CN117421940B

CN117421940B - Global mapping method and device between digital twin lightweight model and physical entity

Info

Publication number: CN117421940B
Application number: CN202311743307.6A
Authority: CN
Inventors: 李伟; 朱秀文; 王晓俊; 刘波; 郭年程; 周帅; 闫安; 孙浩南; 李芸先
Original assignee: Shandong Jiaotong University
Current assignee: Shandong Jiaotong University
Priority date: 2023-12-19
Filing date: 2023-12-19
Publication date: 2024-03-19
Anticipated expiration: 2043-12-19
Also published as: CN117421940A

Abstract

The application discloses a global mapping method and device between a digital twin lightweight model and a physical entity, which belong to the technical field of digital twin, and the method comprises the following steps: collecting data information of the lightweight product, and establishing a digital twin lightweight model; carrying out finite element analysis on a digital twin light model of a light product to obtain a stress strain cloud image in a virtual environment; converting the stress-strain cloud image into an equal proportion stress density image based on the stress-strain constitutive relation and the color gradient change rule; dividing the stress density image to obtain a global mapping region of the digital twin lightweight model and the physical entity; training parameters based on a single variable method to obtain a global mapping region of a final digital twin lightweight model and a physical entity; global mapping between the digital twin lightweight model of the lightweight product and the physical entity is performed. The method and the device improve the efficiency and accuracy of data transmission between the digital twin lightweight model and the physical entity.

Description

Global mapping method and device between digital twin lightweight model and physical entity

Technical Field

The application relates to a global mapping method and device between a digital twin lightweight model and a physical entity, and belongs to the technical field of digital twin.

Background

Traffic departments are the subject of significant attention to energy work. Lightweight is an effective way to achieve the two carbon targets of automobiles. When lightweight products are oriented to application scenarios, complex conditions, large dimensions and heavy loads are often faced, and these factors have a significant impact on the lightweight product design cycle. In order to shorten the lightweight product design cycle, digital twin technology is introduced into the lightweight field.

Digital twinning techniques effectively utilize virtual analytics to develop lightweight products, which must incorporate feedback information of physical entities. This means that a good feedback mechanism needs to be established between the virtual analysis and the physical entity to ensure the accuracy of the virtual analysis results. When physical entity information is required to be fed back to the twin model, operational data of the physical entities must be collected in the real scene and mapped into the virtual space to further optimize the design of the lightweight product. However, the precise physical entity-virtual space global mapping is affected by a plurality of factors, and it is difficult to accurately find a limited global mapping area.

At present, most of methods for searching mapping areas between a digital twin model and a physical entity are based on simulation analysis results of a virtual model, and areas with larger stress or strain are subjectively selected and used as mapping areas between the virtual model and the physical entity, so that a data acquisition area of the physical entity is mapped. However, the above method for searching the mapping region between the digital twin model and the physical entity has larger subjectivity, is easy to have wrong selection and miss selection, and is difficult to accurately cover the whole of the virtual twin model. Therefore, when the digital twin technology is applied to the design process of the automobile lightweight product, particularly when the automobile lightweight product faces to a real application scene, the actual conditions of large size, complex multiple working conditions and the like are often encountered, so that the problem that mapping between a twin model in a digital twin virtual environment and a physical entity in a real environment is incomplete and inaccurate is caused.

Disclosure of Invention

In order to solve the problems, the application provides a global mapping method and device between a digital twin lightweight model and a physical entity, which can realize global mapping between the digital twin lightweight model and the physical entity of a lightweight product and improve the efficiency and accuracy of data transmission between the digital twin lightweight model and the physical entity.

The technical scheme adopted for solving the technical problems is as follows:

in a first aspect, a global mapping method between a digital twin lightweight model and a physical entity provided in an embodiment of the present application includes the following steps:

collecting data information of the lightweight product, and establishing a digital twin lightweight model; the data information comprises design parameters of the lightweight product, physical entity real data and physical principle information;

carrying out finite element analysis on a digital twin light weight model of the light weight product to obtain a stress strain cloud picture of the light weight product in a virtual environment;

converting the stress-strain cloud image into an equal proportion stress density image based on the stress-strain constitutive relation and the color gradient change rule;

dividing the stress density image based on a mean shift method to obtain a global mapping region of the digital twin lightweight model and the physical entity;

training parameters based on a single variable method to obtain a global mapping region of a final digital twin lightweight model and a physical entity;

and performing global mapping between the digital twin light weight model of the light weight product and the physical entity by utilizing the global mapping area of the final digital twin light weight model and the physical entity.

As a possible implementation manner of this embodiment, the finite element analysis is performed on a digital twin lightweight model of a lightweight product to obtain a stress and strain cloud image under a virtual environment of the lightweight product, including:

establishing a virtual analysis module of the digital twin light weight model, and performing finite element analysis on the digital twin light weight model of the light weight product after setting a solver and analysis options to obtain a stress and strain cloud image under the virtual environment of the light weight product; the virtual analysis module comprises a geometric model module, a boundary condition module, a material attribute module, a grid division module and an application loading module.

As a possible implementation manner of this embodiment, the dividing the stress density image based on the mean shift method to obtain a global mapping area of the digital twin lightweight model and the physical entity includes:

dividing the stress density image to form a limited clustering mass center;

and mapping a data acquisition area of the physical entity by taking the limited clustering centroids as feature mapping areas of the digital twin lightweight model and the physical entity to obtain a global mapping area of the digital twin lightweight model and the physical entity.

As a possible implementation manner of this embodiment, the training parameters based on the single variable method to obtain a global mapping area of a final digital twin lightweight model and a physical entity includes:

performing multi-combination training on the drift radius and the variance of the density estimation function in the image segmentation process by adopting a univariate contrast method, and determining the optimal drift radius and the variance of the density estimation function;

and obtaining a global mapping region of the final digital twin lightweight model and the physical entity based on the optimal drift radius and the variance of the density estimation function.

As a possible implementation manner of this embodiment, the converting the stress-strain cloud chart into the equal-proportion stress density image based on the stress and strain constitutive relation and the color gradient change rule includes:

and A, determining the stress and strain constitutive relation:

the relationship between stress and strain is described by the modulus of elasticity E:、/>、/>wherein->For stress->Is strain; />Is the magnitude of the acting force; />Is the area of force application; />Indicating the amount of change in length; />Is the initial length.

B, determining the equal proportional relation between the discrete unit and the pixel unit:

the proportional relationship between discrete cell size and pixel size is: Wherein->Stress for digital twin light modelAnd (3) the discrete unit size of the strain module; />The whole model size of the digital twin light model stress and strain module is provided;is the pixel cell size; />To obtain the final density image through pixel adjustment for the whole pixel size of the color cloud image which needs data conversion.

C, converting the color cloud image into a density image through pixel units and RGB red channel values of each pixel unit as media:

storing color stress strain cloud image pixel units in a 3D array of m multiplied by n multiplied by 3, and taking a red channel value of each pixel unit as a reference value of stress density; defining a locationg=（i,j) The density storage unit of the stress density image converted by the pixel unit isIts density value ismWherein, is less than or equal to 1i≤m；1≤j≤n；0≤m255 or less; red channel values are extracted and returned to a matrix of size mxn: />Wherein->In the matrix, a density storage unit is provided, the value of which is equal to the RGB red channel value of each pixel unit, and the position of which is g= (i, j); matrix density storage unit->The density value in the memory is the density memory cell of stress density image +.>Density value of (2)mThe method comprises the steps of carrying out a first treatment on the surface of the Base groupDensity storage cell in stress Density image with Gaussian distribution function >Internal random generationmAnd (3) carrying out point separation to obtain a point set D, wherein the point set D is a density point set capable of carrying out density image segmentation.

a, performing nuclear density estimation:

different working conditions are in three-dimensional spaceN data units are given->，/>，/>Belongs to point clouds converted from stress cloud pictures; by kernel function->And a symmetrically positive 3 x 3 bandwidth matrix D as parameters, calculated at the point +.>Multi-kernel density estimation at: />，/>Wherein->Is a kernel function with a bandwidth matrix D, where D represents the determinant of the bandwidth matrix D.

Three-dimensional nuclear letterNumber of digitsIs a bounded function that satisfies the following conditions: />，，/>，/>The method comprises the steps of carrying out a first treatment on the surface of the Wherein->Is a constant, I is an integer set, | +.>II represents vector +.>Is (are) norms of->Representation vector->Is a transpose of (a).

The multidimensional kernel function is derived from a symmetric univariate kernel function by two different methodsGenerating:，/>wherein->Obtained from the product of the univariate kernels; />By->Middle rotation->Obtained, i.e.)>Is radially symmetrical,>is constant.

Considering a radially symmetric kernel function, then the symmetric multidimensional kernel function satisfies: Wherein +.>Need to make->Normalized constant->When strictly positive, kernel function->The integral is 1.

Making the bandwidth matrix D proportional to the same matrixThe method comprises the steps of carrying out a first treatment on the surface of the Providing a bandwidth parameter->At this time, the kernel density estimator is: />。

And B, estimating probability density gradient:

calculating secretGradient of degree:。

assume thatThe derivative of (2) is +.>If the kernel function exists in the interval, the kernel function is defined>The method comprises the following steps:wherein->，/>Is a kernel contour function +.>Is a derivative of (2); />Is a corresponding normalization function, and is introduced into a density gradient formula to obtain the following components:。

calculated outThe density estimate at that point is proportional to: />。

The mean shift vector is:。

use of coresAs a weight +.>As the core, there are: />，。

Will be as followsWindow function with a bandwidth h for the center +.>Along the mean shift vector->Translation, obtaining a new window function->。

C, all density storages in the stress density image are seed points of the mean shift program; for each seed pointSetting the drift radius h, calculating to +.>Kernel density weighted average vector +.f for all other sample points within the centered, radius h window>For each seed point +.>Performing translation, i.e.)>Until the seed point converges, i.e. +. >The method comprises the steps of carrying out a first treatment on the surface of the And (5) carrying out mean shift iteration to realize density image segmentation.

And D, finding out a twin model virtual analysis stress dangerous point based on stress density image segmentation, mapping out a data sampling area when the physical entity runs, and constructing a global mapping area of the digital twin lightweight model and the physical entity.

As a possible implementation manner of this embodiment, the global mapping method between the digital twin lightweight model and the physical entity further includes:

verifying the digital twin lightweight model by using physical entity real data: and acquiring and analyzing the light product operation data based on the sampling points in a real operation scene, and comparing the light product operation data with the analysis data in the digital twin mapping area in the virtual environment to verify the accuracy of the digital twin light model.

In a second aspect, an apparatus for global mapping between a digital twin lightweight model and a physical entity provided in an embodiment of the present application includes:

the data acquisition module is used for acquiring data information of the lightweight product and establishing a digital twin lightweight model; the data information comprises design parameters of the lightweight product, physical entity real data and physical principle information;

The finite element analysis module is used for carrying out finite element analysis on the digital twin light weight model of the light weight product to obtain a stress strain cloud image of the light weight product in a virtual environment;

the cloud image conversion module is used for converting the stress-strain cloud image into an equal-proportion stress density image based on the stress-strain constitutive relation and the color gradient change rule;

the image segmentation module is used for segmenting the stress density image based on a mean shift method to obtain a global mapping region of the digital twin lightweight model and the physical entity;

the parameter training module is used for training parameters based on a single variable method to obtain a global mapping area of the final digital twin lightweight model and the physical entity;

and the global mapping module is used for carrying out global mapping between the digital twin light weight model of the light weight product and the physical entity by utilizing the final global mapping area of the digital twin light weight model and the physical entity.

In a third aspect, an embodiment of the present application provides a computer device, including a processor, a memory, and a bus, where the memory stores machine-readable instructions executable by the processor, and when the computer device is running, the processor communicates with the memory through the bus, and the processor executes the machine-readable instructions to perform steps of a global mapping method between any digital twin lightweight model and a physical entity as described above.

In a fourth aspect, embodiments of the present application provide a storage medium having a computer program stored thereon, which when executed by a processor performs the steps of the global mapping method between any of the digital twinning lightweight models and physical entities described above.

The technical scheme of the embodiment of the application can have the following beneficial effects:

in order to realize more comprehensive, more accurate and faster data transmission between the digital twin lightweight model and the physical entity, the method searches for a more representative mapping area between the digital twin lightweight model and the physical entity through automatic global twin model identification; and by using a stress density image segmentation technology, the feature mapping area of the digital twin lightweight model is automatically and accurately identified, and then the feature sampling area of the physical entity is mapped, so that the global mapping between the digital twin lightweight model of the lightweight product and the physical entity is realized.

The cloud image-density image conversion is realized based on the stress and strain constitutive relation, the color gradient change rule and the proportional relation of the discrete units and the pixel units, and a better foundation is provided for subsequent image segmentation by converting the stress cloud image into a stress density image capable of executing image segmentation; the method realizes stress density image segmentation based on a mean shift method, and can quickly and accurately automatically identify the virtual global stress dangerous area of the twin model, thereby mapping the whole physical entity sampling area. In order to find out limited key virtual-real mapping areas under the complex multi-working conditions, large size and heavy load real scenes of lightweight products, the method and the device screen more comprehensive mapping areas, so that the found limited twin models and physical entity mapping areas are high in matching rate and strong in representativeness.

Drawings

FIG. 1 is a flow chart illustrating a method of global mapping between a digital twinning lightweight model and a physical entity, according to an example embodiment;

FIG. 2 is a schematic diagram illustrating a global mapping arrangement between a digital twin lightweight model and a physical entity, according to an example embodiment;

FIG. 3 is a digital twin schematic of a cargo box shown according to an exemplary embodiment;

FIG. 4 is a simulated stress cloud for a condition according to an exemplary embodiment;

FIG. 5 is a local stress cloud of a cargo box twinning model under static load conditions, according to an example embodiment;

FIG. 6 is a plan view of a density image after the conversion of the density image is complete, according to an exemplary embodiment;

FIG. 7 is a three-dimensional representation of a density image shown according to an exemplary embodiment;

fig. 8 is a graph of a density image segmentation result, according to an example embodiment.

Detailed Description

The present application is further described with reference to the accompanying drawings and examples:

in order to clearly illustrate the technical features of the present solution, the present application will be described in detail below with reference to the following detailed description and the accompanying drawings. The following disclosure provides many different embodiments, or examples, for implementing different structures of the application. In order to simplify the disclosure of the present application, the components and arrangements of specific examples are described below. Furthermore, the present application may repeat reference numerals and/or letters in the various examples. This repetition is for the purpose of simplicity and clarity and does not in itself dictate a relationship between the various embodiments and/or configurations discussed. It should be noted that the components illustrated in the figures are not necessarily drawn to scale. Descriptions of well-known components and processing techniques and processes are omitted so as to not unnecessarily obscure the present application.

As shown in fig. 1, an embodiment of the present application provides a global mapping method between a digital twin lightweight model and a physical entity, including the following steps:

dividing the stress density image to form a limited clustering mass center;

and A, determining the stress and strain constitutive relation:

the relationship between stress and strain is described by the modulus of elasticity E:

wherein->For stress->；/>For strain->；/>Is the magnitude of the acting force; />Is the area of force application; />Indicating the amount of change in length; />Is the initial length.

the proportional relationship between discrete cell size and pixel size is:

wherein->The stress and strain module discrete unit size of the digital twin light model is adopted; />The whole model size of the digital twin light model stress and strain module is provided; / >Is the pixel cell size; />The whole pixel size of the color cloud image which needs to be subjected to data conversion; the final density image is obtained by pixel adjustment.

storing color stress strain cloud image pixel units in a 3D array of m multiplied by n multiplied by 3, and taking a red channel value of each pixel unit as a reference value of stress density; defining the density storage unit of the stress density image converted by the pixel unit of the position g= (i, j) asIts density value is->Wherein i is more than or equal to 1 and less than or equal to m; j is more than or equal to 1 and less than or equal to n; 0.ltoreq.L>255 or less; red channel values are extracted and returned to a matrix of size mxn: />Wherein->In the matrix, a density storage unit is provided, the value of which is equal to the RGB red channel value of each pixel unit, and the position of which is g= (i, j); matrix density storage unit->The density value in the memory is the density memory cell of stress density image +.>Density value of->The method comprises the steps of carrying out a first treatment on the surface of the Density storage cell in stress Density image based on Gaussian distribution function>Internal random generation->And (3) carrying out point separation to obtain a point set D, wherein the point set D is a density point set capable of carrying out density image segmentation.

a, performing nuclear density estimation:

Three-dimensional kernel functionIs a bounded function that satisfies the following conditions: />，，/>，/>The method comprises the steps of carrying out a first treatment on the surface of the Wherein->Is a constant, I is an integer set, | +.>II represents vector +.>Is (are) norms of->Representation vector->Is a transpose of (a).

Considering a radially symmetric kernel function, then the symmetric multidimensional kernel function satisfies:wherein +.>Need to make- >Normalized constant->When strictly positive, kernel function->The integral is 1.

And B, estimating probability density gradient:

calculating the density gradient:。

calculated outThe density estimate at that point is proportional to: />。

The mean shift vector is:。

use of coresAs a weight +.>As the core, there are: />，。

C, all density storages in the stress density image are seed points of the mean shift program; for each seed pointSetting the drift radius h, calculating to +.>Kernel density weighted average vector +.f for all other sample points within the centered, radius h window>For each seed point +.>Performing translation, i.e.)>Until the seed point converges, i.e. +.>The method comprises the steps of carrying out a first treatment on the surface of the And (5) carrying out mean shift iteration to realize density image segmentation.

verifying the digital twin lightweight model by using physical entity real data: and acquiring and analyzing the light product operation data based on the sampling area in the real operation scene, and comparing the light product operation data with the analysis data at the digital twin mapping area in the virtual environment to verify the accuracy of the digital twin light model.

As shown in fig. 2, a global mapping device between a digital twin lightweight model and a physical entity provided in an embodiment of the present application includes:

the data acquisition module is used for acquiring data information of the lightweight product and establishing a digital twin lightweight model; the digital twin model comprises design parameters of a lightweight product, physical entity real data and physical principle information;

The specific process of global mapping between the digital twin lightweight model of the lightweight product and the physical entity by using the device of the invention is as follows.

Step one, collecting data information of a lightweight product, and establishing a digital twin lightweight model; the data information comprises design parameters of the lightweight product, physical entity real data and physical principle information.

The digital twin light model is actually a virtual model, and an initial model is a virtual model established in computer software and comprises information such as geometric shapes, material properties, physical characteristics, engineering parameters and the like of various parts of the vehicle. Fig. 3 is a digital twinning schematic of a cargo box. Through the digital twin light model, simulation analysis, prediction and evaluation can be carried out on the vehicle in the design, manufacture and use processes, and product optimization is realized on the basis of ensuring the safe and reliable performance. The most important mechanical properties of dump truck containers with large geometry and load characteristics are represented by stress and strain conditions. Therefore, the application firstly carries out analysis pretreatment on the initial digital twin lightweight model, including finite element meshing, material parameter confirmation, boundary condition confirmation and multi-task loading. And then performing virtual calculation analysis to obtain virtual stress and strain data of the lightweight product. And finally, converting the obtained ground stress and strain cloud image into a stress density image for image segmentation.

And step two, carrying out finite element analysis on a digital twin light model of the light product to obtain a stress strain cloud image of the light product in a virtual environment.

The finite element analysis is carried out on the digital twin light model of the light product to obtain a stress and strain cloud image under the virtual environment of the light product, and the method comprises the following steps: establishing a virtual analysis module of the digital twin light weight model, and performing finite element analysis on the digital twin light weight model of the light weight product after setting a solver and analysis options to obtain a stress and strain cloud image under the virtual environment of the light weight product; the virtual analysis module comprises a geometric model module, a boundary condition module, a material attribute module, a grid division module and an application loading module.

And discretizing the container twin model, adding boundary conditions, and performing finite element simulation analysis to obtain the distribution condition of the overall stress and strain of the container. The analysis result includes values and direction components of various types of stress and strain, as shown in fig. 4, where fig. 4 is a color cloud chart, so that the stress and strain conditions of the whole container can be intuitively reflected, and an engineer usually decides the position of the data mapping area according to the color distribution of the stress cloud chart. However, conventional methods tend to be time consuming and labor intensive and subject to errors in determining the location of the mapped region. Another possible approach is to output the stress and strain values for all cells using software and then determine the mapped region by simple screening. However, the mapped regions obtained by such simple screening often do not represent the features of the overall system well. It is therefore necessary to find a more efficient way to determine the mapping region location.

And thirdly, converting the stress-strain cloud image into an equal-proportion stress density image based on the stress-strain constitutive relation and the color gradient change rule.

Stress and strain clouds cannot be directly segmented, and need to be converted into stress density images and image segmentation is accomplished by performing a mean shift procedure. The process realizes the conversion from the cloud image to the density image based on the stress and strain constitutive relation, the color gradient change rule and the proportion relation between the discrete units and the pixel units.

3.1 stress and strain constitutive relations.

For elastic materials, the relationship between stress and strain can be described by hooke's law. Hooke's law states that within elastic limits, stress is proportional to strain and the proportionality constant is the modulus of elasticity. The physical entity of the elastic material is elastically deformed under the action of external stresses and this deformation is directly proportional to the stresses to which it is subjected. Stress refers to the magnitude of an applied force per unit area, which is an amount that describes the ability of a material to resist the action of an external forceBy symbolsThe expression is: />Wherein->Is the magnitude of the acting force; />Is the area of force application.

Strain refers to the degree of change of an object relative to its original length, volume or shape under the action of force, symbolized by And (3) representing. The formula is as follows: />Wherein->Indicating the amount of change in length; />Is the initial length.

The relationship between stress and strain can be described by the modulus of elasticity E. The elastic modulus is a physical quantity reflecting the deformation resistance of an object in the elastic range. The formula is as follows:typically, the modulus of elasticity is a constant, i.e. each material has a fixed value under certain conditions.

Strictly speaking, stress and strain are two different quantities and cannot be directly expressed in terms of one quantity. However, in elastic materials, there is a linear relationship between stress and strain, which can be related by the modulus of elasticity. This linear relationship allows the elastic modulus to be used to represent the proportional relationship between stress and strain. Thus, based on this constitutive relationship of stress and strain, the stress and strain values can be mutually converted under certain conditions.

3.2 color gradient law.

The clustering object of the density image segmentation method is density points, and the color cloud image can not be directly segmented, and the color cloud image needs to pass through pixel units and RGB red channel values of each pixel unitAs a medium into a density image. The RGB law of color images is based on the additive color mixing principle of the three primary colors (red, green, blue). R, G, B are respectively the brightness values of the three channels of red, green and blue, and images with different colors can be obtained by adjusting the brightness values of the three channels. In the RGB model, the color of each pixel consists of intensity values of three components, red, green, and blue. The intensity value of each component ranges from 0 to 255, and the intensity value of the component represents the brightness intensity of the color. Where 0 represents the lowest luminance and 255 represents the highest luminance. The color cloud image of the prediction result output by the computer software used in the implementation process accords with the RGB rule, so that the stress density of the color cloud image can be extracted according to the RGB rule of the color cloud image.

Fig. 5 is a stress cloud diagram of the container twin model under the static load condition, and the specific position is the front panel area of the container.

The color cloud image pixel units are stored in a 3D array of m x n x 3, and the red channel value of each pixel unit is used as a reference value of stress density. Defining the density storage unit of the stress density image converted by the pixel unit of the position g= (i, j) asIts density value is->. Wherein i is more than or equal to 1 and less than or equal to m; j is more than or equal to 1 and less than or equal to n; 0.ltoreq.L>And is less than or equal to 255. Then extract the red channelValue, acquisition ofThe magnitude of the value is such that it returns a matrix of size mxn, which ultimately contains the red channel information for all pixel cells in the image.

First, an m×n matrix is constructed:

wherein->In the matrix is a density storage unit, that is, RGB red channel value of each pixel unit, where g= (i, j); matrix density storage unit->The density value in the memory is the density memory cell of stress density image +.>Density value of->。/>

Then, based on Gaussian distribution function, density storage unit of stress density imageInternal random generation->A point. Finally, a point set D is obtained. The point set D is the density point set capable of carrying out density image segmentation.

3.3 proportional relationship of discrete units to pixel units.

In general, the pixels of a color cloud image that is directly output are large, and if it is directly converted into a density image, the workload is great. However, larger pixels do not have a more favorable effect on the segmentation result of the density image. Therefore, the pixels of the color cloud need to be adjusted to the appropriate size before density conversion can be performed. Considering that the stress and strain module of the digital twin lightweight model in the implementation of the present application is a discrete model, the discrete units have a certain size. Thus, the present application proposes a proportional relationship between discrete cell size and pixel size, which is an innovative view. Thus, a more reasonable density image can be obtained through proper pixel adjustment, and a better foundation is provided for subsequent image segmentation.

The proportional relationship between discrete cell size and pixel size is:

wherein->The stress and strain module discrete unit size of the digital twin light model is adopted; />The whole model size of the digital twin light model stress and strain module is provided; />Is the pixel cell size; />The whole pixel size is for a color cloud requiring data conversion.

The pixels of fig. 5 are adjusted to 108 x 101 according to the proportional relationship between the discrete cell size and the pixel size, and a 108 x 101 matrix is further constructed according to the m x n matrix. By gaussian distribution, a set of density points a, i.e. a stress density image, is obtained. Fig. 6 is a plan view of a stress density image of the completed transformation, and fig. 7 is a three-dimensional representation of the density image. In fig. 6 and 7, the abscissa and ordinate represent the m-vector and n-vector in the matrix, i.e., the i-coordinate and j-coordinate of the stress density storage unit in the matrix, respectively, and the vertical coordinate represents the density magnitude within each matrix unit, i.e., the stress density magnitude within the stress density storage unit. By comparing fig. 5 and 6, it can be intuitively found that the gradient change in the color cloud is the same as the density gradient change in the density image it translates into. The conversion method can accurately capture the stress and strain characteristics of the cloud image, thereby providing a basis for image segmentation processing.

And step four, segmenting the stress density image based on a mean shift method to obtain a global mapping region of the digital twin lightweight model and the physical entity.

The method for dividing the stress density image based on the mean shift to obtain a global mapping region of the digital twin lightweight model and the physical entity comprises the following steps:

A, performing nuclear density estimation:

different working conditions are in three-dimensional spaceIs given by->Data unit->，/>，/>Belongs to point clouds converted from stress cloud charts. By kernel function->And symmetrically positive->Bandwidth matrix->For parameters, calculate at the point +.>Multi-kernel density estimation at: />，/>。

Is a density estimation function,/-, a>Representing the point at which the density estimate is currently being calculated, whereas +.>Each individual data point in the dataset is represented. />Is a specific location or point for calculating a density estimate at that location. But->Is an index of data points in the dataset for loop iteration and for computing the sum of the accumulation functions. Nuclear density estimation by calculating each data point +.>For the current position +.>And weight sum them to generate the contribution at that locationA density estimate at the location.

Is a kernel function with a bandwidth matrix D. Is a symmetric non-negative function that takes values over a domain and is used to calculate a density estimate; |d| represents the determinant of the bandwidth matrix D. The determinant represents the scaling factor of the matrix, i.e. the ready-to-useIn adjusting the bandwidth scale. In the kernel density estimation, the smoothness and sensitivity of the estimation can be adjusted by adjusting the scale of the bandwidth; the latter half of this formula is to perform the inverse operation on the bandwidth matrix D, i.e. +. >The method comprises the steps of carrying out a first treatment on the surface of the Then multiply it by the dot +.>And passes the result as input to the kernel function. Will result in kernel function with bandwidth matrix D +.>Values. This is done to take into account the influence of the bandwidth matrix in the parameters of the kernel function, thereby better adjusting the window size and shape of the density estimate.

Three-dimensional kernel functionIs a bounded function that satisfies the following conditions: />，，/>，/>Wherein->Is a constant, I is a set of integers; II->II represents vector +.>Is also referred to as a norm (alternatively called a modulus or length). The norm is a function of mapping the vector to a non-negative value for providingFor vector size or distance metrics; />Representation vector->Is a transpose of (a).

The multidimensional kernel can be derived from a symmetric univariate kernel by two different methodsGenerating:，/>wherein->Obtained from the product of the univariate kernels; />By->Middle rotation->Obtained, i.e.)>Is radially symmetrical,>is constant.

Considering a radially symmetric kernel function, then the symmetric multidimensional kernel function satisfies:wherein +.>Need to make->. Normalization constant->When strictly positive, kernel function->The integral is 1.

Making the bandwidth matrix D proportional to the same matrixThe method comprises the steps of carrying out a first treatment on the surface of the Providing a bandwidth parameter- >At this time, the kernel density estimator is: />。

And B, estimating probability density gradient:

calculating the density gradient:。

assume thatThe derivative of (2) is +.>If the kernel function exists in the interval, the kernel function is defined>The method comprises the following steps: />Wherein->；/>Is a corresponding normalization function, and can be introduced into a density gradient formulaObtaining:。

calculated outThe density estimate at that point is proportional to: />。

The mean shift vector is:。

use of coresAs a weight +.>As the core, there are: />，。

All density stores in the stress density image are seed points for the mean shift procedure. For each seed pointIts drift radius h is set. Then calculate to +.>Kernel density weighted average vector +.f for all other sample points within the centered, radius h window>. Finally for each seed point +.>Performing translation, i.e.)>Until the seed point converges, i.eThe method comprises the steps of carrying out a first treatment on the surface of the And (5) carrying out mean shift iteration to realize density image segmentation.

The stress density image is segmented, the segmented density image can form a limited clustering centroid, the clustering centroid is a dangerous area of a virtual analysis result of the twin model, and a data acquisition area of a physical entity is mapped, so that a global mapping area of the digital twin lightweight model and the physical entity is found.

And inputting the density image converted by the color cloud image into an iterative program to obtain a stress density image segmentation result shown in fig. 8. Each iteration in the process calculates a weighted average of the densities around each seed point and moves the surrounding density points toward the average. The process gradually moves the density points to the area of greatest density and forms a cluster centroid. Fig. 8 shows the location of cluster centroids in a density image formed after the mean shift is completed. In fig. 8, ", is the stress density drift centroid, which is referred to as the stress hazard point, i.e., the twinning model map region. By observing fig. 8, it can be found that the distribution rule of the mapping points is consistent with the density distribution rule of the density image, however, the mapping area also faces some problems, such as overlapping or being distributed in the area with smaller stress and smaller gradient. Therefore, the mapping region needs to be further screened, so that the screened mapping region can be ensured to construct an accurate mapping region.

And fifthly, training parameters based on a single variable method to obtain a global mapping area of the final digital twin lightweight model and the physical entity.

The quality of the distribution of the twin model map region depends on the quality of the image segmentation, while the mean shift based image segmentation is affected by two main factors:

1) Distribution of seed points: the distribution of seed points is an important factor affecting the segmentation result. In the implementation of the present application, the density image is randomly generated based on a gaussian distribution function, so that the variance s of the density estimation function is a factor affecting the image segmentation quality. The larger variance s can lead to wider distribution of seed points, so that the density estimation in the clustering process is smoother, and the segmentation result can be excessively fuzzy; a smaller variance s may concentrate the distribution of seed points, resulting in finer density estimation during clustering, which may result in too fine segmentation results. Therefore, selecting the appropriate variance s may affect the accuracy and effectiveness of the segmentation result.

2) Size of drift radius: the drift radius is a parameter that determines the color space and spatial distance range that the density point takes into account when achieving the drift. The larger drift radius can enable wider color difference and space distance to be considered in the drift process, so that the separation result is too smooth, and the obtained characteristic area is less; a smaller drift radius may ignore some detailed information, and the obtained feature area is more, so that the segmentation result is not accurate enough. Therefore, a proper drift radius size needs to be selected to maintain accuracy and representativeness of the segmentation result.

In summary, the distribution of seed points and the size of the drift radius are two main factors affecting the image segmentation quality based on the mean shift procedure. By selecting the appropriate variance s and drift radius, a more accurate and efficient segmentation result can be obtained, resulting in a representative mapping region. The specific implementation process of the method sets a plurality of Gaussian random distribution variance and drift radius operation combinations for training for a plurality of times. Specific training parameter settings are listed in table 1.

Table 1: image segmentation training parameters

In the implementation process of the method, aiming at two parameters of Gaussian variance s and drift radius h, a univariate comparison method is adopted for multi-combination training. Specifically, the parameter variation range of the gaussian variance s is 0.1 to 1, and the scale of each variation is 0.1; the parameter of the drift radius h varies in the range 5 to 50, with the scale of each variation being 5.

Through a univariate comparison method, the values of the Gaussian variance s and the drift radius h can be respectively changed, and then the segmentation effect under each group of parameter combination is evaluated. Through multi-combination training, the influence of different parameter combinations on the segmentation result can be analyzed, and the optimal parameter combination is selected to improve the segmentation quality. The value of the drift radius h is fixed first and then the magnitude of the gaussian variance s value is changed. For example, the first set of fixed h has a value of 5,s changed, ranging from 0.1 to 1; the second set of fixed h has a value of 10, s has a value of 0.1 to 1, and so on.

The value of the gaussian variance s is then fixed and then the magnitude of the drift radius h value is changed. For example, the first set of fixtures s has a value of 0.1, h has a value of 5 to 50, and h has a value of change; the second set of fixtures s has a value of 0.2, s has a value of 5 to 50, and so on.

According to the training results, the Gaussian variance s has little effect on the image segmentation result, and the drift radius has larger effect on the image segmentation. This means that the change in the drift radius has a more pronounced effect on the result of the image segmentation when adjusting the parameters. Therefore, in optimizing the image segmentation, adjustment of the drift radius needs to be focused. By changing the size of the drift radius, the image segmentation can be optimized to obtain more accurate and effective segmentation results. By optimizing the image segmentation parameters, the distribution of the mapping region can be better controlled, and the conditions of overlapping and unreasonable distribution are avoided. By continuously improving algorithm parameters, the accuracy and the representativeness of the selection of the mapping region can be improved, so that the interaction effect of the digital twin lightweight model and the physical entity is improved.

And step six, performing global mapping between the digital twin light weight model of the light weight product and the physical entity by utilizing a global mapping area of the final digital twin light weight model and the physical entity.

Based on the above process, a cargo box twin model-physical entity mapping region can be constructed.

And seventh, verifying the digital twin lightweight model by using the physical entity real data.

And acquiring and analyzing the light product operation data based on the sampling area in the real operation scene, and comparing the light product operation data with the analysis data at the digital twin mapping area in the virtual environment to verify the accuracy of the digital twin light model.

There is a certain error between the physical entity and the twin model, and the error rate directly affects the accuracy of the twin model. In order to evaluate the accuracy of the container twin model, an accuracy evaluation standard is provided, and factors such as historical experience, existing research, engineering actual requirements and the like are comprehensively considered, and a safety coefficient is introduced. The accuracy evaluation standard requires that the error between the virtual data of the twin model and the real data of the physical entity is not more than 15%, and the ratio of the number of the mapping areas meeting the error requirement to the total number of the mapping areas is not less than 80%. And comparing and verifying the virtual data of the container twin model with the real data of the physical entity based on the standard. Table 2 lists the verification results for 84 mapped regions.

Table 2: error range and ratio

From table 2, it can be seen that the ratio of the measurement points with the error range within 15% is 88.09%. This meets the accuracy standard, indicating the accuracy of the method for finding the real-scene-oriented physical entity data sampling area based on the research of the application. Meanwhile, the method also shows that under the existing condition, the container twin model is accurate, the reliability and the accuracy of the twin model in the aspect of simulating actual operation data are ensured, and a reliable basis is provided for subsequent application and analysis.

According to the method, the stress cloud image capable of capturing the stress and strain constitutive relation is successfully generated by analyzing and extracting the characteristics of the twin model in the virtual environment; dividing the stress density image subjected to equal proportion conversion by using a mean shift method to obtain a clustering centroid as a feature mapping region of a digital space and a physical entity, thereby mapping a physical entity sampling region and realizing global mapping; the result shows that the twin model accords with the accuracy standard by comparing the stress data acquired by the physical entity with the virtual analysis data of the twin model. The method provided by the invention can accurately identify the important stress area required by mapping between the digital twin lightweight model and the physical entity, and accurately find the corresponding mapping area, thereby mapping out the sampling area of the physical entity and establishing a reliable mapping area. The method provides an effective auxiliary means for the application and development of the digital twin technology in the aspect of automobile weight reduction.

Through the application of the digital twin technology, the automobile light-weight design and optimization can be more accurate and efficient. Through accurate mapping, researchers and engineers can simulate various optimization schemes in a digital model, thereby reducing the actual test and development costs. In summary, the present application provides powerful support for automobile weight saving and demonstrates its potential in achieving two carbon targets and driving sustainable development.

The embodiment of the application provides a computer device, which comprises a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, when the device runs, the processor and the memory are communicated through the bus, and the processor executes the machine-readable instructions to execute the steps of the global mapping method between any digital twin lightweight model and a physical entity.

In particular, the above memory and the processor can be general-purpose memories and processors, which are not limited herein, and when the processor runs a computer program stored in the memory, the above method for global mapping between the digital twin lightweight model and the physical entity can be performed.

It will be appreciated by those skilled in the art that the structure of the computer device is not limiting of the computer device and may include more or fewer components than shown, or may be combined with or separated from certain components, or may be arranged in a different arrangement of components.

In some embodiments, the computer device may further include a touch screen operable to display a graphical user interface (e.g., a launch interface of an application) and to receive user operations with respect to the graphical user interface (e.g., launch operations with respect to the application). A particular touch screen may include a display panel and a touch panel. The display panel may be configured in the form of an LCD (Liquid Crystal Display), an OLED (Organic Light-Emitting Diode), or the like. The touch panel may collect touch or non-touch operations on or near the user and generate preset operation instructions, for example, operations of the user on or near the touch panel using any suitable object or accessory such as a finger, a stylus, or the like. In addition, the touch panel may include two parts, a touch detection device and a touch controller. The touch detection device detects the touch azimuth and the touch gesture of a user, detects signals brought by touch operation and transmits the signals to the touch controller; the touch controller receives touch information from the touch detection device, converts the touch information into information which can be processed by the processor, sends the information to the processor, and can receive and execute commands sent by the processor. In addition, the touch panel may be implemented by various types such as resistive, capacitive, infrared, and surface acoustic wave, or may be implemented by any technology developed in the future. Further, the touch panel may overlay the display panel, and a user may operate on or near the touch panel overlaid on the display panel according to a graphical user interface displayed by the display panel, and upon detection of an operation thereon or thereabout, the touch panel is transferred to the processor to determine a user input, and the processor then provides a corresponding visual output on the display panel in response to the user input. In addition, the touch panel and the display panel may be implemented as two independent components or may be integrated.

Corresponding to the method for starting the application program, the embodiment of the application also provides a storage medium, and the storage medium stores a computer program, and the computer program executes the steps of the global mapping method between any digital twin lightweight model and a physical entity when being run by a processor.

The starting device of the application program provided by the embodiment of the application program can be specific hardware on the equipment or software or firmware installed on the equipment. The device provided in the embodiments of the present application has the same implementation principle and technical effects as those of the foregoing method embodiments, and for a brief description, reference may be made to corresponding matters in the foregoing method embodiments where the device embodiment section is not mentioned. It will be clear to those skilled in the art that, for convenience and brevity, the specific operation of the system, apparatus and unit described above may refer to the corresponding process in the above method embodiment, which is not described in detail herein.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of modules is merely a logical function division, and there may be additional divisions in actual implementation, and for example, multiple modules or components may be combined or integrated into another system, or some features may be omitted, or not performed. In addition, the coupling or direct coupling or communication connection shown or discussed with respect to each other may be through some communication interface, indirect coupling or communication connection of devices or modules, electrical, mechanical, or other form.

The modules illustrated as separate components may or may not be physically separate, and components shown as modules may or may not be physical modules, i.e., may be located in one place, or may be distributed over a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional module in the embodiments provided in the present application may be integrated in one processing module, or each module may exist alone physically, or two or more modules may be integrated in one module.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present application and not for limiting the same, and although the present application is described in detail with reference to the above embodiments, it should be understood by those of ordinary skill in the art that: modifications and equivalents may be made to the specific embodiments of the present application without departing from the spirit and scope of the application, and any modifications or equivalents are intended to be encompassed within the scope of the claims of the present application.

Claims

1. The global mapping method between the digital twin lightweight model and the physical entity is characterized by comprising the following steps of:

performing global mapping between the digital twin light weight model of the light weight product and the physical entity by utilizing a global mapping area of the final digital twin light weight model and the physical entity;

a, performing nuclear density estimation:

different working conditions are in three-dimensional spaceIs given by->Data unit->，/>，/>Belongs to point clouds converted from stress cloud pictures; by kernel function->And symmetrically positive->Bandwidth matrix->For parameters, calculate at the point +. >Multi-kernel density estimation at:

wherein,is a kernel function with a bandwidth matrix D, D represents the determinant of the bandwidth matrix D;

three-dimensional kernel functionIs a bounded function that satisfies the following conditions:

wherein,is a constant, I is an integer set, | +.>II represents vector +.>Is (are) norms of->Representation vector->Is a transpose of (2);

the multidimensional kernel function is derived from a symmetric univariate kernel function by two different methodsGenerating:

wherein,obtained from the product of the univariate kernels; />By->Middle rotation->Obtained, i.e.)>Is radially symmetrical,>is a constant;

considering a radially symmetric kernel function, then the symmetric multidimensional kernel function satisfies:

wherein for a function called kernel function profileNeed to make->Normalized constant->When strictly positive, kernel function->The integral is 1;

making the bandwidth matrix D proportional to the same matrixThe method comprises the steps of carrying out a first treatment on the surface of the Providing a bandwidth parameter->At this time, the kernel density estimator is:

；

and B, estimating probability density gradient:

calculating the density gradient:

assume thatThe derivative of (2) is +.>If the kernel function exists in the interval, the kernel function is defined>The method comprises the following steps:

wherein,，/>is a kernel contour function +.>Is used for the purpose of determining the derivative of (c),

is a corresponding normalization function, and is introduced into a density gradient formula to obtain the following components:

calculated out The density estimate at that point is proportional to:

the mean shift vector is:

use of coresGAs a weight, andas the core, there are:

will be as followsWindow function with a bandwidth h for the center +.>Along the mean shift vector->Translation, obtaining a new window function->；

C, all density storages in the stress density image are seed points of the mean shift program; for each seed pointSetting the drift radius h, calculating to +.>Kernel density weighted average vector +.f for all other sample points within the centered, radius h window>For each seed point +.>Performing translation, i.e.)>Until the seed point converges, i.e. +.>The method comprises the steps of carrying out a first treatment on the surface of the Performing mean shift iteration to realize density image segmentation;

2. The global mapping method between a digital twin lightweight model and a physical entity according to claim 1, wherein the performing finite element analysis on the digital twin lightweight model of the lightweight product to obtain a stress and strain cloud image under a virtual environment of the lightweight product comprises:

3. The global mapping method between a digital twin lightweight model and a physical entity according to claim 2, wherein the dividing the stress density image based on the mean shift method to obtain a global mapping region of the digital twin lightweight model and the physical entity comprises:

dividing the stress density image to form a limited clustering mass center;

4. The global mapping method between a digital twin lightweight model and a physical entity according to claim 3, wherein the training parameters based on a single variable method to obtain a global mapping region of a final digital twin lightweight model and the physical entity comprises:

5. The global mapping method between a digital twin lightweight model and a physical entity according to any of claims 1-4, wherein the converting the stress-strain cloud into an equal proportion stress density image based on stress and strain constitutive relations and color gradient change rules comprises:

and A, determining the stress and strain constitutive relation:

wherein,for stress->Is strain; />Is the magnitude of the acting force; />Is the area of force application; />Indicating the amount of change in length; />Is the initial length;

the proportional relationship between discrete cell size and pixel size is:

wherein,the stress and strain module discrete unit size of the digital twin light model is adopted; / >The whole model size of the digital twin light model stress and strain module is provided; />Is the pixel cell size; />The whole pixel size of the color cloud image which needs to be subjected to data conversion;

obtaining a final density image through pixel adjustment;

storing color stress strain cloud image pixel units in a 3D array of m multiplied by n multiplied by 3, and taking a red channel value of each pixel unit as a reference value of stress density;

defining the density storage unit of the stress density image converted by the pixel unit of the position g= (i, j) asIts density value is->Wherein i is more than or equal to 1 and less than or equal to m; j is more than or equal to 1 and less than or equal to n; 0.ltoreq.L>≤255；

Red channel values are extracted and returned to a matrix of size mxn:

wherein,in the matrix, a density storage unit is provided, the value of which is equal to the RGB red channel value of each pixel unit, and the position of which is g= (i, j); matrix density storage unit->The density value in the memory is the density memory cell of the stress density imageDensity value of->；

Density storage unit in stress density image based on Gaussian distribution functionInternal random generation- >And (3) carrying out point separation to obtain a point set D, wherein the point set D is a density point set capable of carrying out density image segmentation.

6. The method of global mapping between a digital twin lightweight model and a physical entity according to any of claims 1-4, further comprising:

7. A global mapping device between a digital twin lightweight model and a physical entity, comprising:

the global mapping module is used for carrying out global mapping between the digital twin light weight model of the light weight product and the physical entity by utilizing the final global mapping area of the digital twin light weight model and the physical entity;

a, performing nuclear density estimation:

different working conditions are in three-dimensional spaceIs given by->Data unit->，/>，/>Belongs to point clouds converted from stress cloud pictures; by kernel function->And symmetrically positive->Bandwidth matrix->For parameters, calculate at the point +.>Multi-kernel density estimation at:

Wherein,is a constant, I is an integer set, | +.>II represents vector +.>Is of (2)Count (n)/(l)>Representation vector->Is a transpose of (2);

wherein, for the sake of scaleAs a function of kernel function contoursNeed to make->Normalized constant->When strictly positive, kernel function->The integral is 1;

；

and B, estimating probability density gradient:

calculating the density gradient:

calculated outThe density estimate at that point is proportional to:

the mean shift vector is:

use of coresGAs a weight, andas the core, there are:

will be as followsWindow function with a bandwidth h for the center +.>Along the mean shift vector- >Translation, obtaining a new window function->；

8. A computer device comprising a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication via the bus when the computer device is in operation, the processor executing the machine-readable instructions to perform the steps of the method of global mapping between a digital twin lightweight model as defined in any one of claims 1-6 and a physical entity.

9. A storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method of global mapping between a digital twin lightweight model and a physical entity as claimed in any of claims 1 to 6.