CN112926152A - Accurate control and optimization method for clamping force of thin-walled part driven by digital twin - Google Patents

Accurate control and optimization method for clamping force of thin-walled part driven by digital twin Download PDF

Info

Publication number
CN112926152A
CN112926152A CN202110186148.9A CN202110186148A CN112926152A CN 112926152 A CN112926152 A CN 112926152A CN 202110186148 A CN202110186148 A CN 202110186148A CN 112926152 A CN112926152 A CN 112926152A
Authority
CN
China
Prior art keywords
thin
wall part
clamping
clamp
deformation
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
CN202110186148.9A
Other languages
Chinese (zh)
Other versions
CN112926152B (en
Inventor
张映锋
王刚
税浩轩
曹彦生
李适
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.)
Northwestern Polytechnical University
Beijing Xinfeng Aerospace Equipment Co Ltd
Original Assignee
Northwestern Polytechnical University
Beijing Xinfeng Aerospace Equipment Co Ltd
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 Northwestern Polytechnical University, Beijing Xinfeng Aerospace Equipment Co Ltd filed Critical Northwestern Polytechnical University
Priority to CN202110186148.9A priority Critical patent/CN112926152B/en
Publication of CN112926152A publication Critical patent/CN112926152A/en
Application granted granted Critical
Publication of CN112926152B publication Critical patent/CN112926152B/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
    • G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • 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
    • 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]

Abstract

The invention provides a method for accurately controlling and optimizing clamping force of a thin-wall part driven by digital twins, which comprises the steps of firstly obtaining the clamping force and clamping deformation data of a thin-wall part manufacturing main body, and realizing means such as configuring various sensors for the thin-wall part, a clamp and a machine tool by adopting the technology of the Internet of things; on the basis of acquiring data, establishing a thin-wall part geometric model, a finite element deformation simulation model, a thin-wall part clamping deformation prediction and other digital twin virtual models, and mapping and simulating the thin-wall part clamping process in real time; and finally, establishing a thin-wall part clamping force optimization model, optimizing the clamping force of the thin-wall part to form a thin-wall part clamping force selection rule, and improving the machining precision of the thin-wall part. The thin-wall part clamping device can solve the problem that the clamping force range meeting the clamping deformation error is difficult to determine in the thin-wall part clamping process.

Description

Accurate control and optimization method for clamping force of thin-walled part driven by digital twin
Technical Field
The invention relates to the technical field of thin-wall part clamping, in particular to a method for accurately controlling and optimizing clamping force of a thin-wall part driven by digital twins.
Background
Parts with wall thickness or ratio of plate thickness to profile dimension less than 1:20 are referred to as thin-walled parts. The thin-wall part has the advantages of high specific strength, light relative weight and the like, and is widely applied to the field of aerospace. However, the thin-wall part is easy to generate large deformation in the processing process due to the thin wall thickness and the general weak rigidity. The clamping quality is one of the main factors influencing the deformation of the thin-wall part in the machining process and directly influences the performance of the workpiece. However, the clamping force in the existing clamping process is determined by the experience of operators, so that the defects of large randomness, non-uniform standard, difficult optimization and the like exist, the clamping force is difficult to be accurately controlled, and the fluctuation of the quality of thin-walled parts is aggravated. Therefore, how to accurately control the clamping force has important theoretical value for improving the processing quality of thin-walled parts.
Chinese patent thin-wall part residual stress deformation perception prediction method based on clamping force monitoring (201710964936X) discloses a thin-wall part residual stress deformation perception prediction method based on clamping force monitoring, the method firstly predicts the residual stress deformation trend of the thin-wall part by a finite element simulation method, and adds a clamping force perception point in a large deformation area; then designing a sensing clamp, and monitoring the change of clamping force in the machining process at a sensing point through a pressure sensor; and finally, by establishing a finite element model of the static fixed base of the clamping system and applying the counter force of the change value of the clamping force at the sensing point, the residual stress deformation of the part is obtained, and the prediction of the residual stress deformation of the thin-wall part is realized. The method has an effect on the perception prediction of the residual stress deformation of the thin-wall part, but because the mapping relation between the clamping force and the actual clamping deformation is not considered, and further the clamping force is not optimally controlled, the clamping force range meeting the clamping deformation error cannot be obtained, the clamping force is difficult to accurately control, and the actual use requirement cannot be met.
Disclosure of Invention
In order to solve the problem that the clamping force range meeting the clamping deformation error is difficult to determine in the thin-wall part clamping process, the invention provides a method for accurately controlling and optimizing the clamping force of a thin-wall part driven by a digital twin.
The basic concept of the invention is as follows:
firstly, acquiring clamping force and clamping deformation data of a thin-wall part manufacturing main body, and configuring various sensors for thin-wall parts, clamps and machine tools by means of the Internet of things technology; on the basis of acquiring data, establishing a thin-wall part geometric model, a finite element deformation simulation model, a thin-wall part clamping deformation prediction and other digital twin virtual models, and mapping and simulating the thin-wall part clamping process in real time; and finally, establishing a thin-wall part clamping force optimization model, optimizing the clamping force of the thin-wall part to form a thin-wall part clamping force selection rule, and improving the machining precision of the thin-wall part.
The technical scheme of the invention is as follows:
the method for accurately controlling and optimizing the clamping force of the thin-walled part driven by the digital twin comprises the following steps of:
step 1: acquiring offline data of the thin-wall part and the clamp before clamping and online data of the thin-wall part in the clamping process; the off-line data comprises geometric parameters and material properties; the online data comprises input torque applied to the clamp, deformation displacement data of a monitoring point position arranged on the thin-wall part, and clamping environment data;
step 2: establishing a physical space model and a virtual space model of the thin-wall part-clamp based on a digital twinning technology; the virtual space model comprises a geometric information model, a finite element simulation model and a thin-wall part clamping deformation prediction model
And step 3: on the basis of the step 1 and the step 2, a precise control and optimization process of the clamping force of the thin-wall part based on the digital twin is executed:
step 3.1: transmitting the clamping process data of the real thin-walled workpiece in the physical space to a physical space model, calling real-time information of the physical space model by using a virtual space model, and displaying clamping process pictures in real time through human-computer interaction equipment;
step 3.2: performing analog simulation in a virtual space model of the thin-wall part-clamp, and predicting the deformation of the thin-wall part according to a thin-wall part clamping deformation prediction model:
step 3.3: after the thin-wall part clamping activity is finished, the full-flow data is stored in a thin-wall part-clamp physical space model, and on the basis of a thin-wall part clamping deformation prediction model, a group of optimal positioning points and clamping force parameters are obtained through an optimization algorithm to enable the clamping deformation to be minimum, so that the size of the initial clamping force of the thin-wall part is determined.
Further, in the step 1, the internet of things technology is applied to the thin-wall part clamping process, a torque wrench and sensor equipment are correspondingly configured on the thin-wall part and the clamp, offline data of the thin-wall part and the clamp before clamping is obtained, monitoring points are set, and online data in the thin-wall part clamping process are obtained.
Further, step 1 specifically includes the following steps:
step 1.1: before clamping, detecting a thin-wall part to be clamped and a clamp to obtain geometric parameters and material attributes of the thin-wall part and the clamp;
step 1.2: the method comprises the steps of acquiring input torque applied to a clamp by using a torque wrench, measuring clamping force of a contact position of a thin-wall part and the clamp by using a pressure sensor, establishing acquisition points on the thin-wall part, arranging distance measurement sensing equipment to acquire deformation displacement data of monitoring points of the thin-wall part, and configuring temperature and humidity sensors in a clamping environment to acquire clamping environment data.
Further, step 2 specifically includes the following steps:
step 2.1: establishing a physical space model of the thin-wall part-clamp, converting a thin-wall part clamping system from a physical entity into a Web identified data model, and tracking and calling thin-wall part-clamp information;
step 2.2: and constructing a thin-wall part-clamp virtual space model which comprises a geometric information model, a finite element simulation model and a thin-wall part clamping deformation prediction model, and mapping the clamping deformation state of the thin-wall part in real time.
Further, step 2.2 specifically includes the following steps:
step 2.2.1: constructing a geometric model of the thin-wall part and the clamp by adopting three-dimensional software according to the data acquired in the step 1;
step 2.2.2: adopting finite element software to construct a finite element simulation model of deformation of the thin-wall part;
step 2.2.3: based on a deep neural network, a thin-wall part clamping deformation prediction model is established, a nonlinear mapping relation between the clamping force and the clamping deformation is established, deformation data after the simulation of a finite element simulation model is used as a training sample set of the thin-wall part clamping deformation prediction model, and the thin-wall part clamping deformation prediction model is trained.
Further, step 2.2.2 specifically includes the following steps:
a. importing a geometric model: importing the thin-wall part and the geometric model of the clamp built in the step 2.2.1 into finite element software, wherein the clamp is set as an analytic rigid body, and the thin-wall part is set as an elastic body with isotropy;
b. setting material properties: reading basic information of thin-wall parts and clamp objects of the physical space model established in the step 2.1, and setting material attributes of the thin-wall parts and the clamps;
c. assembling: assembling the thin-wall part and the clamp in finite element software according to an actual clamping mode, setting a contact position, and defining the thin-wall part and the clamp by adopting 'surface-to-surface contact';
d. boundary conditions and loads: the six degrees of freedom of the bottom surface of the thin-wall part are completely restrained, and a load is applied to the contact position of the thin-wall part and a clamp;
e. grid division: and selecting hexahedrons, sweep-Jones and centerline-axis algorithms for the grid control attributes, and dividing the grid by adopting a C3D20R unit.
Further, step 2.2.3 specifically includes the following steps:
step 2.2.3.1: compiling a script file according to the finite element simulation model established in the step 2.2.2 by using a parameterized analysis function of finite element software, and carrying out finite element calculation by changing the contact position coordinates and the clamping force magnitude parameters to obtain thin-wall part deformation data corresponding to the measuring points;
step 2.2.3.2: determining the input and the output of a thin-wall part clamping deformation prediction model; the thin-wall part clamping deformation prediction model input unit is a coordinate parameter of a positioning point and a clamping load applied to the thin-wall part, and is expressed as x ═ xk}(1≤k≤K)，xk＝{sk,Fk}，skRepresenting the coordinates of the anchor point in the kth sample, FkRepresents the clamp load in the kth sample; the thin-wall part clamping deformation prediction model output unit is clamping deformation data of thin-wall part measuring points and is expressed as y ═ yk}(1≤k≤K)；
Step 2.2.3.3: establishing a deep neural network with L layers, wherein the node numbers of an input layer, an i-th hidden layer and an output layer are r and L respectivelyiAnd c, selecting the tanh function as the activation function, and expressing the function as follows:
Figure BDA0002943153030000041
the error function selects a square loss function;
step 2.2.3.4: and performing gradient calculation by adopting a back propagation algorithm, correcting the weight parameter and the bias parameter of the deep neural network, and ending iteration when the iteration times or the error is less than or equal to a preset value to obtain a trained thin-wall part clamping deformation prediction model.
Further, in step 2.2.3, the sample is normalized: assuming the mean and standard deviation of the input features over the entire sample data set is u, the normalization operation is to subtract u from each value of the input features and then divide by the standard deviation.
Further, the step 3.2 comprises the following specific steps:
step 3.2.1: the action and system control instructions in the thin-wall part clamping process are transmitted to the thin-wall part-clamp virtual space model in real time, the thin-wall part-clamp virtual space model simulates the thin-wall part clamping deformation, and three-dimensional visual simulation of the thin-wall part clamping action is realized;
step 3.2.2: the method comprises the following steps of predicting clamping deformation of the thin-wall part through a thin-wall part clamping deformation prediction model:
a. according to the formula F ═ kMnInput torque M collected by torque wrenchnThe transformation into a clamping load F exerted on the surface of the thin-walled part, where k is the transformation coefficient, is determined by the following experiment: when the torque wrench inputs the torque MnAcquiring a clamping force value F ' through the pressure sensors arranged in the step 1.2 to obtain a quotient value k ', performing multiple tests to obtain multiple quotient values k ', and taking an average value as a conversion coefficient k;
b. inputting the converted clamping load and the positioning position coordinate into a thin-wall part clamping deformation prediction model to obtain a predicted value of clamping deformation; and comparing the data with historical data under the operating condition, updating the thin-wall part clamping deformation prediction model when the error between the data and the historical data exceeds an allowable value, wherein the updating step is to firstly add the real-time clamping deformation data into a training set of the thin-wall part clamping deformation prediction model, and then train the thin-wall part clamping deformation prediction model to obtain the updated prediction model.
Further, the optimization algorithm in step 3.3 adopts a genetic algorithm, and an objective function of the genetic algorithm is defined as:
min f(NET,s,F)
Figure BDA0002943153030000051
wherein f represents a thin-wall part clamping deformation prediction model, and NET represents the prediction modelThe established neural network, S represents the coordinates of the positioning points, S represents the point set of the positioning points, F represents the magnitude of clamping force, and F represents the magnitude of the clamping forceminAnd FmaxRepresents the minimum and maximum values of the clamping force;
randomly generating an initial population comprising p individuals; decoding the individuals one by one, eliminating the individuals which do not meet the constraint conditions in the optimization model, and predicting clamping deformation of the individuals which meet the constraint conditions through a deep neural network;
the individual fitness is defined as
Figure BDA0002943153030000052
In the above formula, Ui(i is not less than 1 and not more than p) is the individual dyed in the population, and delta is a value predetermined according to the objective function value of each generation.
Advantageous effects
Compared with the prior art, the invention has the advantages that: the technology of the Internet of things is applied to a manufacturing workshop, and the clamping real-time data of the thin-wall part is collected, so that the active sensing of the clamping process of the thin-wall part is realized; introducing a digital twinning technology, constructing a digital twinning model of the thin-wall part-clamp, and performing real-time simulation to obtain a deformation state in clamping of the thin-wall part, thereby performing dimensionality expansion on physical data; the deep neural network can well model the situations that the coupling mechanism among factors is not clear and the exact relation between input and output is difficult to determine through extremely strong self-adaption and self-learning capabilities, so that the deformation of the thin-wall part can be accurately and quickly predicted, the clamping force which enables the clamping deformation to be minimum is obtained by applying a genetic algorithm, a thin-wall part clamping force selection rule is formed, and the machining precision of the thin-wall part is improved.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a technical roadmap for the present invention;
FIG. 2 is a work flow diagram of the method of the present invention;
FIG. 3 is a geometric model diagram of a clamp three-jaw chuck;
FIG. 4 is a geometric model view of a thin-walled part;
FIG. 5 is a timing diagram of the physical-virtual space interaction of the method of the present invention.
Detailed Description
The following detailed description of embodiments of the invention is intended to be illustrative, and not to be construed as limiting the invention.
The invention designs a digital twin-driven thin-wall part clamping force accurate control optimization method, aiming at establishing a thin-wall part geometric model, a finite element deformation simulation model, a thin-wall part clamping deformation prediction model, a clamping force optimization model and other digital twin virtual models on the basis of acquiring the clamping force and the clamping deformation data of a thin-wall part manufacturing main body, mapping and simulating the thin-wall part clamping process in real time, optimizing the thin-wall part clamping force and finally improving the machining precision of the thin-wall part. The specific implementation steps are as follows:
step 1: the method is characterized in that the technology of the Internet of things is applied to the clamping process of the thin-wall part, a digital display torque wrench, a pressure sensor and other equipment are configured for physical main bodies such as the thin-wall part and a clamp, off-line data such as geometric parameters, material attributes and the like of the thin-wall part and the clamp before clamping is obtained, a proper acquisition point is set, and on-line data in the clamping process of the thin-wall part is obtained, and the method specifically comprises the following steps:
step 1.1: detecting a thin-wall part to be clamped and a clamp, and acquiring data such as geometric parameters, material attributes and the like of the thin-wall part;
step 1.2: the method comprises the steps of acquiring torque applied to a clamp by a digital display torque wrench, measuring clamping force of a contact position of a thin-wall part and the clamp by a pressure sensor, setting acquisition points on the thin-wall part, arranging a laser range finder distance measurement sensing device to acquire deformation displacement data of monitoring points of the thin-wall part, and finally configuring a temperature sensor and a humidity sensor in a clamping environment to acquire clamping processing environment data.
Step 2: based on a digital twinning technology, a physical space model and a virtual space model of a thin-wall part-clamp are established, and the method specifically comprises the following steps:
step 2.1: establishing a physical space model of the thin-wall part-clamp, converting a clamping system consisting of the thin-wall part-clamp into a Web identified data model from a physical entity, and tracking and calling thin-wall part-clamp information; the physical space model of the thin-wall part-clamp is represented by a five-dimensional tuple:
TCGPM＝{TCBI,GBI,SSL,EI,TCGSI}
in the formula, TCBI (Thin-walled component Basic Information) represents Basic Information of a Thin-walled object, gbi (Gig Basic Information) represents Basic Information of a jig object, ssl (smart Sensor list) represents a Sensor list configured in a Thin-walled part clamping system, ei (environmental Information) represents environmental Information, and TCGSI (Thin-walled component and Gig Status Information) represents real-time Status Information of the Thin-walled part-jig.
Step 2.2: and constructing a thin-wall part-clamp virtual space model which comprises a geometric model, a finite element simulation model and a thin-wall part clamping deformation prediction model, and mapping the clamping deformation state of the thin-wall part in real time.
Step 2.2.1: adopting SolidWorks three-dimensional software to construct a geometric model of the thin-wall part and the clamp according to the data obtained in the step 1;
step 2.2.2: and (3) constructing a finite element simulation model of the deformation of the thin-wall part by adopting Abaqus finite element software and applying a finite element statics cutting simulation technology. The method specifically comprises the following steps:
f. importing a geometric model: importing the thin-wall part and the geometric model of the clamp built in the step 2.2.1 into Abaqus software, setting the clamp as an analytic rigid body, and setting the thin-wall part as an elastic body with isotropy;
g. setting material properties: reading basic information of the thin-wall part and the clamp object of the physical space model established in the step 2.1, and setting material attributes such as elastic modulus, Poisson ratio, density and the like of the thin-wall part and the clamp;
h. assembling: assembling the thin-wall part and the clamp in finite element software according to an actual clamping mode, setting a contact position (positioning point), and defining the thin-wall part and the clamp by adopting 'surface-to-surface contact';
i. boundary conditions and loads: the six degrees of freedom of the bottom surface of the thin-wall part are completely constrained, so that the thin-wall part is guaranteed not to move in the loading process, and a load is applied to the contact position of the thin-wall part and a clamp;
j. grid division: selecting hexahedrons, sweep-Jones and centerline-axis algorithms for grid control attributes, and dividing grids by adopting a C3D20R unit;
step 2.2.3: based on a deep neural network, a thin-wall part clamping deformation prediction model is established so as to establish a nonlinear mapping relation between the clamping force and the clamping deformation. The finite element simulation is the simulation of clamping deformation under the condition of neglecting other factors except clamping force in a machining environment, namely under an ideal condition, and when actual clamping deformation data does not exist at the beginning or the actual clamping data is small in scale, the deformation data after the finite element simulation model is simulated is used as a training set of a thin-wall part clamping deformation prediction model to train the thin-wall part clamping deformation prediction model. The method specifically comprises the following steps:
step 2.2.3.1: compiling a script file according to the finite element simulation model established in the step 2.2.2 by using the parameterized analysis function of the Abaqus finite element software, and automatically performing finite element calculation by changing the coordinates of the contact position (positioning point) and the magnitude parameter of the clamping force to obtain thin-wall part deformation data corresponding to the measuring point;
step 2.2.3.2: determining the input and the output of a thin-wall part clamping deformation prediction model; the model input unit is a coordinate parameter of a positioning point and a clamping load applied to a thin-wall part, and is expressed as x ═ xk}(1≤k≤K)，xk＝{sk,Fk}，skRepresenting the coordinates of the anchor point in the kth sample, FkRepresents the clamp load in the kth sample; the model output unit is clamping deformation data of the thin-wall part measuring point and is expressed as y ═ yk}(1≤k≤K)；
Step 2.2.3.3: in order to ensure network convergence, input samples are subjected to standardization processing; assuming the mean and standard deviation of the input features over the entire data set is u and σ, the normalization operation is to subtract u from each value of the input features and divide by the standard deviation σ. After the standardization is finished, dividing the obtained data into two groups, wherein one group is a training sample, the data size of the training sample is 80% of the original data size, and the other group is a testing sample, the data size of the testing sample is 20% of the original data size;
step 2.2.3.4: establishing a deep neural network with L layers, wherein the node numbers of an input layer, an i-th hidden layer and an output layer are r and L respectivelyiAnd c, selecting the tanh function as the activation function, and expressing the function as follows:
Figure BDA0002943153030000081
the error function selects a square loss function;
step 2.2.3.5: and performing gradient calculation by adopting a back propagation algorithm, correcting the weight parameter and the bias parameter of the deep neural network, and ending iteration when the iteration times or the error is less than or equal to a preset value to obtain a trained thin-wall part clamping deformation prediction model.
And step 3: on the basis of the step 1 and the step 2, a precise control and optimization process of the clamping force of the thin-wall part based on digital twinning is executed, and the method specifically comprises the following steps:
step 3.1: transmitting the clamping process data of the real thin-walled workpiece in the physical space to a physical space model, calling real-time information of the physical space model by using a virtual space model, and displaying a clamping process picture in front of a manager in real time through human-computer interaction equipment (a computer and the like);
step 3.2: and performing analog simulation in the thin-wall part-clamp virtual space model, and predicting the deformation of the thin-wall part according to the prediction model. The method comprises the following specific steps:
step 3.2.1: the action and system control instructions in the thin-wall part clamping process are transmitted to the thin-wall part-clamp virtual space model in real time, the thin-wall part-clamp virtual space model carries out simulation of thin-wall part clamping deformation, and three-dimensional visual simulation of thin-wall part clamping behaviors is realized;
step 3.2.2: the method comprises the following steps of predicting clamping deformation of the thin-wall part through a thin-wall part clamping deformation prediction model:
c. the digital display torque wrench converts the input torque collected by the digital display torque wrench into a clamping load applied to the surface of a thin-wall part, and the conversion formula of the three-jaw chuck is as follows: k m ═ Fn，MnFor input torque, k is the conversion factor. The k value was determined by the following experiment: input torque M of torque wrench with digital displayn' obtaining the clamping force value F ' through the pressure sensor arranged in the step 1.2, thereby obtaining a quotient k '. K is averaged after multiple trials. The pressure sensor can not control the clamping force to be input, only can measure the clamping force and has large fluctuation of the clamping force measured in a complex processing environment, so that the pressure sensor is adopted to obtain the clamping force value only in an experiment with a single environment to determine the conversion coefficient. The stable digital display torque wrench is selected to control the input torque in the actual clamping process, so that the clamping force is controlled.
d. And inputting the converted clamping load and the positioning position coordinates into a prediction model to obtain a predicted value of clamping deformation. And comparing the data with historical data under the operating condition to verify the accuracy and effectiveness of the thin-wall part clamping deformation prediction model. When the error of the two exceeds an allowable value, updating the prediction model, wherein the updating step is to add the real-time clamping deformation data into a training set of the prediction model, and then train the prediction model to obtain the updated prediction model so as to ensure that the prediction model can accurately map the thin-wall clamping deformation;
step 3.3: after the thin-wall part clamping activity is finished, the full-flow data is stored in a thin-wall part-clamp physical space model, and on the basis of a thin-wall part clamping deformation prediction model, a group of optimal positioning points and clamping force parameters are obtained through a genetic algorithm to enable the clamping deformation to be minimum, so that the size of the initial clamping force of the thin-wall part is determined. The genetic algorithm objective function is defined as:
min f(NET,s,F)
Figure BDA0002943153030000101
wherein F represents a thin-wall part clamping deformation prediction model, NET represents a neural network established in the prediction model, S represents coordinates of positioning points, S represents a point set where the positioning points are located, F represents the clamping force, and F represents the clamping forceminAnd FmaxIndicating the minimum and maximum values of the clamping force.
An initial population containing p individuals was randomly generated. And decoding the individuals one by one, wherein the individuals which do not meet the constraint condition in the optimization model are removed, and the clamping deformation of the individuals which meet the constraint condition can be predicted by the deep neural network.
The individual fitness is defined as
Figure BDA0002943153030000102
In the above formula, Ui(i is not less than 1 and not more than p) is the individual dyed in the population, and delta is a relatively large value which is preset according to the objective function value of each generation.
The research is supported by Beijing aerospace New trend mechanical equipment (LTD) with great force, and funding projects (No: D5204200210) and research and innovation seed funding (No: CX2020100) of northwest university of Industrial university.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made in the above embodiments by those of ordinary skill in the art without departing from the principle and spirit of the present invention.

Claims (10)

1. A method for accurately controlling and optimizing clamping force of a thin-walled part driven by a digital twin is characterized by comprising the following steps:
step 1: acquiring offline data of the thin-wall part and the clamp before clamping and online data of the thin-wall part in the clamping process; the off-line data comprises geometric parameters and material properties; the online data comprises input torque applied to the clamp, deformation displacement data of a monitoring point position arranged on the thin-wall part, and clamping environment data;
step 2: establishing a physical space model and a virtual space model of the thin-wall part-clamp based on a digital twinning technology; the virtual space model comprises a geometric information model, a finite element simulation model and a thin-wall part clamping deformation prediction model
And step 3: on the basis of the step 1 and the step 2, a precise control and optimization process of the clamping force of the thin-wall part based on the digital twin is executed:
step 3.1: transmitting the clamping process data of the real thin-walled workpiece in the physical space to a physical space model, calling real-time information of the physical space model by using a virtual space model, and displaying clamping process pictures in real time through human-computer interaction equipment;
step 3.2: performing analog simulation in a virtual space model of the thin-wall part-clamp, and predicting the deformation of the thin-wall part according to a thin-wall part clamping deformation prediction model:
step 3.3: after the thin-wall part clamping activity is finished, the full-flow data is stored in a thin-wall part-clamp physical space model, and on the basis of a thin-wall part clamping deformation prediction model, a group of optimal positioning points and clamping force parameters are obtained through an optimization algorithm to enable the clamping deformation to be minimum, so that the size of the initial clamping force of the thin-wall part is determined.
2. The method for accurately controlling and optimizing the clamping force of the thin-walled part driven by the digital twin according to claim 1 is characterized in that in the step 1, the technology of the internet of things is applied to the clamping process of the thin-walled part, a torque wrench and sensor equipment are correspondingly configured on the thin-walled part and a clamp, offline data of the thin-walled part and the clamp before clamping is obtained, monitoring points are set, and online data in the clamping process of the thin-walled part is obtained.
3. The method for precisely controlling and optimizing the clamping force of the thin-walled part driven by the digital twin according to claim 2, wherein the step 1 specifically comprises the following steps:
step 1.1: before clamping, detecting a thin-wall part to be clamped and a clamp to obtain geometric parameters and material attributes of the thin-wall part and the clamp;
step 1.2: the method comprises the steps of acquiring input torque applied to a clamp by using a torque wrench, measuring clamping force of a contact position of a thin-wall part and the clamp by using a pressure sensor, establishing acquisition points on the thin-wall part, arranging distance measurement sensing equipment to acquire deformation displacement data of monitoring points of the thin-wall part, and configuring temperature and humidity sensors in a clamping environment to acquire clamping environment data.
4. The method for precisely controlling and optimizing the clamping force of the thin-walled part driven by the digital twin according to claim 3, wherein the step 2 specifically comprises the following steps:
step 2.1: establishing a physical space model of the thin-wall part-clamp, converting a thin-wall part clamping system from a physical entity into a Web identified data model, and tracking and calling thin-wall part-clamp information;
step 2.2: and constructing a thin-wall part-clamp virtual space model which comprises a geometric information model, a finite element simulation model and a thin-wall part clamping deformation prediction model, and mapping the clamping deformation state of the thin-wall part in real time.
5. The method for precisely controlling and optimizing the clamping force of the thin-walled part driven by the digital twin according to claim 4, wherein the step 2.2 specifically comprises the following steps:
step 2.2.1: constructing a geometric model of the thin-wall part and the clamp by adopting three-dimensional software according to the data acquired in the step 1;
step 2.2.2: adopting finite element software to construct a finite element simulation model of deformation of the thin-wall part;
step 2.2.3: based on a deep neural network, a thin-wall part clamping deformation prediction model is established, a nonlinear mapping relation between the clamping force and the clamping deformation is established, deformation data after the simulation of a finite element simulation model is used as a training sample set of the thin-wall part clamping deformation prediction model, and the thin-wall part clamping deformation prediction model is trained.
6. The method for precisely controlling and optimizing the clamping force of the thin-walled part driven by the digital twin according to claim 5, wherein the step 2.2.2 specifically comprises the following steps:
step 2.2.2.1: importing a geometric model: importing the thin-wall part and the geometric model of the clamp built in the step 2.2.1 into finite element software, wherein the clamp is set as an analytic rigid body, and the thin-wall part is set as an elastic body with isotropy;
step 2.2.2.2: setting material properties: reading basic information of thin-wall parts and clamp objects of the physical space model established in the step 2.1, and setting material attributes of the thin-wall parts and the clamps;
step 2.2.2.3: assembling: assembling the thin-wall part and the clamp in finite element software according to an actual clamping mode, setting a contact position, and defining the thin-wall part and the clamp by adopting 'surface-to-surface contact';
step 2.2.2.4: boundary conditions and loads: the six degrees of freedom of the bottom surface of the thin-wall part are completely restrained, and a load is applied to the contact position of the thin-wall part and a clamp;
step 2.2.2.5: grid division: and selecting hexahedrons, sweep-Jones and centerline-axis algorithms for the grid control attributes, and dividing the grid by adopting a C3D20R unit.
7. The method for precisely controlling and optimizing the clamping force of the thin-walled part driven by the digital twin according to claim 5 or 6, wherein the step 2.2.3 specifically comprises the following steps:
step 2.2.3.1: compiling a script file according to the finite element simulation model established in the step 2.2.2 by using a parameterized analysis function of finite element software, and carrying out finite element calculation by changing the contact position coordinates and the clamping force magnitude parameters to obtain thin-wall part deformation data corresponding to the measuring points;
step 2.2.3.2: determining the input and the output of a thin-wall part clamping deformation prediction model; the thin-wall part clamping deformation prediction model input unit is a coordinate parameter of a positioning point and a clamping load applied to the thin-wall part, and is expressed as x ═ xk}(1≤k≤K)，xk＝{sk,Fk}，skRepresenting the coordinates of the anchor point in the kth sample, FkRepresents the clamp load in the kth sample; the thin-wall part clamping deformation prediction model output unit is clamping deformation data of thin-wall part measuring points and is expressed as y ═ yk}(1≤k≤K)；
Step 2.2.3.3: establishing a deep neural network with L layers, wherein the node numbers of an input layer, an i-th hidden layer and an output layer are r and L respectivelyiAnd c, selecting the tanh function as the activation function, and expressing the function as follows:
Figure FDA0002943153020000031
the error function selects a square loss function;
step 2.2.3.4: and performing gradient calculation by adopting a back propagation algorithm, correcting the weight parameter and the bias parameter of the deep neural network, and ending iteration when the iteration times or the error is less than or equal to a preset value to obtain a trained thin-wall part clamping deformation prediction model.
8. The method for precisely controlling and optimizing the clamping force of the digital twin driven thin-walled workpiece according to claim 5, is characterized in that in step 2.2.3, a sample is subjected to standardization: assuming the mean and standard deviation of the input features over the entire sample data set is u, the normalization operation is to subtract u from each value of the input features and then divide by the standard deviation.
9. The method for accurately controlling and optimizing the clamping force of the thin-walled part driven by the digital twin according to claim 5, wherein the step 3.2 comprises the following specific steps:
step 3.2.1: the action and system control instructions in the thin-wall part clamping process are transmitted to the thin-wall part-clamp virtual space model in real time, the thin-wall part-clamp virtual space model simulates the thin-wall part clamping deformation, and three-dimensional visual simulation of the thin-wall part clamping action is realized;
step 3.2.2: the method comprises the following steps of predicting clamping deformation of the thin-wall part through a thin-wall part clamping deformation prediction model:
step 3.2.2.1: according to the formula F ═ kMnInput torque M collected by torque wrenchnThe transformation into a clamping load F exerted on the surface of the thin-walled part, where k is the transformation coefficient, is determined by the following experiment: when the torque wrench inputs the torque MnAcquiring a clamping force value F ' through the pressure sensors arranged in the step 1.2 to obtain a quotient value k ', performing multiple tests to obtain multiple quotient values k ', and taking an average value as a conversion coefficient k;
step 3.2.2.2: inputting the converted clamping load and the positioning position coordinate into a thin-wall part clamping deformation prediction model to obtain a predicted value of clamping deformation; and comparing the data with historical data under the operating condition, updating the thin-wall part clamping deformation prediction model when the error between the data and the historical data exceeds an allowable value, wherein the updating step is to firstly add the real-time clamping deformation data into a training set of the thin-wall part clamping deformation prediction model, and then train the thin-wall part clamping deformation prediction model to obtain the updated prediction model.
10. The method for precisely controlling and optimizing the clamping force of the digital twin driven thin-walled workpiece according to claim 5, wherein the optimization algorithm in the step 3.3 adopts a genetic algorithm, and an objective function of the genetic algorithm is defined as:
minf(NET,s,F)
Figure FDA0002943153020000041
wherein F represents a thin-wall part clamping deformation prediction model, NET represents a neural network established in the prediction model, S represents coordinates of positioning points, S represents a point set where the positioning points are located, F represents the clamping force, and F represents the clamping forceminAnd FmaxRepresents the minimum and maximum values of the clamping force;
randomly generating an initial population comprising p individuals; decoding the individuals one by one, eliminating the individuals which do not meet the constraint conditions in the optimization model, and predicting clamping deformation of the individuals which meet the constraint conditions through a deep neural network;
the individual fitness is defined as
Figure FDA0002943153020000051
In the above formula, Ui(i is not less than 1 and not more than p) is the individual dyed in the population, and delta is a value predetermined according to the objective function value of each generation.
CN202110186148.9A 2021-02-15 2021-02-15 Digital twin-driven thin-wall part clamping force precise control and optimization method Active CN112926152B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110186148.9A CN112926152B (en) 2021-02-15 2021-02-15 Digital twin-driven thin-wall part clamping force precise control and optimization method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110186148.9A CN112926152B (en) 2021-02-15 2021-02-15 Digital twin-driven thin-wall part clamping force precise control and optimization method

Publications (2)

Publication Number Publication Date
CN112926152A true CN112926152A (en) 2021-06-08
CN112926152B CN112926152B (en) 2023-04-28

Family

ID=76169749

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110186148.9A Active CN112926152B (en) 2021-02-15 2021-02-15 Digital twin-driven thin-wall part clamping force precise control and optimization method

Country Status (1)

Country Link
CN (1) CN112926152B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113591333A (en) * 2021-09-30 2021-11-02 南方电网科学研究院有限责任公司 Construction method of GIS (gas insulated switchgear) temperature simulation model based on digital twinning
CN114547826A (en) * 2022-04-25 2022-05-27 长江空间信息技术工程有限公司(武汉) Operation method of engineering deformation monitoring network optimization design system based on digital twin
CN114803472A (en) * 2022-04-22 2022-07-29 深圳航天科技创新研究院 Robot-based clamping control method and control system
CN117763926A (en) * 2024-02-22 2024-03-26 大连理工大学 digital twin information driven high-reliability structure deformation monitoring method

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007285832A (en) * 2006-04-14 2007-11-01 Nippon Steel Corp System and method for acquiring breaking point, system and method for estimating break, programs and recording media of these method
CN104239599A (en) * 2014-07-07 2014-12-24 西安工业大学 Dynamics simulated analysis method on basis of multipoint positioning flexibility tool system
CN106874624A (en) * 2017-03-15 2017-06-20 中南大学 The method evaluated the online virtual detection of the yielding cylindrical member Forming Quality of ultra-thin-wall
CN107657129A (en) * 2017-10-17 2018-02-02 西北工业大学 Thin-wall part residual stress deformation based on clamping power monitoring perceives Forecasting Methodology
KR20200066932A (en) * 2018-12-03 2020-06-11 한국건설기술연구원 Apparatus and Method for Monitoring Damage of Structure with Measuring Strain and Digital Twin
EP3685969A1 (en) * 2019-01-28 2020-07-29 Siemens Aktiengesellschaft Computer-aided optimization of a numerically controlled machining of a workpiece
CN112327774A (en) * 2020-11-09 2021-02-05 东北大学 Digital twinning-based thin-wall part riveting quality control method

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007285832A (en) * 2006-04-14 2007-11-01 Nippon Steel Corp System and method for acquiring breaking point, system and method for estimating break, programs and recording media of these method
CN104239599A (en) * 2014-07-07 2014-12-24 西安工业大学 Dynamics simulated analysis method on basis of multipoint positioning flexibility tool system
CN106874624A (en) * 2017-03-15 2017-06-20 中南大学 The method evaluated the online virtual detection of the yielding cylindrical member Forming Quality of ultra-thin-wall
CN107657129A (en) * 2017-10-17 2018-02-02 西北工业大学 Thin-wall part residual stress deformation based on clamping power monitoring perceives Forecasting Methodology
KR20200066932A (en) * 2018-12-03 2020-06-11 한국건설기술연구원 Apparatus and Method for Monitoring Damage of Structure with Measuring Strain and Digital Twin
EP3685969A1 (en) * 2019-01-28 2020-07-29 Siemens Aktiengesellschaft Computer-aided optimization of a numerically controlled machining of a workpiece
CN112327774A (en) * 2020-11-09 2021-02-05 东北大学 Digital twinning-based thin-wall part riveting quality control method

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
ALEXANDER GALKIN 等: "Integrated Simulation of Process of Steel Casting on the Continuous Steel Casting Unit", 《2020 2ND INTERNATIONAL CONFERENCE ON CONTROL SYSTEMS, MATHEMATICAL MODELING, AUTOMATION AND ENERGY EFFICIENCY (SUMMA)》 *
焦辉等: "薄壁件铣削过程仿真与分析", 《中国制造业信息化》 *
郭诗瑶等: "基于ABAQUS的异型薄壁件装夹变形控制研究", 《制造技术与机床》 *
龚智鹏等: "2124铝合金桁梁薄壁件铣削变形仿真优化", 《机械制造与自动化》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113591333A (en) * 2021-09-30 2021-11-02 南方电网科学研究院有限责任公司 Construction method of GIS (gas insulated switchgear) temperature simulation model based on digital twinning
CN114803472A (en) * 2022-04-22 2022-07-29 深圳航天科技创新研究院 Robot-based clamping control method and control system
CN114803472B (en) * 2022-04-22 2023-12-26 深圳航天科技创新研究院 Clamping control method and system based on robot
CN114547826A (en) * 2022-04-25 2022-05-27 长江空间信息技术工程有限公司(武汉) Operation method of engineering deformation monitoring network optimization design system based on digital twin
CN114547826B (en) * 2022-04-25 2022-07-12 长江空间信息技术工程有限公司(武汉) Operation method of engineering deformation monitoring network optimization design system based on digital twin
CN117763926A (en) * 2024-02-22 2024-03-26 大连理工大学 digital twin information driven high-reliability structure deformation monitoring method

Also Published As

Publication number Publication date
CN112926152B (en) 2023-04-28

Similar Documents

Publication Publication Date Title
CN112926152B (en) Digital twin-driven thin-wall part clamping force precise control and optimization method
CN110609531B (en) Workshop scheduling method based on digital twin
CN113051831B (en) Modeling method and thermal error control method for thermal error self-learning prediction model of machine tool
CN106125574B (en) Piezoelectric ceramics mini positioning platform modeling method based on DPI model
CN106842922B (en) Numerical control machining error optimization method
CN111723440B (en) Thin-wall part machining precision prediction hybrid modeling method
CN110007645B (en) Feeding system hybrid modeling method based on dynamics and deep neural network
CN109918614B (en) Global dynamic strain measurement method based on modal learning
CN109948207A (en) A kind of aircraft engine high pressure rotor rigging error prediction technique
CN107063060A (en) A kind of method and device for determining surface planarity
CN111368466B (en) Mechanical vibration prediction method based on frequency response function parameter correction
CN110039373A (en) A kind of main shaft of numerical control machine tool thermal deformation prediction technique and system
CN110631792A (en) Seismic hybrid test model updating method based on convolutional neural network
Hu et al. Design, implementation, and testing of advanced virtual coordinate-measuring machines
Ostrowska et al. Comparison of accuracy of virtual articulated arm coordinate measuring machine based on different metrological models
KR20120050597A (en) Hot forming apparatus using thermal deformation predicting for curved plates in ship and method therof
CN113504768A (en) High-precision product digital twin computability method for assembly quality prediction
CN116227045B (en) Local stress strain field construction method and system for structural test piece
CN114417681B (en) Two-dimensional structure deformation monitoring method and device based on dynamic decision and neural network
CN114139772B (en) Method for predicting workpiece surface quality based on optimized generalized regression neural network
CN113570147B (en) Method and equipment for rapidly predicting machining errors of thin-wall parts in multiple modes
CN113704922B (en) Method for predicting surface roughness based on sound vibration and texture characteristics
Karadogan et al. Potential use of machine learning to determine yield locus parameters
CN115618709A (en) Method for establishing digital twinning model of textile surface quality under periodic frictional wear
CN111210877B (en) Method and device for deducing physical parameters

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