of Model Reduction Passive Quarter Car Suspension System.

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