Tracing diagnosis method and system for size error of welded frame

Abstract

The invention relates to the technical field of welded vehicle frame manufacturing and detection, in particular to a traceability diagnosis method and system for welded vehicle frame size errors, and the method comprises the steps: obtaining actual measurement three-dimensional coordinates of a series of preset measurement target spots on a welded vehicle frame; wherein the measurement target spot comprises a functional reference point group for establishing a coordinate reference and a diagnosis feature point group for deformation analysis; on the basis of the actually measured coordinates of the function reference point group, calculating overall space transformation parameters between the frame design model and the actually measured frame to eliminate overall pose deviation; on the basis, the actually-measured coordinates of the diagnosis feature point group are compensated, residual deviation reflecting welding deformation is obtained, and a welding deformation mode is extracted through geometric analysis and mode recognition; and further matching and reasoning the overall space transformation parameters and the welding deformation mode with a process error source mapping knowledge base, and outputting a suspected process error source and a corresponding process optimization suggestion, so that accurate analysis and effective traceability of the size error of the welded frame are realized.

Description

Technical Field

[0001] This invention relates to the field of welding frame manufacturing and inspection technology, and in particular to a method and system for tracing and diagnosing dimensional errors in welding frames. Background Technology

[0002] As a core load-bearing component of two-wheeled electric vehicles and other products, the welded frame is prone to deviations from design values ​​due to thermal deformation and constraint deformation during the welding process, resulting in dimensional errors. The geometric quality of the welded frame directly affects the assembly accuracy and safety of electric two-wheeled vehicles. The dimensional errors of the welded frame not only include structural deformation caused by welding, but may also be compounded by overall posture deviations introduced by factors such as measurement placement posture and benchmark establishment methods. This causes different error sources in the measurement results to be coupled with each other, increasing the difficulty of subsequent analysis and diagnosis.

[0003] Currently, the industry generally adopts two types of technologies for dimensional control of welded vehicle frames: one is to use dedicated physical inspection fixtures designed for specific vehicle models to quickly and qualitatively judge key positions through simulated assembly. However, this method has poor versatility and is difficult to provide quantitative information reflecting the welding deformation mechanism, thus failing to provide clear guidance for process optimization. The other method is to use a coordinate measuring machine to acquire high-precision point cloud data and use software to align and compare it with the CAD model to generate an inspection report. However, this type of method usually does not effectively distinguish between overall pose deviation and welding deformation, resulting in a mixture of various error factors in the deviation data. The analysis results are highly dependent on the engineer's experience and judgment, making it difficult to systematically identify representative deformation features from the measurement data and stably correlate them with welding process parameters. As a result, dimensional control is mostly limited to the post-inspection stage, making it difficult to achieve precise traceability and feedforward control of the welding process.

[0004] The information disclosed in this background section is intended only to enhance the understanding of the general background of the invention and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0005] This invention provides a method and system for tracing and diagnosing dimensional errors in welded vehicle frames, thereby effectively solving the problems in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: a method for tracing and diagnosing dimensional errors in welded vehicle frames, comprising the following steps: The measured three-dimensional coordinates of a series of preset measurement target points on the welding frame are obtained. The preset measurement target points include a group of functional reference points for establishing coordinate references and a group of diagnostic feature points for deformation analysis. Based on the measured coordinates of the functional reference point group, the overall spatial transformation parameters required to align the chassis design model with the measured chassis are calculated using a spatial registration algorithm. Based on the overall spatial transformation parameters, the measured coordinates of each diagnostic feature point in the diagnostic feature point group are compensated to obtain the residual deviation reflecting the welding deformation. The residual deviation is then subjected to geometric analysis and pattern recognition to extract at least one predefined welding deformation pattern from the pre-constructed deformation pattern library. The types and values ​​of the overall spatial transformation parameters, as well as the identified deformation patterns, are matched and reasoned with a pre-built knowledge base of process error sources to output one or more suspected process error sources and associated process optimization suggestions.

[0007] Furthermore, the process error source mapping knowledge base is constructed in the following manner: Collect historical production data, including chassis design parameters, welding process parameters, measured coordinate data, confirmed sources of process errors, and records of rectification effects; Based on the historical production data, a mapping rule from measurement features to process error sources is established using data mining methods; The analysis results of the physical mechanism model of welding deformation, which is constructed based on the principles of welding thermodynamics and materials mechanics, are incorporated into the mapping rules as prior knowledge.

[0008] Furthermore, the physical mechanism model of welding deformation is a reduced-order proxy model established by the finite element method, and its construction includes: Fully parametric finite element simulation of the welding process is performed to generate a large-scale sample dataset covering the process parameter space; The dominant deformation modes were extracted from the sample dataset using the intrinsic orthogonal decomposition method. By establishing a rapid mapping relationship from key process parameters to dominant modal coefficients through Kriging interpolation or neural networks, a reduced-order surrogate model is formed.

[0009] Furthermore, the matching and reasoning also include: Calculate the overall confidence level for each output suspected process error source; The overall confidence level is calculated based on at least two of the following factors: the consistency of results from different inference paths, the statistical frequency of the error sources in historical data, and the degree of matching between the error sources and the current chassis feature combination.

[0010] Furthermore, the geometric analysis and pattern recognition of the residual deviations include: The diagnostic feature points are grouped according to their structural location on the vehicle frame and the welding process unit to which they belong. Statistical analysis is performed on the residual deviations within each group to extract at least one of the following characteristics: spatial distribution, directional consistency, and magnitude trend of the residual deviations. The extracted features are matched with typical welding deformation patterns in the deformation pattern library to determine the deformation pattern corresponding to the current frame welding deformation state.

[0011] Furthermore, statistical analysis of residual bias within the groups includes: For a group of points distributed on a tubular structure, the tendency of bending or twisting is calculated by curve or surface fitting; For a group of points distributed in the weld area, analyze the deviation vector of symmetrical points on both sides of the weld to calculate the asymmetric shrinkage.

[0012] Furthermore, the matching is achieved through a Siamese neural network, specifically including: The extracted feature vectors are input into one branch of the Siamese neural network, and the features of various typical patterns stored in the deformable pattern library are input into another branch. By calculating the similarity in the feature space, the pattern with the highest similarity to the current feature is identified as the welding deformation pattern, and the similarity value is output as the pattern confidence.

[0013] Furthermore, the spatial registration algorithm is a weighted robust registration algorithm, comprising: Assign a weight to each point in the functional reference point group; With the goal of minimizing the weighted sum of squared residuals, we solve for the optimal transformation parameters to align the design model with the measured chassis. The weights are dynamically allocated based at least on the functional importance of the corresponding benchmark point or historical measurement stability information.

[0014] Furthermore, the location of the preset measurement target point is determined in the following way: Based on the digital design model and welding process documents of the welded frame, the welding joints and assembly reference features are automatically identified. For each weld joint, the corresponding diagnostic target point layout template is invoked according to the type to generate the diagnostic feature point group; The functional reference point group is generated for each assembly reference feature.

[0015] The present invention also includes a traceability and diagnostic system for dimensional errors in welded vehicle frames, the system comprising: The coordinate measurement and acquisition module is used to acquire the measured three-dimensional coordinates of a series of preset measurement target points on the welding frame. The preset measurement target points include a group of functional reference points for establishing coordinate references and a group of diagnostic feature points for deformation analysis. The spatial registration module is used to calculate the overall spatial transformation parameters required to align the chassis design model with the actual chassis based on the measured coordinates of the functional reference point group and through a spatial registration algorithm. The deformation pattern recognition module is used to compensate the measured coordinates of each diagnostic feature point in the diagnostic feature point group based on the overall spatial transformation parameters, obtain the residual deviation reflecting the welding deformation, and perform geometric analysis and pattern recognition on the residual deviation to extract at least one predefined welding deformation pattern from the pre-built deformation pattern library. The error tracing module is used to match and reason with the type and value of the overall spatial transformation parameters and the identified deformation patterns as geometric behavior features of the welding process, and output one or more suspected process error sources and associated process optimization suggestions.

[0016] The beneficial effects of this invention are as follows: By performing structured analysis and layered processing on the measurement data of the welded frame, the overall pose deviation and welding deformation error can be effectively distinguished, avoiding the mutual coupling of different error sources in the measurement results. This improves the accuracy and stability of dimensional deviation analysis. On this basis, the residual deviation after removing the influence of the overall pose is further transformed into welding deformation features with engineering significance. Through deformation pattern recognition and process error source mapping, an effective correlation between measurement results and welding process factors is achieved. This allows dimensional deviations to no longer be judged only after the fact, but can be used to identify potential process problems and their causes. Thus, this invention not only reduces the dependence of dimensional analysis on the personal experience of engineers, but also provides a clear basis for optimizing welding process parameters, adjusting clamping schemes, and improving welding sequence. This is conducive to building a closed-loop control mechanism from measurement detection to process improvement, thereby improving the manufacturing consistency and overall quality level of the welded frame.

[0017] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1A flowchart for the method of tracing and diagnosing dimensional errors in welded vehicle frames; Figure 2 A flowchart illustrating the process of building a knowledge base for mapping process error sources; Figure 3 This is a flowchart illustrating the geometric analysis and pattern recognition of residual deviations. Figure 4 This is a schematic diagram of a system for tracing and diagnosing dimensional errors in welded vehicle frames. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] 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 invention pertains. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] Example 1: like Figures 1 to 3 As shown, this application provides a method for tracing and diagnosing dimensional errors in welded vehicle frames, the method comprising: S10: Obtain the measured three-dimensional coordinates of a series of preset measurement target points on the welding frame. The preset measurement target points include a group of functional reference points for establishing coordinate references and a group of diagnostic feature points for deformation analysis. S20: Based on the measured coordinates of the functional reference point group, the overall spatial transformation parameters required to align the frame design model with the measured frame are calculated through a spatial registration algorithm to eliminate the spatial deviation introduced by the overall positioning and clamping of the frame. S30: Based on the overall spatial transformation parameters, the measured coordinates of each diagnostic feature point in the diagnostic feature point group are compensated to obtain the residual deviation reflecting the welding deformation. The residual deviation is then subjected to geometric analysis and pattern recognition to extract at least one predefined welding deformation mode from the pre-built deformation mode library. S40: Match and reason with the type and value of the overall spatial transformation parameters and the identified deformation mode with the pre-built process error source mapping knowledge base, and output one or more suspected process error sources and related process optimization suggestions.

[0023] Specifically, firstly, a series of measurement target points are pre-set on the welded frame, and the measured three-dimensional coordinates of each target point are obtained using measuring equipment such as a coordinate measuring machine, laser tracker, or industrial photogrammetry system. The measurement target points include at least two categories: one is a group of functional reference points, used to reflect the positioning reference of the frame in the vehicle assembly, which is usually arranged at assembly reference holes, reference surfaces, or key mounting interface positions; the other is a group of diagnostic feature points, used to reflect structural deformations generated during welding, which is usually arranged near welds, areas of abrupt changes in structural stiffness, or key positions prone to thermal deformation. By functionally classifying the target points during the measurement stage, subsequent analysis can address different error sources separately. The acquisition of the functional reference point group... After obtaining the measured three-dimensional coordinates, they are matched with the corresponding nominal coordinates in the chassis design model. A spatial registration algorithm is used to calculate the overall spatial transformation parameters required to align the measured chassis to the design model. These parameters include at least spatial translation and rotation, used to characterize spatial deviations introduced by differences in the overall chassis placement posture, clamping method, or measurement coordinate system. By applying these overall spatial transformation parameters to the measured data, deviations caused by overall positioning or clamping factors are uniformly absorbed, thus avoiding misjudging overall pose errors as welding deformation in subsequent analysis. This step achieves preliminary decoupling of different error sources and is a key prerequisite for the present invention to achieve source tracing and diagnosis. After obtaining the overall spatial transformation parameters, they are... The measured three-dimensional coordinates of each diagnostic feature point in the diagnostic feature point group are applied, and the coordinates are compensated. The compensated coordinates are then compared with the nominal coordinates in the design model to obtain the residual deviation of each diagnostic feature point. The residual deviation reflects the actual welding deformation caused by factors such as welding heat input and structural constraints after removing the influence of the overall pose. In this embodiment, the residual deviation is further subjected to geometric analysis and pattern recognition processing. Specifically, the spatial distribution characteristics, directional consistency, and magnitude change trend of the residual deviation can be analyzed, and the analysis results are matched with a pre-built deformation pattern library. The deformation pattern library stores a variety of typical welding deformation patterns, each of which is associated with its corresponding geometric... The system describes the deformation through feature combinations and uses matching analysis to identify at least one predefined welding deformation mode corresponding to the current frame welding deformation state. This transforms discrete measurement deviations into deformation type descriptions with engineering significance. Finally, the type and value of the overall spatial transformation parameters and the identified welding deformation mode are used as joint inputs and matched and reasoned with a pre-built process error source mapping knowledge base. The process error source mapping knowledge base stores the correlation between measurement features, deformation modes, and welding process factors. By comprehensively reasoning about the above information, one or more suspected process error sources are output, and process optimization suggestions corresponding to the process error sources are further given, thus providing a direct basis for welding process improvement.

[0024] By performing structured analysis and layered processing on the measurement data of the welded frame, the overall pose deviation and welding deformation error can be effectively distinguished, avoiding the mutual coupling of different error sources in the measurement results. This improves the accuracy and stability of dimensional deviation analysis. On this basis, the residual deviation after removing the influence of the overall pose is further transformed into welding deformation features with engineering significance. Through deformation pattern recognition and process error source mapping, an effective correlation between measurement results and welding process factors is achieved. This allows dimensional deviations to no longer be judged only after the fact, but can be used to identify potential process problems and their causes. Thus, this invention not only reduces the dependence of dimensional analysis on the personal experience of engineers, but also provides a clear basis for optimizing welding process parameters, adjusting clamping schemes, and improving welding sequence. This is conducive to building a closed-loop control mechanism from measurement detection to process improvement, thereby improving the manufacturing consistency and overall quality level of the welded frame.

[0025] As a preferred embodiment of the above, in step S40, as Figure 2 As shown, the process error source mapping knowledge base is constructed in the following way: S41: Collect historical production data, including chassis design parameters, welding process parameters, measured coordinate data, confirmed sources of process errors, and records of rectification effects; S42: Based on historical production data, establish mapping rules from measurement features to process error sources through data mining methods; S43: The analysis results of the physical mechanism model of welding deformation, which is constructed based on the principles of welding thermodynamics and materials mechanics, are incorporated into the mapping rules as prior knowledge.

[0026] Specifically, the process begins with systematically collecting historical production data for welded chassis. This data includes at least chassis design parameters, welding process parameters, measured 3D coordinate data after welding, confirmed sources of process errors, and corresponding rectification records. This data is then standardized and structured. After organizing the historical production data, data mining analysis is performed to establish mapping rules between measurement features and process error sources. Measurement features include not only single-point or single-dimensional deviations but also spatial distribution characteristics, directional characteristics, and regional correlation characteristics formed by residual deviations from multiple measurement points. By analyzing the correspondence between measurement features and confirmed process error sources in different historical samples, discrimination rules that characterize specific process anomalies are extracted. These mapping rules can be stored in a knowledge base in rule form, feature weight form, or probability form, enabling them to be invoked during subsequent source tracing and diagnosis. To improve the mapping rules... To ensure reliability and engineering interpretability, this embodiment incorporates the analysis results of a welding deformation physical mechanism model built based on the principles of welding thermodynamics and materials mechanics as prior knowledge into the mapping rules. Specifically, through thermo-mechanical coupling analysis or empirical mechanism models, the deformation trends and typical deformation modes that may occur under different welding parameters, structural stiffness distributions, and constraints can be obtained. The analysis results are then used to verify or correct the mapping relationships obtained based on data mining. By introducing prior physical mechanisms, the constructed process error source mapping knowledge base possesses both data-driven and mechanism-driven characteristics. Through the above implementation method, while maintaining the interpretability of the welding deformation physical mechanism, rapid response modeling of complex welding processes is achieved. This transforms welding deformation analysis from computationally expensive finite element simulation into an efficient proxy model that can be embedded in the traceability and diagnostic process, providing a reliable physical foundation for process error source mapping and welding process control.

[0027] In this embodiment, in step S43, the physical mechanism model of welding deformation is a reduced-order proxy model established by the finite element method, the construction of which includes: Fully parametric finite element simulation of the welding process is performed to generate a large-scale sample dataset covering the process parameter space; The dominant deformation modes are extracted from the sample dataset using the intrinsic orthogonal decomposition method. By establishing a rapid mapping relationship from key process parameters to dominant modal coefficients through Kriging interpolation or neural networks, a reduced-order surrogate model is formed.

[0028] Specifically, firstly, a fully parametric finite element simulation model is performed on the welding process of the welded frame. Based on the frame's structural form, material properties, and welding process characteristics, a thermo-mechanical coupled finite element model is established, including weld geometry, heat source model, material thermo-mechanical properties, and clamping constraints. Welding current, voltage, welding speed, welding sequence, heat input intensity, and clamping constraint parameters are parameterized as variable process parameters. By performing multiple sets of combined simulations within the preset process parameter space, a large-scale sample dataset covering welding deformation responses under different process conditions is obtained. The sample data includes the displacement field, deformation field, or feature point displacement vector of key nodes or regions after welding. Subsequently, the sample dataset is dimensionality reduced to extract the dominant features of welding deformation. In this embodiment, the sample data is analyzed using the intrinsic orthogonal decomposition method, decomposing the high-dimensional deformation response into a set of physically meaningful orthogonal modes and their corresponding modal coefficients. The modes reflect the main deformation forms during the welding process. By retaining only the modes that can be used for deformation reduction, the model is further refined. This paper explains several dominant modes of the main deformation energy, achieving an effective reduced-order representation of the finite element results. After extracting the dominant deformation modes, a rapid mapping relationship between process parameters and dominant mode coefficients is further established. In this embodiment, the sample data is fitted by interpolation or learning methods to form a mapping model from key welding process parameters to dominant mode coefficients. This allows for rapid prediction of welding deformation trends under different process conditions without re-executing high-cost finite element simulations. The reduced-order surrogate model constructed in this way can achieve efficient computation while maintaining the consistency of the physical mechanism of welding deformation. It is convenient to use as a priori mechanism model in the construction of the process error source mapping knowledge base and the traceability diagnosis process. Through the above implementation method, while maintaining the interpretability of the physical mechanism of welding deformation, rapid response modeling of complex welding processes is achieved. This transforms welding deformation analysis from high-computation-cost finite element simulation into an efficient surrogate model that can be embedded in the traceability diagnosis process, providing a reliable physical basis for process error source mapping and welding process control.

[0029] In step S40, the matching and reasoning process further includes: Calculate the overall confidence level for each output suspected process error source; The overall confidence level is calculated based on at least two of the following factors: the consistency of results from different inference paths, the statistical frequency of error sources in historical data, and the degree of matching between error sources and the current chassis feature combination.

[0030] Specifically, in the calculation of the overall confidence level, diagnostic results from different inference paths are introduced. These inference paths can include rule-based inference based on measurement residual deviation characteristics, statistical inference based on historical sample similarity, and mechanism consistency inference based on welding physical mechanism models. When the same process error source is pointed to by multiple inference paths simultaneously, its corresponding consistency evaluation value increases accordingly, thereby enhancing the overall confidence level of that error source. Simultaneously, the statistical frequency of the process error source is quantified based on its occurrence in historical production data. If a process error source is repeatedly identified under the same or similar vehicle models, structural forms, or process conditions, its historical support strength is relatively high, and it occupies a higher weight in the overall confidence level calculation. Conversely, for error sources that are less frequent in history but are identified in the current case, their statistical support strength is relatively low. Furthermore, an evaluation of the matching degree between the process error source and the current chassis feature combination is also introduced. This method is used to measure the rationality of the error source under the current specific working conditions. The frame feature combination can include frame structure type, welding area distribution, spatial pattern of residual deviation of diagnostic feature points, and corresponding welding process parameter configuration, etc. By comparing the current feature combination with the applicable conditions of the process error source in historical samples or mechanism models, the corresponding matching degree evaluation result is obtained. Based on the above multiple factors, the comprehensive confidence of each suspected process error source is obtained through weighted fusion or scoring normalization, and multiple suspected process error sources are sorted or screened accordingly. Through the above implementation method, while outputting multiple suspected process error sources, their credibility can be quantitatively distinguished, making the source tracing diagnosis results more transparent, stable and interpretable, avoiding misjudgment caused by a single reasoning path, and helping engineers to quickly locate the most likely process problem based on the confidence level, thereby improving the guiding value of welding frame size error source tracing diagnosis in actual production.

[0031] As a preferred embodiment of the above, in step S30, such as Figure 3 As shown, geometric analysis and pattern recognition of residual deviations include: S31: Group the diagnostic feature points according to their structural location on the frame and the welding process unit to which they belong; S32: Perform statistical analysis on the residual deviations within each group to extract at least one of the following characteristics: spatial distribution of residual deviations, directional consistency, and trend of magnitude change. S33: Match the extracted features with typical welding deformation patterns in the deformation pattern library to determine the deformation pattern corresponding to the current frame welding deformation state.

[0032] Specifically, the diagnostic feature points are first grouped according to their spatial structural location on the vehicle frame and their corresponding welding process units. Structural locations can include the main beam area, sub-beam area, connection node area, or local reinforcement structure area of ​​the vehicle frame. Welding process units can include structural units formed within the same weld, the same welding station, or the same welding sequence. This grouping method ensures consistency in structural function and welding origin among the diagnostic feature points within each group, thereby reducing the possibility of interference between deformation characteristics of different welding areas. After grouping, statistical analysis and geometric feature extraction are performed on the residual deviations within each group. Specifically, spatial distribution analysis can be performed on the residual deviations of the diagnostic feature points within the same group to determine their concentration or diffusion in three-dimensional space. The process involves several steps: First, a consistency analysis is performed on the direction of the residual deviation vector to assess whether a unified deformation direction exists within the group. Second, the trend of residual deviation magnitude is analyzed to identify whether deformation features gradually increase or decrease along the structural length or weld direction. The results of these analyses can be used individually or in combination to constitute the geometric feature description corresponding to the group. After obtaining the geometric features of each group, these features are matched with typical welding deformation modes in a pre-built deformation mode library. This library stores various welding deformation mode descriptions with engineering significance. By comparing the similarity between the geometric features of the current group and the features of typical deformation modes, the deformation mode corresponding to the current frame welding deformation state is determined, thereby achieving qualitative identification and structured expression of the welding deformation state.

[0033] In this embodiment, step S32, statistical analysis of the residual deviation within the group, includes: For a group of points distributed on a tubular structure, the tendency of bending or twisting is calculated by curve or surface fitting; For a group of points distributed in the weld area, the deviation vector of symmetrical points on both sides of the weld is analyzed to calculate the asymmetric shrinkage. Typical welding deformation modes include at least one of the following: overall bending mode, local welding shrinkage mode, torsional deformation mode, flatness deviation mode, or assembly hole offset mode.

[0034] Specifically, for point groups distributed on tubular structures, these point groups are typically distributed along the axial or circumferential direction of the frame tubing. Their welding deformation often manifests as overall bending or twisting. To address this characteristic, the residual deviation of these point groups is analyzed using curve or surface fitting. Specifically, the centerline of the point group in the axial direction of the tubing can be used as a reference. The deformed spatial curve is obtained through least-squares fitting and compared with the nominal geometric centerline to calculate the bending trend of the tubing at different locations. Alternatively, based on the circumferential distribution characteristics of the point group, the surface morphology of the tubing's outer surface is fitted. The degree of twisting of the tubing is quantified by analyzing the change in the torsion angle of the fitted surface. The resulting bending or twisting trend parameters serve as important statistical indicators describing the welding deformation characteristics of this group. For points distributed in the weld area... These point groups are typically located in the base material regions on both sides of the weld. Their welding deformation is mainly caused by uneven weld shrinkage and tends to be asymmetric. To address this characteristic, this embodiment selects diagnostic feature points located on both sides of the weld and having a symmetrical relationship in the nominal model as the analysis objects. By comparing the residual deviation vectors of corresponding point pairs, the difference in deviation between the two sides of the weld in the normal or specified direction is calculated, thereby obtaining the asymmetric shrinkage amount. This asymmetric shrinkage amount can intuitively reflect the weld deformation characteristics caused by uneven heat input or inconsistent constraint conditions during the welding process. By extracting statistical features such as bending, torsion, or asymmetric shrinkage for different structural types, the residual deviation analysis results within the group correspond to the specific welding deformation mechanism, providing feature inputs with clear physical meaning for subsequent deformation pattern recognition.

[0035] In step S33, the matching is achieved through a Siamese neural network, specifically including: The extracted feature vectors are input into one branch of the Siamese neural network, and the features of various typical patterns stored in the deformable pattern library are input into another branch. By calculating the similarity in the feature space, the pattern with the highest similarity to the current feature is identified as the welding deformation pattern, and the similarity value is output as the pattern confidence.

[0036] Specifically, the two branches of the Siamese neural network use the same network structure and parameter configuration to map the input feature vectors into a unified feature space. This makes features with the same or similar welding deformation mechanisms closer together in the feature space, while features corresponding to different deformation modes are farther apart. After feature mapping, the similarity between the feature vector to be identified and the feature vectors of each typical deformation mode is calculated in the feature space. Different deformation modes are compared and analyzed. The similarity can be obtained based on Euclidean distance, cosine similarity, or other distance metrics, and is numerically reflected to show the degree of closeness between the current welding deformation state and each typical deformation mode. Finally, the typical deformation mode with the highest similarity is determined as the welding deformation mode corresponding to the current frame welding deformation state. At the same time, the corresponding similarity value is output as the mode confidence of the pattern recognition result, which is used to characterize the reliability of the recognition result. In this way, the recognition of welding deformation modes no longer relies on manually set fixed criteria, but is based on the overall similarity in the feature space, thereby enhancing the adaptability to complex or mixed deformation states.

[0037] As a preferred embodiment of the above embodiments, in step S20, the spatial registration algorithm is a weighted robust registration algorithm, including: Assign a weight to each point in the functional benchmark group; With the goal of minimizing the weighted sum of squared residuals, we solve for the optimal transformation parameters to align the design model with the measured chassis. The weights are dynamically allocated based at least on the functional importance of the corresponding benchmark or historical measurement stability information.

[0038] Specifically, in one embodiment, when spatially aligning the design model and the actual vehicle frame based on a group of functional reference points, each functional reference point is no longer considered as an equally weighted point. Instead, a corresponding weight parameter is assigned to each reference point in the functional reference point group. The weight is used to reflect the reliability and constraint capability of different reference points in the overall spatial registration. When assigning weights, multiple factors related to engineering applications can be considered comprehensively. Among them, the functional importance factor is used to describe the criticality of a certain reference point in the vehicle assembly or functional positioning; the historical measurement stability factor is used to reflect the repeatability and dispersion of the reference point in historical measurement data. For reference points with smaller fluctuations in multiple measurement results, their corresponding weights are relatively higher, while the weights of reference points with larger fluctuations are relatively lower. Weight information can be determined before registration or dynamically updated after multiple measurement data accumulations. After weight allocation, the optimal spatial transformation parameters for aligning the design model to the coordinate system of the measured vehicle frame are solved with the objective of minimizing the weighted sum of squared residuals between the functional reference points in the design model and the measured vehicle frame. The spatial transformation parameters include at least three-dimensional translation and rotation parameters. By solving the registration problem in the sense of weighted least squares, the high-weight reference points exert stronger constraints on the overall registration results, thereby reducing the adverse effects of low-confidence reference points on the registration results. Through the above weighted robust registration method, stable and reliable overall spatial transformation parameters can still be obtained even in the presence of local measurement anomalies or slight deformation of reference points, providing a consistent coordinate basis for subsequent residual deviation analysis.

[0039] In this embodiment, in step S10, the position of the preset measurement target point is determined in the following way: Based on the digital design model and welding process documents of the welded frame, the welding joints and assembly reference features are automatically identified. For each welded joint, the corresponding diagnostic target point layout template is called according to the type to generate a group of diagnostic feature points; Generate a group of functional reference points for each assembly reference feature.

[0040] Specifically, before measuring the dimensions of the welded frame, the digital design model of the welded frame and the corresponding welding process documents are first read. The digital design model can be a 3D CAD model, and the welding process documents can include information such as weld numbers, weld joint types, welding sequence, and assembly datum definitions. Based on the above input data, the frame structure is analyzed, and the weld joints and assembly datum features in the frame are automatically identified. Weld joints can be identified based on the intersection relationship of components, weld geometric features, or process annotations. Assembly datum features can include positioning holes, datum surfaces, or assembly reference boundaries. After the weld joint identification is completed, a corresponding diagnostic target point layout template is called for each weld joint according to its structural form and process type. The diagnostic target point layout template predefines the number of measurement points, spatial distribution, and relative weld dimensions that are compatible with this type of weld joint. Positional relationships: For plate-to-plate fillet welds, measurement points reflecting angular deformation and shrinkage deformation can be arranged in key areas near the weld. By mapping the template to the geometric position of the specific weld joint, a group of diagnostic feature points for weld deformation analysis is automatically generated. At the same time, for the identified assembly datum features, corresponding functional datum point groups are automatically generated according to their role in vehicle assembly and functional positioning. Functional datum point groups are usually set at the center of the assembly datum hole, representative positions on the datum surface, or on key positioning structures to establish stable coordinate datums in subsequent analysis, thereby providing reliable constraints for overall spatial registration. In the above way, the diagnostic feature point group and the functional datum point group can be automatically planned in the design stage, so that the arrangement of measurement target points is consistent with the weld structure features and process units, providing a high-quality input data foundation for subsequent dimensional error analysis and traceability diagnosis.

[0041] Example 2: This invention also includes a traceability and diagnostic system for welded vehicle frame dimensional errors, such as... Figure 4 As shown, the system includes: The coordinate measurement and acquisition module is used to acquire the measured three-dimensional coordinates of a series of preset measurement target points on the welding frame. The preset measurement target points include a group of functional reference points for establishing coordinate references and a group of diagnostic feature points for deformation analysis. The spatial registration module is used to calculate the overall spatial transformation parameters required to align the chassis design model with the actual chassis based on the measured coordinates of the functional reference point group and through the spatial registration algorithm. The deformation pattern recognition module is used to compensate the measured coordinates of each diagnostic feature point in the diagnostic feature point group based on the overall spatial transformation parameters, obtain the residual deviation reflecting the welding deformation, and perform geometric analysis and pattern recognition on the residual deviation to extract at least one predefined welding deformation pattern from the pre-built deformation pattern library. The error tracing module is used to match and reason with the type and value of the overall spatial transformation parameters and the identified deformation patterns as geometric behavior features of the welding process, and output one or more suspected process error sources and associated process optimization suggestions.

[0042] The adjustment system described above in this invention can effectively realize the method for tracing and diagnosing the dimensional errors of welded vehicle frames. The technical effects it can achieve are as described in the above embodiments, and will not be repeated here.

[0043] Similarly, the above-mentioned optimization schemes for the system can also achieve the optimization effects corresponding to the methods in Embodiment 1, which will not be repeated here.

[0044] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and accompanying drawings are merely exemplary illustrations of the application as defined herein, and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Thus, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for tracing and diagnosing dimensional errors in welded vehicle frames, characterized in that, The method includes: The measured three-dimensional coordinates of a series of preset measurement target points on the welding frame are obtained. The preset measurement target points include a group of functional reference points for establishing coordinate references and a group of diagnostic feature points for deformation analysis. Based on the measured coordinates of the functional reference point group, the overall spatial transformation parameters required to align the chassis design model with the measured chassis are calculated using a spatial registration algorithm. Based on the overall spatial transformation parameters, the measured coordinates of each diagnostic feature point in the diagnostic feature point group are compensated to obtain the residual deviation reflecting the welding deformation. The residual deviation is then subjected to geometric analysis and pattern recognition to extract at least one predefined welding deformation pattern from the pre-constructed deformation pattern library. The types and values ​​of the overall spatial transformation parameters, as well as the identified deformation patterns, are matched and reasoned with a pre-built knowledge base of process error sources to output one or more suspected process error sources and associated process optimization suggestions.

2. The method for tracing and diagnosing dimensional errors in welded vehicle frames according to claim 1, characterized in that, The process error source mapping knowledge base is constructed in the following way: Collect historical production data, including chassis design parameters, welding process parameters, measured coordinate data, confirmed sources of process errors, and records of rectification effects; Based on the historical production data, a mapping rule from measurement features to process error sources is established using data mining methods; The analysis results of the physical mechanism model of welding deformation, which is constructed based on the principles of welding thermodynamics and materials mechanics, are incorporated into the mapping rules as prior knowledge.

3. The method for tracing and diagnosing dimensional errors in welded vehicle frames according to claim 2, characterized in that, The physical mechanism model of welding deformation is a reduced-order proxy model established by the finite element method, and its construction includes: Fully parametric finite element simulation of the welding process is performed to generate a large-scale sample dataset covering the process parameter space; The dominant deformation modes were extracted from the sample dataset using the intrinsic orthogonal decomposition method. By establishing a rapid mapping relationship from key process parameters to dominant modal coefficients through Kriging interpolation or neural networks, a reduced-order surrogate model is formed.

4. The method for tracing and diagnosing dimensional errors in welded vehicle frames according to claim 1, characterized in that, The matching and reasoning also include: Calculate the overall confidence level for each output suspected process error source; The overall confidence level is calculated based on at least two of the following factors: the consistency of results from different inference paths, the statistical frequency of the error sources in historical data, and the degree of matching between the error sources and the current chassis feature combination.

5. The method for tracing and diagnosing dimensional errors in welded vehicle frames according to claim 1, characterized in that, Geometric analysis and pattern recognition of the residual deviations include: The diagnostic feature points are grouped according to their structural location on the vehicle frame and the welding process unit to which they belong. Statistical analysis is performed on the residual deviations within each group to extract at least one of the following characteristics: spatial distribution, directional consistency, and magnitude trend of the residual deviations. The extracted features are matched with typical welding deformation patterns in the deformation pattern library to determine the deformation pattern corresponding to the current frame welding deformation state.

6. The method for tracing and diagnosing dimensional errors in welded vehicle frames according to claim 5, characterized in that, Statistical analysis of residual bias within groups includes: For a group of points distributed on a tubular structure, the tendency of bending or twisting is calculated by curve or surface fitting; For a group of points distributed in the weld area, analyze the deviation vector of symmetrical points on both sides of the weld to calculate the asymmetric shrinkage.

7. The method for tracing and diagnosing dimensional errors in welded vehicle frames according to claim 5, characterized in that, The matching is achieved through a Siamese neural network, specifically including: The extracted feature vectors are input into one branch of the Siamese neural network, and the features of various typical patterns stored in the deformable pattern library are input into another branch. By calculating the similarity in the feature space, the pattern with the highest similarity to the current feature is identified as the welding deformation pattern, and the similarity value is output as the pattern confidence.

8. The method for tracing and diagnosing dimensional errors in welded vehicle frames according to claim 1, characterized in that, The spatial registration algorithm is a weighted robust registration algorithm, including: Assign a weight to each point in the functional reference point group; With the goal of minimizing the weighted sum of squared residuals, we solve for the optimal transformation parameters to align the design model with the measured chassis. The weights are dynamically allocated based at least on the functional importance of the corresponding benchmark point or historical measurement stability information.

9. The method for tracing and diagnosing dimensional errors in welded vehicle frames according to claim 1, characterized in that, The location of the preset measurement target point is determined in the following way: Based on the digital design model and welding process documents of the welded frame, the welding joints and assembly reference features are automatically identified. For each weld joint, the corresponding diagnostic target point layout template is invoked according to the type to generate the diagnostic feature point group; The functional reference point group is generated for each assembly reference feature.

10. A system for tracing and diagnosing dimensional errors in welded vehicle frames, characterized in that, The system includes: The coordinate measurement and acquisition module is used to acquire the measured three-dimensional coordinates of a series of preset measurement target points on the welding frame. The preset measurement target points include a group of functional reference points for establishing coordinate references and a group of diagnostic feature points for deformation analysis. The spatial registration module is used to calculate the overall spatial transformation parameters required to align the chassis design model with the actual chassis based on the measured coordinates of the functional reference point group and through a spatial registration algorithm. The deformation pattern recognition module is used to compensate the measured coordinates of each diagnostic feature point in the diagnostic feature point group based on the overall spatial transformation parameters, obtain the residual deviation reflecting the welding deformation, and perform geometric analysis and pattern recognition on the residual deviation to extract at least one predefined welding deformation pattern from the pre-built deformation pattern library. The error tracing module is used to match and reason with the type and value of the overall spatial transformation parameters and the identified deformation patterns as geometric behavior features of the welding process, and output one or more suspected process error sources and associated process optimization suggestions.

Images