A performance twin construction method and system

Abstract

This invention discloses a method and system for constructing a performance twin, comprising: establishing a theoretical simulation model of the assembly object based on its geometric and physical parameters and motion constraints, obtaining low-precision model input samples; collecting measured data of the geometric and physical parameters of the assembly object, obtaining high-precision model input samples; constructing a performance twin by fusing the high-precision and low-precision model input samples based on a variable credibility surrogate modeling method; predicting the assembly accuracy and structural mechanical properties of the assembly object based on the performance twin, and evaluating to obtain a performance twin that meets the performance requirements. This invention constructs the performance twin using a variable credibility surrogate model and considers a dual-function adaptive selection method (kernel function and learning function) instead of traditional manual selection, improving the modeling efficiency of the high-fidelity performance twin system, and is of great significance for performance monitoring, optimization control, and assembly regulation of the assembly object.

Description

Technical Field

[0001] This invention belongs to the field of digital twin technology and relates to a method and system for constructing a performance twin. Background Technology

[0002] For complex precision assemblies, assembly accuracy directly affects their service performance. Precision mechanisms have extremely high requirements for assembly accuracy, such as rotational accuracy, surface accuracy, and pointing accuracy. Generally speaking, assembly accuracy directly determines the service performance of the assembled object; the higher the accuracy, the better the service performance. However, the key to improving assembly accuracy and performance lies first in the positive prediction of the assembly performance of the assembled object, establishing a mapping relationship between the geometric motion parameters of the parts and the assembly performance. Traditional methods for obtaining assembly performance parameters mainly involve theoretical simulation and experimental testing. Theoretical simulation includes geometric modeling methods and finite element modeling methods. However, these methods can only explore the formation and derivation laws of assembly performance from local problems or assumed conditions. When solving complex precision mechanical assembly performance problems with obvious multi-factor, multi-scale, strongly coupled, and highly nonlinear characteristics, they cannot fully reflect the formation process of assembly performance, resulting in generally low analysis and prediction accuracy. However, when considering measured data, the cost of acquiring measured data is usually very high due to factors such as data acquisition costs and production scheduling. Therefore, relying solely on theoretical simulation modeling or measurement data training will have bottlenecks such as insufficient prediction accuracy or lack of a large amount of data, failing to completely solve the problem of accurate prediction of assembly performance.

[0003] The variable confidence approximation model method, which has seen significant development in recent years, offers a new solution. It uses a large number of low-cost, low-precision sample points to predict trends, while employing a very small number of high-cost, high-precision sample points to correct the calculation results. This effectively balances the contradiction between the predictive performance and modeling cost of the approximation model. Generally, considering the sample size and the accuracy of the samples themselves, the surrogate model's prediction accuracy often reflects the true situation of the predicted mapping relationship more accurately with a larger sample size, and the sample accuracy is always directly proportional to the surrogate model's prediction accuracy. A way to significantly improve the predictive performance of the surrogate model is to use a large amount of high-precision measured data as input. However, considering the difficulty in obtaining large amounts of measured data at the assembly site of complex equipment, this approach is not economically feasible from an engineering perspective. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for constructing performance twins, thereby solving the problem that assembly performance cannot be accurately predicted in the prior art.

[0005] To achieve the above objectives, the present invention employs the following technical solution:

[0006] A method for constructing a performance twin includes:

[0007] Based on the geometric and physical parameters and motion constraints of the assembly object, a theoretical simulation model of the assembly object is established as an input sample for the low-precision model.

[0008] Collect measured data of the geometric and physical parameters of the assembled object as input samples for the high-precision model;

[0009] Based on the variable credibility proxy model modeling method, a performance twin is constructed by fusing high-precision model input samples and low-precision model input samples;

[0010] Based on the performance twin, the assembly accuracy and structural mechanical properties of the assembled object are predicted, and a performance twin that meets the performance requirements is obtained.

[0011] Furthermore, the geometric physical parameters include geometry, dimensions, tolerances, elastic modulus, stiffness, Poisson's ratio, and density.

[0012] Furthermore, the motion constraint relationships include component geometric constraints, assembly fit relationships, frictional forces, and contact parameters.

[0013] Furthermore, the theoretical simulation model is a finite element analysis model, which includes a finite element static model and a finite element dynamic model.

[0014] Furthermore, the method for acquiring measured data of the geometric and physical parameters of the assembled object includes:

[0015] For geometric information, a 3D scanner is used to obtain the point cloud coordinates of the geometric structural parameters of the assembly object, and digital measuring equipment is used to measure the dimensional information and manufacturing errors of the parts.

[0016] For physical information, a three-degree-of-freedom stiffness testing system is used to obtain the stiffness information of the assembled object parts, industrial sensors are fixed on the assembled object to obtain motion state parameters, and strain gauges are attached to obtain stress and strain information.

[0017] Furthermore, the variable credibility proxy model employs the Co-Kriging method, and the kernel function of the variable credibility proxy model is selected through an adaptive selection method.

[0018] Furthermore, the method for constructing the performance twin includes:

[0019] Based on the Latin hypercube experimental design method, an experimental design for a performance twin prediction model for assembly objects was completed.

[0020] Based on the initial experimental sample set obtained from the experimental design, the input samples for the low-precision model are substituted into the theoretical simulation model to obtain low-precision output samples, and the input samples for the high-precision model are substituted into the experimental test to obtain high-precision output samples, thus completing the establishment of high- and low-precision input and output samples;

[0021] The kernel function-based adaptive selection method selects relevant functions for the surrogate model. Based on the established high and low precision input and output samples, the variable confidence surrogate model method is used to construct an initial prediction performance approximation model, and a performance twin for predicting the structural parameters and mechanical properties of the assembly object is built.

[0022] Furthermore, the performance twin includes data acquisition and fusion, data transmission, data simulation analysis and algorithm solving, assembly 3D model construction, real-time prediction and display feedback of assembly performance, and interactive mapping and optimization adjustment.

[0023] Furthermore, the evaluation method for the performance twin that meets the performance requirements is as follows:

[0024] Based on the real-time prediction results of the assembly accuracy and structural mechanical performance of the assembled object by the performance twin, the performance twin is trained to determine whether the performance twin meets the evaluation requirements of the assembly accuracy and structural mechanical performance of the assembled object.

[0025] If the performance evaluation requirements are met, there is no need to update the initial sample set;

[0026] If the performance evaluation requirements are not met, the initial sample set of the digital twin that does not meet the performance requirements is updated by using an adaptive selection of learning function.

[0027] A performance twin construction system, comprising:

[0028] The modeling module is used to establish a theoretical simulation model of the assembly object based on its geometric and physical parameters and motion constraints, which serves as a low-precision model input sample.

[0029] The acquisition module is used to acquire measured data of the geometric and physical parameters of the assembly object as input samples for a high-precision model.

[0030] A performance twin construction module is used to construct a performance twin based on a variable credibility proxy model modeling method by fusing high-precision model input samples and low-precision model input samples.

[0031] An evaluation module is used to predict the assembly accuracy and structural mechanical properties of the assembled object based on the performance twin, and to evaluate and obtain a performance twin that meets the performance requirements.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] This invention provides a method for constructing a performance twin. It utilizes a variable reliability surrogate model to fuse input samples from high- and low-precision models, and considers a dual-function adaptive selection method to construct a performance twin based on both models. The constructed performance twin possesses real-time prediction and feedback capabilities for the structural and assembly performance of the equipment entity, and the predicted performance feedback from the performance twin is used for performance evaluation. Key points of samples are dynamically added to the high- and low-precision models to update the digital twin model. If the performance evaluation does not meet the requirements, an adaptive selection learning function method is used to update the initial samples. The updated assembly object's prediction accuracy and performance are then judged to meet the requirements. If they do, an optimized performance twin is output; otherwise, samples are added again to update the performance twin. This invention uses a variable reliability surrogate model to construct the performance twin, and considers a dual-function adaptive selection method using kernel functions and learning functions to replace traditional manual selection, improving the modeling efficiency of high-fidelity performance twin systems. This is of great significance for performance monitoring, optimization control, and assembly regulation of assembly objects.

[0034] Furthermore, the adaptive selection method based on the variable credibility surrogate model framework can add points at the locations of interest based on different learning function point-addition strategies and perform prediction performance evaluation, reducing the need for a large initial sample set. On the other hand, the adaptive selection of the kernel function also improves the prediction performance of the performance twin in reflecting the structural mechanical properties of the assembled object. Attached Figure Description

[0035] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the technical route for the performance twin construction method of the present invention.

[0037] Figure 2 This is a schematic diagram of the performance twin construction process of the present invention.

[0038] Figure 3 This is a schematic diagram of the digital twin system based on performance prediction of assembled objects according to the present invention.

[0039] Figure 4 This is a schematic diagram of the data information processing of the present invention. Detailed Implementation

[0040] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0041] Obviously, the described embodiments are only some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0042] It should be noted that the terminals involved in the embodiments of this application may include, but are not limited to, mobile phones, personal digital assistants (PDAs), wireless handheld devices, tablet computers, personal computers (PCs), MP3 players, MP4 players, wearable devices (e.g., smart glasses, smartwatches, smart bracelets), smart home devices, and other smart devices.

[0043] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0044] The present invention will now be described in further detail with reference to the accompanying drawings:

[0045] See Figure 1 This invention provides a method for constructing a performance twin, comprising the following steps:

[0046] Step 1: Based on the geometric and physical parameters and motion constraints of the assembly object, establish a theoretical simulation analysis model of the assembly object as a low-precision model input sample; obtain measured data by conducting experimental tests on the assembly object through an actual physical measurement system, which will serve as a high-precision model input sample.

[0047] The geometric and physical parameters of the assembled object include its geometric shape, dimensions, tolerances, elastic modulus, stiffness, Poisson's ratio, and density.

[0048] The motion constraints of an assembly include the geometric constraints of parts, assembly fit relationships, friction and contact parameters, etc.

[0049] The theoretical simulation model adopts the finite element analysis model, which mainly includes the finite element static model and the dynamic model. Its advantage is that the theoretical simulation model can be changed at any time as the geometric and physical parameters of the assembled object change, which facilitates the real-time performance twin to predict and update the spatial geometric motion relationship and structural mechanical performance of the physical equipment.

[0050] The low-precision sample data of the theoretical simulation model can also adopt the variable confidence model approach, and construct a new nested model with high and low precision based on the mesh density of the finite element model.

[0051] Methods for real-time acquisition of the geometric and physical parameters of assembled objects include:

[0052] For geometric information, a 3D scanner is used to obtain the point cloud coordinates of the geometric structural parameters of the assembly object, and digital measuring equipment is used to measure the dimensional information and manufacturing errors of the parts.

[0053] For physical information, a three-degree-of-freedom stiffness testing system is used to obtain the stiffness information of the assembled object parts, industrial sensors are fixed on the assembled object to obtain motion state parameters, and strain gauges are attached to obtain stress and strain information.

[0054] Step Two: Based on the variable confidence surrogate modeling method, a performance twin with high and low accuracy models is constructed by integrating theoretical analysis models and measured data. The kernel function (correlation function) of the variable confidence surrogate model is selected using an adaptive selection method, replacing the poor predictive performance of the surrogate model caused by the lack of prior information in traditional manual selection. A multi-source, high-dimensional, high-fidelity, real-time online performance prediction digital twin system for precision equipment is constructed.

[0055] Step 2-1: Based on the Latin hypercube experimental design method, complete the experimental design of the performance twin prediction model for the assembly object;

[0056] Step 2-2: Based on the initial experimental sample set obtained from the experimental design, input the low-precision samples into the theoretical simulation model to obtain low-precision output samples, and input the high-precision samples into the experimental test to obtain high-precision output samples, thus completing the establishment of high- and low-precision input and output samples;

[0057] Steps 2-3: The kernel function-based adaptive selection method selects a correlation function with high prediction efficiency for the surrogate model. This kernel function has an advantage in sample prediction accuracy or prediction time in a single iteration.

[0058] Steps 2-4: Based on the established high and low precision input and output samples, construct an initial prediction performance approximation model using the variable confidence surrogate model method, build a performance twin for predicting the structural parameters and mechanical properties of the assembly object, and provide feedback on real-time prediction results to guide the updating of the performance twin prediction model and the optimization and adjustment of the assembly parameters of the assembly entity.

[0059] The performance twin mainly includes data acquisition and fusion, data transmission, data simulation analysis and algorithm solving, assembly 3D model construction, real-time prediction and display feedback of assembly performance, and human-computer interaction and optimization adjustment.

[0060] Step 3: The performance twin predicts the assembly accuracy and structural mechanical properties of the assembled object in real time to determine whether the performance evaluation requirements are met. If the requirements are met, there is no need to update the initial sample set. If not, the initial sample set of the digital twin that does not meet the performance requirements is updated by adaptively selecting the learning function until the predicted performance requirements are met.

[0061] An adaptive selection method for learning functions is proposed, which overcomes the drawbacks of traditional fixed single learning functions when collecting samples from a large number of engineering instances. The added samples update the initial sample set, thereby improving the uniformity and robustness of the experimental design input sample space.

[0062] Step 4: Based on the real-time performance twin prediction results, add the new sample points to the variable confidence model to update the performance twin. Repeat steps 2 and 3, and then determine whether the product prediction accuracy and performance meet the requirements. If they do, output the updated performance twin.

[0063] The geometric and physical information collected in real time is saved as historical data and stored in the performance twin's data storage system for updating and optimizing the performance twin.

[0064] The specific implementation method of the performance twin construction method of the present invention is as follows:

[0065] (1) Establishing a theoretical simulation model

[0066] Finite element simulation analysis of the assembly object is carried out based on geometric and physical parameters. Specifically, parametric modeling is performed using Ansys finite element analysis software. The errors and flexibility of the components are fully considered. The equivalent modeling method is used to realize the input of part errors and assembly errors, and a theoretical simulation static model of the assembly object is established. Dynamic simulation analysis of the assembly object is carried out. Specifically, parametric modeling is performed using Adams simulation software. Parameter requirements such as contact parameters, friction parameters, and driving parameters are set, and a theoretical simulation dynamic model of the assembly object is established.

[0067] (2) Actual measurement of the geometric and physical parameters of the assembled object

[0068] A 3D scanner is used to obtain the point cloud coordinates of relevant parts, which facilitates the construction of a 3D rendering model of the assembly object and the study of key points. A coordinate measuring machine is used to measure the initial manufacturing error of the parts, and a laser tracker is used to measure the positioning error of the parts in the assembly object. The measured part errors are used as input variables for the model. Corresponding industrial sensors are placed on the physical equipment, and displacement sensors and acceleration sensors are used to measure displacement and motion acceleration, respectively. Strain gauges are attached to obtain the stress and strain magnitudes at key locations of the parts.

[0069] (3) Construction of assembly performance twin based on surrogate model

[0070] Figure 2 This is a schematic diagram of the performance twin construction of the present invention. The specific implementation steps are as follows:

[0071] (3-1) Use the Latin hypercube experimental design method to generate nested high and low fidelity samples as input samples for the variable confidence approximation model;

[0072] (3-2) Low-fidelity samples are substituted into the theoretical simulation model to obtain low-precision sample simulation data, which serves as the low-precision output sample of the variable confidence approximation model. Considering that this variable confidence surrogate model can further reduce computational costs and is applicable to single scenarios with multiple fidelities rather than dual fidelity, the generated low-precision sample data is divided into dense and sparse meshes by theoretical simulation finite element analysis, ultimately producing high and low precision multi-fidelity sample output data under the low-precision sample case.

[0073] (3-3) High-fidelity samples are used in experimental testing to obtain measured data, which serves as the high-precision output samples for the variable confidence approximation model. A multi-fidelity level numerical prediction model is established based on the variable confidence surrogate model by integrating high- and low-fidelity sample models.

[0074] (3-4) Considering multi-fidelity recursive form Co-Kriging surrogate modeling

[0075] Figure 3 The diagram shows the digital twin system of the present invention. The variable fidelity proxy model is modeled based on the nesting relationship between high and low fidelity samples, satisfying the condition that the high fidelity model data is a subset of the low fidelity model data. Under this condition, the performance twin of high and low precision model fusion can be extended to t fidelity levels.

[0076] Assuming the above problem has k-dimensional design variables, sample t fidelity input samples from the high- and low-precision models:

[0077]

[0078] Where S is the design space for multi-fidelity input variables, t is the fidelity level, and the high-precision input samples that satisfy the nesting relationship are a subset of the low-precision input samples, i.e. N represents the number of samples, where N1 > N2 > ... > N t The high and low precision output samples are represented as follows:

[0079]

[0080] The traditional Kriging surrogate model prediction process is defined as follows:

[0081] Y(x)=μ(x)+δ(x)

[0082] In the formula, μ(x) reflects the overall trend of the predicted response Y, and δ(x) represents a mean of 0 and a variance of σ. 2 The spatial correlation between different points in a stochastic process is expressed by the covariance Cov(δ(x),δ(x′))=σ 2 R(x,x′,θ) is used for characterization.

[0083] Considering the high- and low-fidelity model fusion method with varying fidelity, the prediction output of the recursive Co-Kriging variable-fidelity approximation model is defined as:

[0084] Y t (x)=ρ t-1 (x)Y t-1 (x)+δ t (x)

[0085] The above formula represents the recursive relationship between two adjacent fidelity models, where the fidelity t ≥ 2, ρ t-1 It is the scaling factor that connects adjacent fidelity models.

[0086] The prediction results for Co-Kriging high-fidelity output are calculated as follows:

[0087]

[0088] Its prediction variance is calculated as follows:

[0089]

[0090] In the formula:

[0091]

[0092] Here, ⊙ represents the operation rule for multiplying matrix elements one by one.

[0093] (3-5) Adaptive selection of kernel function (correlation function)

[0094] The correlation function defines the correlation or similarity between two input sample points, and its selection is usually based on the characteristics of the problem, such as data periodicity, trends, linear relationships, etc. Model evaluation criteria based on leave-one-out cross-validation error and maximum likelihood loss are used to judge the generalization performance of the current model kernel function by minimizing the cross-validation error and the negative log-likelihood loss. Among common kernel functions such as Gaussian kernel, cubic kernel, and Matern kernel, the surrogate model containing the kernel function with the smallest validation error or loss value is selected. Since the kernel function considers the interrelationships between sample points, combined with the sample point addition strategy, an adaptive selection that minimizes error can be made in each iteration, improving the modeling efficiency of the surrogate model.

[0095] (4) Digital twin system predicts assembly accuracy and performance of assembly objects in real time.

[0096] Figure 3 This is a schematic diagram of the digital twin system of the present invention. At the beginning, the initial error of the assembled object parts is input, and the real-time collected geometric and physical information is saved as historical data and stored in the twin data storage module for subsequent updates and optimization of the performance twin. Figure 4 This is a schematic diagram of the data information processing of the present invention.

[0097] (4-1) For the assembly performance digital twin system for precision equipment, the first step is to use industrial sensors to collect the spatial positional relationships between the parts of the assembly object in real time, focusing on the spatial geometric relationships and the kinematic relationships of the parts, and constructing an assembly performance twin data acquisition module to extract the performance twin input variables. The algorithm simulation module relies on the data transmission module to provide input data, filters and denoises the transmitted sensor data, and then classifies, cleans, and reduces it. The processed geometric and physical parameters provide the data sensed by the sensors as high-dimensional input variables for the core proxy model of the performance twin. The data fusion module uses expert knowledge, AI algorithms, etc. to fuse multi-source, high-dimensional, and heterogeneous input data, and fuses the proxy model input data based on finite element simulation analysis data and measured data. The proxy model is used to construct the mapping relationship between the spatial position changes of the assembly object and the structural mechanical properties.

[0098] The twin display module collects simulation prediction data provided by the algorithm module and establishes a 3D rendering display of the assembly entity based on the point cloud coordinates of the 3D scanner. The 3D rendering model provides geometric and physical parameters, deformation stress cloud diagrams, etc. of the assembly object under the current motion state based on human-computer interaction information. The changes in the twin can be displayed online by changing the spatial position relationship of the assembly in real time. The twin data storage module stores the historical motion position and performance state of the assembly as saved data for updating the performance twin.

[0099] (4-2) Input the initial error of the components of the assembly object, and the performance twin predicts the assembly accuracy and performance of the assembly object in real time, and evaluates the prediction results of the performance twin on the assembly accuracy and performance.

[0100] (5) Performance twin prediction performance evaluation

[0101] The root mean square error (RMSE) is used as the performance evaluation index, and the deviation between the measured data and the predicted results of the performance twin is used as the standard for evaluating the quality of the performance twin. The evaluation index is expressed as follows:

[0102]

[0103] If the model accuracy meets the design requirements, the output is a performance twin prediction model that considers the fusion of high and low accuracy models. If it does not meet the requirements, the initial performance twin is updated by adding sample points.

[0104] (6) Learning function adaptive selection of sample addition strategy

[0105] Given that a single learning function cannot meet the needs of a large number of engineering instances, and that different learning functions exhibit different performance characteristics for different samples, and considering the changes in the sample set caused by updating samples in the performance twin, it is advisable to aggregate the learning functions and adaptively select them during the iterative point addition process. An "investment-portfolio" learning function selection strategy is proposed, defining the prediction weight w(t) and calculating the cumulative reward G(t) during the iteration process as follows:

[0106]

[0107] G i (t)=δG i (t-1)+w i (t)

[0108] To avoid situations where the selection probability calculation results in the same probability when choosing a learning function, the rewards corresponding to different learning functions are normalized to u(t). Finally, the selection probability of each learning function in the function set is calculated as follows:

[0109]

[0110] In the formula, λ is the balance coefficient. Initially, the selection probability of each learning function is equal. The optimal sample points are obtained by using the learning function for the high-precision sample set and the low-precision sample set respectively. The fidelity level of the sample addition is determined by comparing the function values. For the selection problem between different learning functions, the optimal candidate points of the samples are input into the Kriging model to calculate w(t). The best learning function for this iteration is selected by comparing the probability calculation method. The sample points generated by it are the best sample points. The newly added sample points are input into the experimental test and theoretical simulation models of high precision and low precision to obtain the actual and simulation results of high precision and low precision. They are added to the initial sample set to update the performance twin. Steps (3) and (4) are repeated until the Co-Kriging surrogate model meets the accuracy prediction performance requirements. The performance twin prediction model of the assembly object based on the variable credibility surrogate model is output.

[0111] The present invention provides a performance twin construction system, which includes: a modeling module, a data acquisition module, a performance twin construction module, and an evaluation module.

[0112] The modeling module is used to establish a theoretical simulation model of the assembly object based on its geometric and physical parameters and motion constraints, which serves as a low-precision model input sample.

[0113] The acquisition module is used to acquire measured data of the geometric and physical parameters of the assembly object as input samples for a high-precision model.

[0114] A performance twin construction module is used to construct a performance twin based on a variable credibility proxy model modeling method by fusing high-precision model input samples and low-precision model input samples.

[0115] An evaluation module is used to predict the assembly accuracy and structural mechanical properties of the assembled object based on the performance twin, and to evaluate and obtain a performance twin that meets the performance requirements.

[0116] It is understood that the performance twin-based construction system provided by this invention corresponds to the aforementioned performance twin construction methods. The relevant technical features of the performance twin-based construction system can be referred to the relevant technical features of the performance twin construction methods, and will not be repeated here.

[0117] Another object of the present invention is to provide an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor performing the steps of the performance twin construction method.

[0118] The method for constructing the performance twin includes the following steps:

[0119] Based on the geometric and physical parameters and motion constraints of the assembly object, a theoretical simulation model of the assembly object is established as an input sample for the low-precision model.

[0120] Collect measured data of the geometric and physical parameters of the assembled object as input samples for the high-precision model;

[0121] Based on the variable credibility proxy model modeling method, a performance twin is constructed by fusing high-precision model input samples and low-precision model input samples;

[0122] Based on the performance twin, the assembly accuracy and structural mechanical properties of the assembled object are predicted, and a performance twin that meets the performance requirements is obtained.

[0123] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the performance twin construction method.

[0124] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can 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, etc.) containing computer-usable program code.

[0125] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can 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, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0126] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0127] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0128] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a performance twin, characterized in that, include: Based on the geometric and physical parameters and motion constraints of the assembly object, a theoretical simulation model of the assembly object is established as an input sample for the low-precision model. Collect measured data of the geometric and physical parameters of the assembled object as input samples for the high-precision model; Based on the variable credibility proxy model modeling method, a performance twin is constructed by fusing high-precision model input samples and low-precision model input samples; Based on the performance twin, the assembly accuracy and structural mechanical properties of the assembled object are predicted, and a performance twin that meets the performance requirements is obtained. The theoretical simulation model is a finite element analysis model, which includes a finite element static model and a finite element dynamic model. The variable credibility proxy model adopts the Co-Kriging method, and the kernel function of the variable credibility proxy model is selected through an adaptive selection method. The evaluation method for the performance twin that meets the performance requirements is as follows: Based on the real-time prediction results of the assembly accuracy and structural mechanical performance of the assembled object by the performance twin, the performance twin is trained to determine whether the performance twin meets the evaluation requirements of the assembly accuracy and structural mechanical performance of the assembled object. If the performance evaluation requirements are met, there is no need to update the initial sample set; If the performance evaluation requirements are not met, the initial sample set of the digital twin that does not meet the performance requirements is updated by using an adaptive selection of learning function. The method for constructing the performance twin includes: Based on the Latin hypercube experimental design method, an experimental design for a performance twin prediction model for assembly objects was completed. Based on the initial experimental sample set obtained from the experimental design, the input samples for the low-precision model are substituted into the theoretical simulation model to obtain low-precision output samples, and the input samples for the high-precision model are substituted into the experimental test to obtain high-precision output samples, thus completing the establishment of high- and low-precision input and output samples; The kernel function-based adaptive selection method selects relevant functions for the surrogate model. Based on the established high and low precision input and output samples, the variable confidence surrogate model method is used to construct an initial prediction performance approximation model, and a performance twin for predicting the structural parameters and mechanical properties of the assembly object is built. The performance twin includes data acquisition and fusion, data transmission, data simulation analysis and algorithm solving, assembly 3D model construction, real-time prediction and display feedback of assembly performance, interactive mapping and optimization adjustment.

2. The method for constructing a performance twin according to claim 1, characterized in that, The geometric and physical parameters include geometry, dimensions, tolerances, elastic modulus, stiffness, Poisson's ratio, and density.

3. The method for constructing a performance twin according to claim 1, characterized in that, The motion constraint relationships include the geometric constraints of the components, the assembly fit relationships, frictional forces, and contact parameters.

4. The method for constructing a performance twin according to claim 1, characterized in that, The method for collecting measured data of the geometric and physical parameters of the assembled object includes: For geometric information, a 3D scanner is used to obtain the point cloud coordinates of the geometric structural parameters of the assembly object, and digital measuring equipment is used to measure the dimensional information and manufacturing errors of the parts. For physical information, a three-degree-of-freedom stiffness testing system is used to obtain the stiffness information of the assembled object parts, industrial sensors are fixed on the assembled object to obtain motion state parameters, and strain gauges are attached to obtain stress and strain information.

5. A performance twin construction system, characterized in that, The steps for implementing the method according to any one of claims 1 to 4 include: The modeling module is used to establish a theoretical simulation model of the assembly object based on its geometric and physical parameters and motion constraints, which serves as a low-precision model input sample. The acquisition module is used to acquire measured data of the geometric and physical parameters of the assembly object as input samples for a high-precision model. A performance twin construction module is used to construct a performance twin based on a variable credibility proxy model modeling method by fusing high-precision model input samples and low-precision model input samples. An evaluation module is used to predict the assembly accuracy and structural mechanical properties of the assembled object based on the performance twin, and to evaluate and obtain a performance twin that meets the performance requirements.

Images