Vehicle part assembling method and system

Abstract

The invention provides a vehicle part assembly method and system, and the method comprises the steps: creating a corresponding digital twin in real time according to the target model of a vehicle part, and starting a preset multi-dimensional perception array; before the vehicle parts enter the assembly station, target part features corresponding to the vehicle parts are collected in real time through a preset multi-dimensional sensing array; acquiring reference parameters corresponding to the digital twin in real time, and performing multi-dimensional mapping comparison on the target part features and the reference parameters based on a preset rule to predict whether assembly errors occur in the assembly process in real time; and if it is predicted in real time that assembly errors occur in the assembly process, corresponding target compensation parameters are generated in real time through an adaptive genetic algorithm so as to correspondingly complete real-time assembly of the vehicle parts. According to the invention, assembly errors can be effectively avoided, and the assembly efficiency is correspondingly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automobiles, in particular to a vehicle part assembly method and system. BACKGROUND

[0002] With the progress of science and technology and the rapid development of productivity, automobiles have been popularized in people's daily life and have become one of the indispensable means of transportation for people's daily travel, greatly facilitating people's life.

[0003] Among them, the existing automobiles are composed of a large number of parts, and each part needs to be assembled correspondingly to generate a complete automobile. Therefore, the assembly accuracy of each part needs to be ensured to ensure the use performance of the vehicle.

[0004] Further, in the process of real-time assembly, it is necessary to detect whether each part is assembled incorrectly in real time. However, in the actual detection process, most of the prior art can only issue corresponding alarm information when the parts have been assembled incorrectly, and then perform corresponding correction processing. This method will increase the assembly time of the parts and reduce the assembly efficiency of the parts. SUMMARY

[0005] Therefore, the purpose of the present application is to provide a vehicle part assembly method and system to solve the problem that most of the prior art can only issue corresponding alarm information when the parts have been assembled incorrectly, which increases the assembly time of the parts and reduces the assembly efficiency of the parts.

[0006] The first aspect of the embodiment of the present application is proposed: A vehicle part assembly method, wherein the method comprises: According to the target model of the vehicle part, a corresponding digital twin is created in real time, and a preset multi-dimensional perception array is enabled; Before the vehicle part enters the assembly station, the target part features corresponding to the vehicle part are collected in real time by the preset multi-dimensional perception array; The reference parameters corresponding to the digital twin are collected in real time, and the target part features and the reference parameters are mapped and compared in multiple dimensions based on a preset rule to predict whether an assembly error will occur in the assembly process in real time; If it is predicted in real time that an assembly error will occur in the assembly process, a corresponding target compensation parameter is generated in real time by an adaptive genetic algorithm to complete the real-time assembly of the vehicle part.

[0007] The beneficial effects of the present application are: by constructing a digital twin corresponding to the vehicle parts in real time, the reference parameters of the current parts can be obtained correspondingly, based on which, the target part features corresponding to the current vehicle parts are collected in real time through the preset multi-dimensional perception array, based on which, corresponding multi-dimensional mapping comparison processing can be carried out, and whether the current vehicle parts will appear assembly error can be judged in real time according to the comparison result, specifically, if so, the adaptive target compensation parameters can be generated immediately, so that the assembly error of the vehicle parts can be effectively avoided, and the assembly time of the parts can be shortened, and the assembly efficiency is improved.

[0008] Further, the step of collecting the target part features corresponding to the vehicle parts in real time through the preset multi-dimensional perception array comprises: When the vehicle parts enter the sensing range of the preset multi-dimensional perception array, the outer contour vertices of the vehicle parts are marked with coordinates through the preset laser positioning sensor to form a corresponding initial positioning coordinate set; The internal structure of the vehicle parts is tomographically scanned through the preset scanner, and the surface texture, aperture size and chamfer angle of the vehicle parts are imaged through the preset camera to generate a corresponding tomographic image set and an external morphology image set; The initial positioning coordinate set, the tomographic image set and the external morphology image set are characterized to generate the target part features correspondingly.

[0009] Further, the step of characterizing the initial positioning coordinate set, the tomographic image set and the external morphology image set to generate the target part features correspondingly comprises: The initial positioning coordinate set is reconstructed in space topology to convert the generated discrete coordinate points into continuous geometric grids, and the normal vector and curvature value of the grid vertices of the continuous geometric grids are extracted to form corresponding three-dimensional structure feature parameters; The tomographic image set is subjected to multi-scale edge detection, different levels of tomographic profiles are obtained through Gaussian pyramid decomposition, and the profile coincidence degree and offset between adjacent tomographic profiles are calculated to generate internal structure feature parameters; The external morphology image set is subjected to pixel feature extraction, the direction angle and contrast of the surface texture are calculated through the gray level co-occurrence matrix, and the inclination angle is obtained through the least square method fitting chamfer edge line to form appearance morphology feature parameters, and the three-dimensional structure feature parameters, the internal structure feature parameters and the appearance morphology feature parameters are normalized to generate the target part features.

[0010] Further, the step of mapping and comparing the target part feature with the reference parameter in multiple dimensions based on the preset rule to predict in real time whether an assembly error will occur in the assembly process comprises: reconstructing the target part feature in three-dimensional point cloud to extract a three-dimensional feature set containing geometric profile, surface roughness and key hole coordinate, and converting the three-dimensional feature set into a standard data matrix; generating a reference matrix adapted to the three-dimensional feature set in real time according to the reference parameter, and performing dimension alignment on the standard data matrix and the reference matrix by principal component analysis to generate a corresponding feature mapping relationship graph; Based on the feature mapping relationship graph, the cosine of the feature vector and the Euclidean distance are calculated. When the cosine or the Euclidean distance exceeds the preset threshold, it is determined that there is a risk of assembly error, and the error warning coordinate area is output.

[0011] Further, the step of calculating the cosine of the feature vector and the Euclidean distance based on the feature mapping relationship graph comprises: extracting a target feature vector group corresponding to the three-dimensional feature set and a reference feature vector group corresponding to the reference matrix from the feature mapping relationship graph; According to the corresponding relationship established in the feature mapping relationship graph, the target feature vector and the reference feature vector are paired and processed; For each pair of paired vectors, the cosine of the two vectors is calculated by vector dot product formula, and the Euclidean distance of the two vectors is calculated by space rectangular coordinate system for corresponding judgment.

[0012] Further, the step of generating corresponding target compensation parameters in real time by adaptive genetic algorithm to complete real-time assembly of the vehicle parts comprises: Based on the predicted assembly error, an initial gene population containing three-dimensional coordinate compensation amount, angle correction value and pressure adjustment coefficient is constructed; Perform adaptive selection on the initial gene population, adjust the selection probability according to the error correction effect, and recombine the effective compensation features of different gene fragments by crossover operator; According to the real-time working condition of the assembly station, the mutation probability is dynamically adjusted, and the converged gene fragments are output as the target compensation parameters after iteration.

[0013] Further, the step of dynamically adjusting the mutation probability according to the real-time working condition of the assembly station, and outputting the converged gene fragments as the target compensation parameters after iteration comprises: Real-time acquisition of the environmental temperature and humidity, mechanical arm vibration frequency and part surface temperature of the assembly station to construct a working condition feature matrix; Input the working condition characteristic matrix into a preset model to generate a benchmark variation probability coefficient and correct the benchmark variation probability coefficient in combination with an iteration number. An error simulation value is calculated after each iteration, and when the error fluctuation of three consecutive iterations is less than a preset fluctuation threshold, the current gene fragment is determined as the target compensation parameter.

[0014] The second aspect of the embodiment of the application provides: A vehicle part assembly system, wherein the system comprises: A creation module configured to create a corresponding digital twin in real time according to a target model of a vehicle part, and enable a preset multi-dimensional perception array; A collection module configured to collect target part features corresponding to the vehicle part in real time through the preset multi-dimensional perception array before the vehicle part enters an assembly station; A mapping module configured to collect benchmark parameters corresponding to the digital twin in real time, and perform multi-dimensional mapping comparison between the target part features and the benchmark parameters based on a preset rule to predict whether an assembly error will occur in the assembly process in real time; A prediction module configured to generate corresponding target compensation parameters in real time through an adaptive genetic algorithm to complete real-time assembly of the vehicle part if it is predicted that an assembly error will occur in the assembly process in real time.

[0015] Further, the collection module is specifically configured to: When the vehicle part enters a sensing range of the preset multi-dimensional perception array, mark coordinates of an outer contour vertex of the vehicle part through a preset laser positioning sensor to form an initial positioning coordinate set; Perform tomographic scanning on an internal structure of the vehicle part through a preset scanner, and perform image collection on a surface texture, an aperture size, and a chamfer angle of the vehicle part through a preset camera to generate a tomographic image set and an external morphology image set; Perform feature processing on the initial positioning coordinate set, the tomographic image set, and the external morphology image set to generate the target part features correspondingly.

[0016] Further, the collection module is specifically configured to: Reconstruct a spatial topology of the initial positioning coordinate set to convert discrete coordinate points generated into continuous geometric meshes, and extract a normal vector and a curvature value of a mesh vertex of the continuous geometric meshes to form corresponding three-dimensional structure feature parameters; Perform multi-scale edge detection on the tomographic image set, obtain tomographic contours at different levels through Gaussian pyramid decomposition, and calculate a profile coincidence degree and an offset between adjacent tomographic to generate internal structure feature parameters; The external appearance image set is subjected to pixel feature extraction, the direction angle and contrast of surface texture are calculated through a gray level co-occurrence matrix, and the inclination angle is obtained by fitting the chamfer edge line through a least square method to form an appearance and morphology feature parameter, and the three-dimensional structure feature parameter, the internal structure feature parameter and the appearance and morphology feature parameter are subjected to normalization processing to generate the target part feature.

[0017] Further, the mapping module is specifically used for: The target part feature is subjected to three-dimensional point cloud reconstruction to extract a three-dimensional feature set containing a geometric contour, surface roughness and key hole coordinate, and the three-dimensional feature set is converted into a standard data matrix; A reference matrix adapted to the three-dimensional feature set is generated in real time according to the reference parameter, and the standard data matrix and the reference matrix are subjected to dimension alignment through a principal component analysis method to generate a corresponding feature mapping relationship graph; The cosine of the angle and the Euclidean distance of the feature vector are calculated based on the feature mapping relationship graph, when the cosine of the angle or the Euclidean distance exceeds a preset threshold, it is determined that there is an assembly error risk, and an error warning coordinate area is output.

[0018] Further, the mapping module is specifically used for: The target feature vector group corresponding to the three-dimensional feature set and the reference feature vector group corresponding to the reference matrix are extracted from the feature mapping relationship graph; The target feature vector and the reference feature vector are paired according to the corresponding relationship established in the feature mapping relationship graph; For each pair of paired vectors, the cosine of the angle is calculated through the vector dot product formula, and the Euclidean distance is calculated through the space rectangular coordinate system for corresponding judgment.

[0019] Further, the prediction module is specifically used for: Based on the predicted assembly error, an initial gene population containing a three-dimensional coordinate compensation amount, an angle correction value and a pressure adjustment coefficient is constructed; The initial gene population is subjected to adaptive selection, the selection probability is adjusted according to the error correction effect, and the effective compensation features of different gene fragments are recombined through a crossover operator; The mutation probability is dynamically adjusted according to the real-time working condition of the assembly station, and the converged gene fragment is output as the target compensation parameter after iteration.

[0020] Further, the prediction module is specifically used for: Collect the environment temperature and humidity, mechanical arm vibration frequency and part surface temperature of the assembly station in real time, and construct a working condition feature matrix; Input the working condition feature matrix into a preset model, generate a reference variation probability coefficient and correct it in combination with the iteration number; Calculate the error simulation value after each iteration, and when the error fluctuation of three consecutive iterations is less than a preset fluctuation threshold, the current gene fragment is determined as the target compensation parameter.

[0021] The third aspect of the embodiment of the application provides: A computer comprises a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the vehicle part assembly method as described above when executing the computer program.

[0022] The fourth aspect of the embodiment of the application provides: A readable storage medium has a computer program stored thereon, wherein the program is executed by a processor to implement the vehicle part assembly method as described above.

[0023] Additional aspects and advantages of the application will be described in part in the description which follows, will become apparent from the description, or will be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A flow chart of the vehicle part assembly method provided by the first embodiment of the application; Figure 2 A structural block diagram of the vehicle part assembly system provided by the third embodiment of the application.

[0025] The following specific implementation will further illustrate the application in combination with the above-mentioned drawings. DETAILED DESCRIPTION

[0026] In order to facilitate the understanding of the application, the application will be described more fully below with reference to the related drawings. The drawings show several embodiments of the application. However, the application can be realized in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the application more thorough and comprehensive.

[0027] It should be noted that when an element is referred to as being "fixed" to another element, it can be directly on the other element or there can be a middle element. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or there can be a middle element. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description herein is for describing particular embodiments only and is not intended to be limiting of the application. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0029] Referring to Figure 1 , a vehicle part assembly method is shown, the vehicle part assembly method can accurately and effectively complete the assembly of vehicle parts, and the assembly efficiency is improved.

[0030] Specifically, the embodiment provides: A vehicle part assembly method, specifically comprising the following steps: Step S10, a corresponding digital twin is created in real time according to the target model of the vehicle part, and a preset multi-dimensional perception array is enabled; It should be noted that for the target model of the vehicle part to be assembled (such as engine cylinder, gearbox gear, etc.), the system constructs a digital twin that is completely consistent with its physical properties in real time — this virtual model can dynamically reflect the geometric parameters, material properties and assembly constraints of the part. At the same time, a preset multi-dimensional perception array (composed of laser sensors, industrial cameras, force sensors, etc.) is enabled to provide hardware support for subsequent feature collection. This step is the basis of digital assembly, and through virtual-real mapping, accurate digital control of physical parts is realized. In order to facilitate subsequent processing.

[0031] Step S20, before the vehicle part enters the assembly station, the target part features corresponding to the vehicle part are collected in real time through the preset multi-dimensional perception array; It should be noted that before the part enters the assembly station, the multi-dimensional perception array scans and detects it in all directions, collecting key features including external contour, internal structure, surface texture, etc. For example, the laser sensor obtains the three-dimensional size of the part, the camera captures the integrity of the surface thread, and the tomography records the internal cavity structure. This step converts the physical part into quantifiable digital features, providing data basis for subsequent error prediction. In order to facilitate subsequent processing.

[0032] Step S30, real-time collection of reference parameters corresponding to the digital twin, and multi-dimensional mapping and comparison of the target part features and the reference parameters based on a preset rule to predict whether an assembly error will occur during assembly in real time; It should be noted that the system extracts preset reference parameters (such as design size, tolerance range, assembly pose standard) from the digital twin, and maps and compares the collected target part features with the reference parameters through preset rules (such as geometric size deviation, hole coordinate offset, surface roughness difference, etc.). Through this comparison, it is judged in real time whether the parts may appear too large fitting gap, position misalignment and other errors in the assembly process. This step is the core of quality control, which identifies potential assembly problems in advance. In order to facilitate subsequent processing.

[0033] Step S40, if it is predicted that assembly error will occur in the assembly process, corresponding target compensation parameters are generated in real time through adaptive genetic algorithm to complete the real-time assembly of the vehicle parts.

[0034] It should be noted that if the assembly error is predicted, the system calculates the corresponding target compensation parameters (such as mechanical arm posture adjustment angle, clamping force correction value, etc.) through adaptive genetic algorithm, and sends the parameters to the assembly equipment in real time to guide the dynamic adjustment of assembly action, and finally completes the precise assembly of the parts. This step realizes the closed-loop control of error, ensures that the assembly precision meets the design requirements. In order to facilitate subsequent judgment and processing.

[0035] Second embodiment Further, the step of collecting the target part features corresponding to the vehicle parts in real time through the preset multi-dimensional perception array comprises: When the vehicle parts enter the sensing range of the preset multi-dimensional perception array, the coordinates of the outer contour vertices of the vehicle parts are marked through the preset laser positioning sensor to form a corresponding initial positioning coordinate set; The internal structure of the vehicle parts is tomographically scanned through the preset scanner, and the surface texture, aperture size and chamfer angle of the vehicle parts are imaged through the preset camera to generate a corresponding tomographic image set and an external morphology image set; The initial positioning coordinate set, the tomographic image set and the external morphology image set are characterized to correspondingly generate the target part features.

[0036] It should be noted that when the parts enter the sensing range of the multi-dimensional perception array, the preset laser positioning sensor emits a laser beam to mark the coordinates of the key vertices (such as corners, hole edge) of the part outer contour, forming an initial positioning coordinate set (such as (x1, y1, z1), (x2, y2, z2)…). These coordinate points accurately reflect the spatial position and overall shape of the part, providing a basic framework for subsequent three-dimensional modeling. Based on this, the part is subjected to internal structure tomography by a preset scanner (such as an industrial CT or ultrasonic scanner), generating a series of tomographic images parallel to different cross sections, completely presenting hidden features such as internal cavities, wall thickness, embedded parts, etc. Through a preset high-resolution camera, the part surface is photographed from multiple angles, the texture pattern, aperture size, chamfer angle, etc. are collected, and an external morphology image set is formed. The system digitally processes the initial positioning coordinate set, the tomographic image set, and the external morphology image set (such as coordinate fitting, image segmentation, parameter extraction), converts the original data into structured feature parameters (such as contour deviation value, aperture tolerance, texture uniformity, etc.), and finally integrates them into “target part features”. This step converts massive raw data into analyzable feature indicators, providing standardized input for subsequent comparison.

[0037] Further, the step of feature processing of the initial positioning coordinate set, the tomographic image set, and the external morphology image set to correspondingly generate the target part features comprises: spatial topology reconstruction of the initial positioning coordinate set to convert the generated discrete coordinate points into continuous geometric grids, and extraction of normal vectors and curvature values of grid vertices of the continuous geometric grids to form corresponding three-dimensional structure feature parameters; multi-scale edge detection of the tomographic image set, obtaining tomographic contours at different levels through Gaussian pyramid decomposition, and calculating profile coincidence and offset between adjacent tomographic images to generate internal structure feature parameters; pixel feature extraction of the external morphology image set, calculating the direction angle and contrast of surface texture through a gray level co-occurrence matrix, and obtaining the inclination angle through least squares fitting of chamfer edge lines to form appearance morphology feature parameters, and normalizing the three-dimensional structure feature parameters, the internal structure feature parameters, and the appearance morphology feature parameters to generate the target part features.

[0038] It should be noted that the initial positioning coordinate set is spatially topologically reconstructed, and the discrete coordinate points are connected into continuous geometric grids (simulating the three-dimensional form of the part) through a triangulation algorithm. On this basis, the normal vector (reflecting the surface orientation) and the curvature value (reflecting the surface bending degree) of the grid vertex are calculated to form the three-dimensional structural feature parameters. For example, a large curvature value may indicate that the part has deformation and needs to be focused on assembly adaptability. Based on this, multi-scale edge detection is performed on the set of cross-sectional images (through Gaussian pyramid decomposition technology, the outline is identified at different resolutions), and the boundary lines of each cross section are extracted. By calculating the coincidence degree (such as 90% or more is qualified) and the offset (such as an allowable deviation of ±0.02mm) of the adjacent cross-sectional outlines, internal structural feature parameters are generated to reflect the consistency and integrity of the internal structure of the part. For example, if the outline offset of two cross sections exceeds the standard, it may mean that the internal channel has deformation. Finally, the direction angle (such as the spiral direction of the thread) and the contrast (reflecting the surface smoothness) of the surface texture are calculated by the gray level co-occurrence matrix, the bevel edge line is fitted by the least squares method, and the inclination angle is accurately calculated. Finally, the three-dimensional structure, internal structure, and appearance form three types of feature parameters are uniformly converted to the interval of 0 to 1 to eliminate the dimensional difference and form the standardized "target part feature". This step ensures that different types of features can be compared comprehensively to improve the analysis accuracy for subsequent processing.

[0039] Further, the step of mapping and comparing the target part feature with the reference parameter in multiple dimensions based on the preset rule to predict in real time whether an assembly error will occur in the assembly process comprises: reconstructing the target part feature into a three-dimensional point cloud to extract a three-dimensional feature set containing geometric outlines, surface roughness, and key hole coordinates, and converting the three-dimensional feature set into a standard data matrix; generating a reference matrix adapted to the three-dimensional feature set in real time according to the reference parameter, and aligning the dimensions of the standard data matrix and the reference matrix by principal component analysis to generate a corresponding feature mapping relationship graph; calculating the cosine of the angle and the Euclidean distance of the feature vector based on the feature mapping relationship graph, and determining that there is a risk of assembly error when the cosine of the angle or the Euclidean distance exceeds a preset threshold, and outputting an error warning coordinate area.

[0040] It should be noted that the system imports the target part features into the three-dimensional point cloud reconstruction algorithm, generates a point cloud model composed of millions of points, extracts key features (such as key line segments of geometric contours, numerical distribution of surface roughness, three-dimensional coordinates of key hole positions) from it, and forms a three-dimensional feature set. Then, these features are converted into a standard data matrix (such as each row representing a feature parameter and each column representing a sampling point), realizing the structured storage of features. Based on this, according to the reference parameters (design standards) of the digital twin, a reference matrix (such as the theoretical value of the hole position coordinates, the upper limit standard of roughness) corresponding to the three-dimensional feature set is generated. Through principal component analysis (PCA), the standard data matrix and the reference matrix are aligned in dimension, specifically, by reducing dimension to retain core features and eliminate data redundancy, ensuring that they are comparable in the same feature space, and finally generating a feature mapping relationship graph (directly showing the correspondence between actual features and reference features). Finally, based on the feature mapping relationship graph, the cosine of the angle between the corresponding feature vectors in the two matrices is calculated (measuring the direction similarity, the value is closer to 1 the more similar) and the Euclidean distance (measuring the numerical difference, the smaller the value the smaller the difference). When any index exceeds the preset threshold (such as the cosine of the angle <0.95 or the Euclidean distance >0.05mm), it is determined that there is a risk of assembly error, and the specific error area (such as the coordinate offset of a certain hole position) is located through the graph, and the error warning coordinates are output. This step realizes the quantification and localization of errors, providing a clear target for subsequent compensation.

[0041] Further, the step of calculating the cosine of the angle between the feature vectors and the Euclidean distance based on the feature mapping relationship graph comprises: extracting a target feature vector group corresponding to the three-dimensional feature set and a reference feature vector group corresponding to the reference matrix from the feature mapping relationship graph; According to the corresponding relationship established in the feature mapping relationship graph, the target feature vector and the reference feature vector are paired and processed; For each pair of paired vectors, the cosine of the angle between them is calculated through the vector dot product formula, and the Euclidean distance between them is calculated through the space rectangular coordinate system, for corresponding judgment.

[0042] It should be noted that two groups of vectors are separated from the feature mapping relationship graph, namely the target feature vector group and the reference feature vector group. It should be noted that the target feature vector group is converted from the target part features, and each vector represents a specific feature (such as "hole diameter size vector" and "chamfer angle vector"). The reference feature vector group is converted from the reference parameters, and is one-to-one corresponding to the target vector. According to the corresponding rule in the feature mapping relationship graph (such as "actual hole diameter" corresponding to "theoretical hole diameter"), the target feature vector and the reference feature vector are paired one by one to form a plurality of vector pairs. This step ensures the consistency of the comparison objects and avoids errors caused by cross comparison. Based on this, the cosine value of the included angle and the Euclidean distance generated between the two are immediately calculated. It should be noted that the cosine value of the included angle is calculated by the formula cosθ=(A•B) / (|A|•|B|), where A and B are paired vectors. The closer the result is to 1, the more consistent the vector direction (feature trend) is. The smaller the result of the Euclidean distance is, the closer the vector value is (the feature size is consistent). For example, if the cosine value of the hole diameter vector pair is 0.99 (the direction is consistent) but the Euclidean distance is 0.2mm (the numerical deviation), it may still be determined that there is a risk of error. This step improves the reliability of error judgment through double indicators. In order to facilitate subsequent processing.

[0043] Further, the step of generating corresponding target compensation parameters in real time through an adaptive genetic algorithm to correspondingly complete real-time assembly of the vehicle parts comprises: Based on the predicted assembly error, an initial gene population containing three-dimensional coordinate compensation, angle correction value and pressure adjustment coefficient is constructed; The initial gene population is subjected to adaptive selection, the selection probability is adjusted according to the error correction effect, and the effective compensation characteristics of different gene fragments are recombined through a crossover operator; The mutation probability is dynamically adjusted according to the real-time working condition of the assembly station, and the converged gene fragments are output as the target compensation parameters after iteration.

[0044] It should be noted that, for the predicted assembly error (such as hole position deviation 0.1mm, angle deviation 1°), the system constructs an initial gene population, wherein each gene represents a set of potential compensation parameter combinations, wherein it should be pointed out that the compensation parameter combination contains three-dimensional coordinate compensation (such as X axis + 0.08mm, Y axis - 0.05mm), angle correction value (such as mechanical arm rotation + 0.8°), pressure adjustment coefficient (such as assembly pressure reduction 10%), based on this, the selection probability is dynamically adjusted according to the error correction effect of the gene (such as the reduction of the simulated assembly deviation), the gene with better effect is retained with higher probability, the effective fragments of different genes (such as coordinate compensation of methyl gene + angle correction of ethyl gene) are recombined to generate new genes with the advantages of both, and the population diversity is improved. Finally, according to the real-time working condition of the assembly station (such as mechanical arm load, environment temperature), the gene mutation probability is dynamically adjusted (such as increasing the mutation probability when the working condition is unstable to explore more compensation schemes). After multiple iterations (such as 50 generations), when the error correction effect of the gene fragment tends to be stable (convergent), it is output as the target compensation parameter. This step ensures that the compensation parameter can adapt to real-time working condition changes and improve assembly robustness. In order to facilitate subsequent processing.

[0045] Further, the step of dynamically adjusting the mutation probability according to the real-time working condition of the assembly station and outputting the convergent gene fragment after iteration as the target compensation parameter comprises: Real-time acquisition of the environment temperature and humidity, mechanical arm vibration frequency and part surface temperature of the assembly station to construct a working condition feature matrix; Input the working condition feature matrix into a preset model to generate a reference mutation probability coefficient and correct it combined with the iteration number; Calculate the error simulation value after each iteration, and when the error fluctuation of three consecutive iterations is less than a preset fluctuation threshold, the current gene fragment is determined as the target compensation parameter.

[0046] It should be noted that the key working condition parameters of the assembly station are collected in real time, i.e. environmental temperature and humidity, mechanical arm vibration frequency, and part surface temperature. These parameters are integrated into a working condition feature matrix to reflect the stability of the current assembly environment. The working condition feature matrix is input into a preset model (such as a neural network trained based on historical data) to generate a benchmark variation probability coefficient (such as 5% under stable working conditions and 15% under unstable conditions). At the same time, the algorithm iteration number is modified: the variation probability is increased in the early iterations to expand the search range, and the probability is reduced in the later iterations to focus on the optimal solution (such as reducing the probability from 10% to 3% after 40 iterations). After each iteration, the assembly process is simulated through the digital twin to calculate the error simulation value (actual deviation after compensation). When the error fluctuation value of three consecutive iterations (such as the error difference between adjacent two iterations is less than 0.0001mm) is less than the preset threshold, it is determined that the gene fragment has converged, and it is determined as the final target compensation parameter for subsequent processing.

[0047] Please refer to Figure 2 The third embodiment of the present application provides: A vehicle part assembly system, wherein the system comprises: A creation module for creating a corresponding digital twin in real time according to the target model of the vehicle part, and enabling a preset multi-dimensional perception array; An acquisition module for acquiring target part features corresponding to the vehicle part in real time through the preset multi-dimensional perception array before the vehicle part enters the assembly station; A mapping module for acquiring reference parameters corresponding to the digital twin in real time, and performing multi-dimensional mapping comparison between the target part features and the reference parameters based on a preset rule to predict in real time whether an assembly error will occur during assembly; A prediction module for generating corresponding target compensation parameters in real time through an adaptive genetic algorithm if it is predicted in real time that an assembly error will occur during assembly, so as to complete the real-time assembly of the vehicle part.

[0048] Further, the acquisition module is specifically configured to: When the vehicle part enters the sensing range of the preset multi-dimensional perception array, the outer contour vertices of the vehicle part are marked with coordinates through a preset laser positioning sensor to form an initial positioning coordinate set; The internal structure of the vehicle part is tomographically scanned through a preset scanner, and the texture, aperture size and chamfer angle of the outer surface of the vehicle part are imaged through a preset camera to generate a tomographic image set and an external morphology image set; characterize the initial positioning coordinate set, the tomographic image set and the external morphological image set to generate the target part feature correspondingly.

[0049] Further, the acquisition module is specifically used for: topologically reconstruct the initial positioning coordinate set to convert the generated discrete coordinate points into a continuous geometric grid, and extract normal vectors and curvature values of grid vertices of the continuous geometric grid to form corresponding three-dimensional structural feature parameters; perform multi-scale edge detection on the tomographic image set, obtain tomographic profiles at different levels through Gaussian pyramid decomposition, and calculate profile coincidence degree and offset between adjacent tomographic profiles to generate internal structural feature parameters; extract pixel features from the external morphological image set, calculate direction angle and contrast of surface texture through a gray level co-occurrence matrix, and obtain an inclination angle through least squares fitting of a chamfer edge line to form appearance morphological feature parameters, and normalize the three-dimensional structural feature parameters, the internal structural feature parameters and the appearance morphological feature parameters to generate the target part feature.

[0050] Further, the mapping module is specifically used for: reconstruct the target part feature into a three-dimensional point cloud to extract a three-dimensional feature set containing geometric profiles, surface roughness and key hole coordinates, and convert the three-dimensional feature set into a standard data matrix; generate a reference matrix adapted to the three-dimensional feature set in real time according to the reference parameters, and perform dimension alignment on the standard data matrix and the reference matrix through principal component analysis to generate a corresponding feature mapping relationship graph; calculate the cosine of the angle and the Euclidean distance of the feature vectors based on the feature mapping relationship graph, and when the cosine of the angle or the Euclidean distance exceeds a preset threshold, determine that there is a risk of assembly error, and output an error warning coordinate area.

[0051] Further, the mapping module is specifically used for: extract a target feature vector group corresponding to the three-dimensional feature set and a reference feature vector group corresponding to the reference matrix from the feature mapping relationship graph; pair the target feature vector and the reference feature vector according to the corresponding relationship established in the feature mapping relationship graph; For each pair of paired vectors, calculate the cosine of the angle through the vector dot product formula, and calculate the Euclidean distance through the space rectangular coordinate system to make a corresponding judgment.

[0052] Further, the prediction module is specifically used for: Based on the predicted assembly error, an initial gene population containing three-dimensional coordinate compensation, angle correction value and pressure adjustment coefficient is constructed; The initial gene population is executed adaptive selection, the selection probability is adjusted according to error correction effect, and the effective compensation characteristics of different gene fragments are recombined through a crossover operator; The mutation probability is dynamically adjusted according to the real-time working condition of the assembly station, and the converged gene fragment is output as the target compensation parameter after iteration.

[0053] Further, the prediction module is specifically used for: The environmental temperature and humidity, mechanical arm vibration frequency and part surface temperature of the assembly station are collected in real time to construct a working condition feature matrix; The working condition feature matrix is input into a preset model to generate a reference mutation probability coefficient and is corrected in combination with the iteration number; The error simulation value is calculated after each iteration, and when the error fluctuation of three consecutive iterations is less than a preset fluctuation threshold, the current gene fragment is determined as the target compensation parameter.

[0054] The fourth embodiment of the application provides a computer, including a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to realize the vehicle part assembly method as described above.

[0055] The fifth embodiment of the application provides a readable storage medium, which stores a computer program, wherein the program is executed by a processor to realize the vehicle part assembly method as described above.

[0056] In summary, the vehicle part assembly method and system provided by the above embodiments of the application can accurately and effectively complete the assembly of vehicle parts, thereby improving the assembly efficiency.

[0057] It should be noted that the above-mentioned modules can be functional modules or program modules, which can be implemented by software or hardware. For modules implemented by hardware, the above-mentioned modules can be located in the same processor;Or the above-mentioned modules can also be located in different processors in any combination.

[0058] The logic and / or steps represented in the flowcharts and / or described herein, for example, can be embodied in non-transitory computer-readable media, executed by one or more computing devices, and / or in any other way. The logic and / or steps represented in the flowcharts and / or described herein, for example, can be considered a list of executable instructions for implementing logic functions, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor- containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For purposes of this specification, a "computer-readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber (optical), and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example via an optical scanner, then compiled, interpreted, or otherwise processed, and stored in a computer memory in a form that can be later executed by a computer. In this context, the computer-readable medium can be paper or another suitable medium that can act to transfer a program, for example, from the Internet, over a wired or wireless network, over the airwaves, etc., in a cartridged, downloadable, or other format. Thus, a computer-readable medium can take many forms of hardware and software.

[0059] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber (optical), and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example via an optical scanner, then compiled, interpreted, or otherwise processed, and stored in a computer memory in a form that can be later executed by a computer. In this context, the computer-readable medium can be paper or another suitable medium that can act to transfer a program, for example, from the Internet, over a wired or wireless network, over the airwaves, etc., in a cartridged, downloadable, or other format. Thus, a computer-readable medium can take many forms of hardware and software.

[0060] It should be understood that aspects of the application can be implemented in hardware, software, firmware, or combinations thereof. In the embodiments described above, various steps or methods can be implemented, for example, in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be used: discrete logic circuitry having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having appropriate combinational logic gates, programmable gate arrays (PGA), field programmable gate arrays (FPGA), and the like.

[0061] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the particular feature, structure, material, or characteristic being described is included in at least one embodiment or example of the application. The illustrative descriptions of the above terms in this specification are not necessarily all referring to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0062] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A vehicle component assembly method, characterized by, The method comprises: According to the target model of the vehicle parts, a corresponding digital twin is created in real time, and a preset multi-dimensional perception array is enabled; Before the vehicle parts enter the assembly station, the target part features corresponding to the vehicle parts are collected in real time by the preset multi-dimensional perception array; The reference parameters corresponding to the digital twin are collected in real time, and the target part features are compared with the reference parameters based on a preset rule in multiple dimensions to predict whether an assembly error will occur in the assembly process in real time; If it is predicted in real time that an assembly error will occur in the assembly process, a corresponding target compensation parameter is generated in real time by an adaptive genetic algorithm to complete the real-time assembly of the vehicle parts.

2. The vehicle part assembly method according to claim 1, characterized by The step of collecting the target part features corresponding to the vehicle parts in real time by the preset multi-dimensional perception array comprises: When the vehicle parts enter the sensing range of the preset multi-dimensional perception array, the outer contour vertices of the vehicle parts are marked with coordinates by a preset laser positioning sensor to form a corresponding initial positioning coordinate set; The internal structure of the vehicle parts is tomographically scanned by a preset scanner, and the surface texture, aperture size and chamfer angle of the vehicle parts are imaged by a preset camera to generate a corresponding tomographic image set and an external morphology image set; The initial positioning coordinate set, the tomographic image set and the external morphology image set are feature processed to correspondingly generate the target part features.

3. The vehicle part assembly method according to claim 2, characterized by The step of feature processing the initial positioning coordinate set, the tomographic image set and the external morphology image set to correspondingly generate the target part features comprises: The initial positioning coordinate set is spatially topologically reconstructed to convert the generated discrete coordinate points into a continuous geometric grid, and the normal vector and curvature value of the grid vertices of the continuous geometric grid are extracted to form a corresponding three-dimensional structure feature parameter; The tomographic image set is subjected to multi-scale edge detection, different levels of tomographic profiles are obtained by Gaussian pyramid decomposition, and the profile coincidence degree and offset between adjacent tomographic profiles are calculated to generate internal structure feature parameters; The external morphology image set is subjected to pixel feature extraction, the direction angle and contrast of the surface texture are calculated by a gray level co-occurrence matrix, and the inclination angle is obtained by least squares fitting of the chamfer edge line to form appearance morphology feature parameters, and the three-dimensional structure feature parameters, the internal structure feature parameters and the appearance morphology feature parameters are normalized to generate the target part features.

4. The vehicle part assembly method according to claim 1, characterized by The step of comparing the target part features with the reference parameters based on a preset rule in multiple dimensions to predict whether an assembly error will occur in the assembly process in real time comprises: The target part features are reconstructed into a three-dimensional point cloud to extract a three-dimensional feature set containing geometric profiles, surface roughness and key hole coordinates, and the three-dimensional feature set is converted into a standard data matrix; According to the reference parameter, a reference matrix adapting to the three-dimensional feature set is generated in real time, and the standard data matrix and the reference matrix are dimensionally aligned through principal component analysis to generate a corresponding feature mapping relationship atlas; Based on the feature mapping relationship atlas, the cosine of the angle and the Euclidean distance of the feature vectors are calculated, and when the cosine of the angle or the Euclidean distance exceeds a preset threshold, it is determined that there is a risk of assembly error, and an error warning coordinate area is output.

5. The vehicle part assembly method according to claim 4, characterized in that, The step of calculating the cosine of the angle and the Euclidean distance of the feature vectors based on the feature mapping relationship atlas comprises: From the feature mapping relationship atlas, a target feature vector group corresponding to the three-dimensional feature set and a reference feature vector group corresponding to the reference matrix are extracted; According to the corresponding relationship established in the feature mapping relationship atlas, the target feature vector and the reference feature vector are paired and processed; For each pair of paired vectors, the cosine of the angle between them is calculated through the vector dot product formula, and the Euclidean distance between them is calculated through the space rectangular coordinate system for corresponding judgment.

6. The vehicle part assembly method according to claim 1, characterized by The step of generating the corresponding target compensation parameter in real time through the adaptive genetic algorithm to correspondingly complete the real-time assembly of the vehicle parts comprises: Based on the predicted assembly error, an initial gene population containing three-dimensional coordinate compensation, angle correction value and pressure adjustment coefficient is constructed; The initial gene population is subjected to adaptive selection, the selection probability is adjusted according to the error correction effect, and the effective compensation features of different gene fragments are recombined through the crossover operator; The mutation probability is dynamically adjusted according to the real-time working condition of the assembly station, and the converged gene fragment is output as the target compensation parameter after iteration.

7. The vehicle part assembly method according to claim 6, characterized by The step of dynamically adjusting the mutation probability according to the real-time working condition of the assembly station and outputting the converged gene fragment as the target compensation parameter after iteration comprises: Real-time collection of the environmental temperature and humidity, mechanical arm vibration frequency and part surface temperature of the assembly station to construct a working condition feature matrix; The working condition feature matrix is input into a preset model to generate a reference mutation probability coefficient and correct it combined with the number of iterations; After each iteration, the error simulation value is calculated, and when the error fluctuation of three consecutive iterations is less than a preset fluctuation threshold, the current gene fragment is determined as the target compensation parameter.

8. A vehicle component assembly system characterized by, The system comprises: A creation module is configured to create a corresponding digital twin in real time according to a target model of a vehicle part, and enable a preset multi-dimensional perception array; A collection module is configured to collect target part features corresponding to the vehicle part in real time through the preset multi-dimensional perception array before the vehicle part enters an assembly station; A mapping module is configured to collect reference parameters corresponding to the digital twin in real time, and perform multi-dimensional mapping comparison between the target part features and the reference parameters based on a preset rule to predict whether an assembly error will occur during assembly in real time; A prediction module is configured to generate a corresponding target compensation parameter in real time through an adaptive genetic algorithm to correspondingly complete the real-time assembly of the vehicle part if it is predicted that an assembly error will occur during assembly.

9. A computer comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor, when executing the computer program, realizes the vehicle part assembly method as claimed in any one of claims 1 to 7.

10. A readable storage medium, having stored thereon a computer program, characterized in that, The program, when executed by the processor, realizes the vehicle part assembly method as claimed in any one of claims 1 to 7.

Images