A commercial vehicle assembly inspection method and system

Abstract

The application provides a commercial vehicle assembly inspection method and system, which comprises the following steps: collecting a torque-angle curve, a bolt ultrasonic echo signal and a tool vibration frequency of bolt assembly through a preset distributed sensing unit based on an assembly station coordinate, inputting them into a preset feature separation algorithm, constructing a target mapping relationship between the torque-angle curve and the bolt ultrasonic echo signal through the preset feature separation algorithm, and generating a corresponding full-cycle torque feature vector; extracting a dynamic torque threshold interval suitable for the current bolt characteristics from a preset assembly parameter database, and judging whether the bolt is normally assembled according to the dynamic torque threshold interval and the full-cycle torque feature vector; if not, performing process backtracking, locating the torque deviation root cause by combining fault tree analysis; generating corresponding assembly parameter optimization instructions according to the torque deviation root cause, and feeding back to an execution system to complete the normal assembly of the bolt. The application can effectively improve the inspection efficiency of the commercial vehicle.

Description

Technical Field

[0001] This invention relates to the field of commercial vehicle manufacturing technology, and in particular to a method and system for assembling and inspecting commercial vehicles. Background Technology

[0002] The assembly quality of key bolts in a commercial vehicle chassis (such as suspension, steering, and braking system bolts) directly determines the safety and durability of the entire vehicle. These bolts need to withstand complex loads and vibration impacts, and failure in their tightened state can easily lead to serious accidents such as steering failure and braking malfunction. Therefore, accurate inspection of their assembly quality is a critical step in commercial vehicle production.

[0003] Currently, most commercial vehicle assembly lines employ visual inspection technology combining industrial cameras and image processing. This technology identifies bolt positions, angles, and appearance defects to determine if assembly is correct. While this technology replaces manual visual inspection, effectively improving the efficiency and accuracy of defect identification and providing a fundamental guarantee for assembly quality control, visual inspection only detects external features and cannot obtain core quality parameters such as bolt torque and preload, presenting significant limitations. The core of bolt assembly completion lies in whether the torque meets the standard; specifically, insufficient torque leads to loosening, while excessive torque leads to breakage—potential risks that cannot be identified through visual inspection.

[0004] Furthermore, existing visual inspection technologies have significant functional limitations. Their core deficiency lies in their ability to only inspect the appearance of bolts, failing to directly obtain core quality parameters such as tightening torque and preload. This results in a serious blind spot in assessing assembly quality. The completeness of bolt assembly is not only reflected in visual compliance but also depends on whether the torque meets the manufacturer's specific standards. Specifically, insufficient torque can easily lead to bolt loosening, while excessive torque may cause thread damage or bolt breakage—potential risks that cannot be identified through visual inspection. Simultaneously, commercial vehicle chassis bolts face technological challenges such as the "stick-slip effect." Even if the appearance assembly is satisfactory, insufficient preload may leave safety hazards. Existing inspection methods are insufficient to meet the high-precision assembly quality requirements of commercial vehicles. Summary of the Invention

[0005] Therefore, the purpose of this invention is to provide a method and system for assembling and inspecting commercial vehicles to solve the problems of the prior art.

[0006] The first aspect of the present invention proposes:

[0007] A commercial vehicle assembly inspection method, specifically including the following steps:

[0008] Based on the assembly station coordinates, the torque-rotation curve of bolt assembly, bolt ultrasonic echo signal and tool vibration frequency are collected by a preset distributed sensing unit. The data are then input into a preset feature separation algorithm, and the preset feature separation algorithm is used to construct the target mapping relationship between the torque-rotation curve and the bolt ultrasonic echo signal to generate the corresponding full-cycle torque feature vector.

[0009] Extract the dynamic torque threshold range that is compatible with the current bolt characteristics from the preset assembly parameter database, and determine whether the bolt is properly assembled based on the dynamic torque threshold range and the full-cycle torque feature vector.

[0010] If the bolt assembly is determined to be abnormal based on the dynamic torque threshold range and the full-cycle torque feature vector, then process backtracking is performed, and the root cause of the torque deviation is located by combining fault tree analysis.

[0011] Based on the root cause of the torque deviation, corresponding assembly parameter optimization instructions are generated and fed back to the execution system to complete the normal assembly of the bolts.

[0012] The beneficial effects of this invention are as follows: This commercial vehicle assembly inspection method accurately collects the torque-angle curve, ultrasonic echo signal, and tool vibration frequency of bolt assembly through a preset distributed sensing unit based on the assembly station coordinates. A preset feature separation algorithm is used to construct a target mapping relationship to generate a full-cycle torque feature vector. Combined with a preset database of dynamic torque threshold ranges adapted to the current bolt characteristics, this effectively solves the problem of inaccurate bolt torque feature collection in existing technologies, significantly improving the accuracy of assembly inspection to meet the high-precision assembly requirements of commercial vehicles. When an assembly abnormality is detected, process backtracking and fault tree analysis can quickly locate the root cause of torque deviation. The generated assembly parameter optimization instructions are fed back to the execution system, enabling efficient normal bolt assembly and further ensuring assembly quality and efficiency.

[0013] Furthermore, the step of constructing the target mapping relationship between the torque-rotation angle curve and the bolt ultrasonic echo signal through the preset feature separation algorithm to generate the corresponding full-cycle torque feature vector includes:

[0014] The timestamps of the torque-rotation curve and the acquisition time of the ultrasonic echo signal of the bolt are aligned by a dynamic time warping algorithm to construct a corresponding target mapping relationship, and the effective vibration features in the vibration frequency of the tool are extracted by a variational mode decomposition algorithm.

[0015] Based on the target mapping relationship, dynamic correlation coefficients between torque parameters and effective vibration characteristics under different assembly stages are established to construct the corresponding correlation coefficient matrix.

[0016] Based on the correlation coefficient matrix, the torque-angle curves within the entire assembly cycle are subjected to feature enhancement processing, and bolt material and specification parameters are introduced as constraints to generate the full-cycle torque feature vector.

[0017] Furthermore, the step of performing feature enhancement processing on the torque-angle curve throughout the entire assembly cycle based on the correlation coefficient matrix, and introducing bolt material and specification parameters as constraints to generate the full-cycle torque feature vector includes:

[0018] Based on the correlation coefficient matrix, density peak clustering is used to locate abnormal fluctuation segments in the torque-angle curve. An adaptive weighted algorithm is then used to correct outliers. The inverse of the correlation coefficient matrix is ​​used as the weight to dynamically gain the corrected torque-angle curve and output target feature data.

[0019] The core parameters of the bolt are extracted, and the constraint factors corresponding to the core parameters are screened out through grey relational analysis. The target feature data and the constraint factors are then fused across dimensions to generate a corresponding fused feature set.

[0020] The fused feature set is quantized to generate the full-cycle torque feature vector.

[0021] Furthermore, the step of determining whether the bolt is properly assembled based on the dynamic torque threshold range and the full-cycle torque feature vector includes:

[0022] A Mahalanobis distance algorithm is used to construct an adaptation network between the full-cycle torque feature vector and the dynamic torque threshold interval. Based on the adaptation network, the full-cycle torque feature vector is projected onto the two-dimensional evaluation plane corresponding to the dynamic torque threshold interval through feature dimension mapping, so as to output the minimum distance value between the feature point and the threshold boundary.

[0023] Combining the three stages of bolt assembly—preload, yield, and tightening—the full-cycle torque feature vector is decomposed into several stage sub-vectors, and the matching degree between each stage sub-vector and the standard stage feature template is calculated.

[0024] If the minimum distance value exceeds the preset distance threshold or the matching degree of any of the stage sub-vectors is lower than the preset matching degree threshold, it is marked as a preliminary anomaly. The preliminary anomaly result is then used to initiate multi-source data cross-validation to determine whether the bolt is properly assembled.

[0025] Furthermore, the step of initiating multi-source data cross-validation of the preliminary abnormal results to determine whether the bolts are properly assembled includes:

[0026] The target assembly stage corresponding to the preliminary abnormal results is detected, and a multi-source verification feature set is constructed based on the abnormality type of the target assembly stage;

[0027] The ultrasonic resonant frequency deviation, vibration phase abrupt change value and torque characteristic anomaly parameters in the multi-source verification feature set are extracted, and converted into corresponding fuzzy confidence scores through fuzzy inference. The conditional probability of the multi-source verification feature set is calculated using a Bayesian network.

[0028] The fuzzy confidence and the conditional probability are fused to obtain a multi-dimensional anomaly confidence, and the bolt assembly status is determined based on the magnitude of the multi-dimensional anomaly confidence.

[0029] Furthermore, the step of performing process backtracking and combining fault tree analysis to locate the root cause of torque deviation includes:

[0030] The entire process data chain of the assembly is traced through the preset distributed sensing unit, and the corresponding abnormal features contained in the entire process data chain are extracted.

[0031] Triggered by the aforementioned abnormal features, the assembly fault knowledge graph is invoked, and a corresponding fault tree is generated based on entity association rules;

[0032] Based on the fault tree, a corresponding fault propagation path is generated, and the fault propagation path is verified across domains to output the root cause of the torque deviation.

[0033] Furthermore, the step of performing cross-domain verification on the fault propagation path to output the root cause of the torque deviation includes:

[0034] Extract the feature values ​​corresponding to each underlying node in the fault propagation path, and select the core verification objects based on the magnitude of the feature values;

[0035] A verification matrix adapted to the core verification object is created based on the assembly fault knowledge graph, so as to simulate the corresponding target deviation data based on the verification matrix and the core verification object;

[0036] In the fault propagation path, a target bottom-level node that matches the target deviation data is selected, and the target bottom-level node is set as the root cause of the torque deviation.

[0037] The second aspect of the present invention proposes:

[0038] A commercial vehicle assembly inspection system, wherein the system includes:

[0039] The acquisition module is used to acquire the torque-rotation curve of bolt assembly, bolt ultrasonic echo signal and tool vibration frequency based on the assembly station coordinates through a preset distributed sensing unit, so as to input them into a preset feature separation algorithm, and construct the target mapping relationship between the torque-rotation curve and the bolt ultrasonic echo signal through the preset feature separation algorithm to generate the corresponding full-cycle torque feature vector.

[0040] The extraction module is used to extract the dynamic torque threshold range that is suitable for the current bolt characteristics from the preset assembly parameter database, and to determine whether the bolt is assembled normally based on the dynamic torque threshold range and the full-cycle torque feature vector.

[0041] The analysis module is used to perform process backtracking and locate the root cause of torque deviation by combining fault tree analysis if it is determined that the bolt assembly is abnormal based on the dynamic torque threshold range and the full-cycle torque feature vector.

[0042] The generation module is used to generate corresponding assembly parameter optimization instructions based on the root cause of the torque deviation, and feed them back to the execution system to complete the normal assembly of the bolts.

[0043] Furthermore, the acquisition module is specifically used for:

[0044] The timestamps of the torque-rotation curve and the acquisition time of the ultrasonic echo signal of the bolt are aligned by a dynamic time warping algorithm to construct a corresponding target mapping relationship, and the effective vibration features in the vibration frequency of the tool are extracted by a variational mode decomposition algorithm.

[0045] Based on the target mapping relationship, dynamic correlation coefficients between torque parameters and effective vibration characteristics under different assembly stages are established to construct the corresponding correlation coefficient matrix.

[0046] Based on the correlation coefficient matrix, the torque-angle curves within the entire assembly cycle are subjected to feature enhancement processing, and bolt material and specification parameters are introduced as constraints to generate the full-cycle torque feature vector.

[0047] Furthermore, the acquisition module is specifically used for:

[0048] Based on the correlation coefficient matrix, density peak clustering is used to locate abnormal fluctuation segments in the torque-angle curve. An adaptive weighted algorithm is then used to correct outliers. The inverse of the correlation coefficient matrix is ​​used as the weight to dynamically gain the corrected torque-angle curve and output target feature data.

[0049] The core parameters of the bolt are extracted, and the constraint factors corresponding to the core parameters are screened out through grey relational analysis. The target feature data and the constraint factors are then fused across dimensions to generate a corresponding fused feature set.

[0050] The fused feature set is quantized to generate the full-cycle torque feature vector.

[0051] Furthermore, the extraction module is specifically used for:

[0052] A Mahalanobis distance algorithm is used to construct an adaptation network between the full-cycle torque feature vector and the dynamic torque threshold interval. Based on the adaptation network, the full-cycle torque feature vector is projected onto the two-dimensional evaluation plane corresponding to the dynamic torque threshold interval through feature dimension mapping, so as to output the minimum distance value between the feature point and the threshold boundary.

[0053] Combining the three stages of bolt assembly—preload, yield, and tightening—the full-cycle torque feature vector is decomposed into several stage sub-vectors, and the matching degree between each stage sub-vector and the standard stage feature template is calculated.

[0054] If the minimum distance value exceeds the preset distance threshold or the matching degree of any of the stage sub-vectors is lower than the preset matching degree threshold, it is marked as a preliminary anomaly. The preliminary anomaly result is then used to initiate multi-source data cross-validation to determine whether the bolt is properly assembled.

[0055] Furthermore, the extraction module is specifically used for:

[0056] The target assembly stage corresponding to the preliminary abnormal results is detected, and a multi-source verification feature set is constructed based on the abnormality type of the target assembly stage;

[0057] The ultrasonic resonant frequency deviation, vibration phase abrupt change value and torque characteristic anomaly parameters in the multi-source verification feature set are extracted, and converted into corresponding fuzzy confidence scores through fuzzy inference. The conditional probability of the multi-source verification feature set is calculated using a Bayesian network.

[0058] The fuzzy confidence and the conditional probability are fused to obtain a multi-dimensional anomaly confidence, and the bolt assembly status is determined based on the magnitude of the multi-dimensional anomaly confidence.

[0059] Furthermore, the analysis module is specifically used for:

[0060] The entire process data chain of the assembly is traced through the preset distributed sensing unit, and the corresponding abnormal features contained in the entire process data chain are extracted.

[0061] Triggered by the aforementioned abnormal features, the assembly fault knowledge graph is invoked, and a corresponding fault tree is generated based on entity association rules;

[0062] Based on the fault tree, a corresponding fault propagation path is generated, and the fault propagation path is verified across domains to output the root cause of the torque deviation.

[0063] Furthermore, the analysis module is specifically used for:

[0064] Extract the feature values ​​corresponding to each underlying node in the fault propagation path, and select the core verification objects based on the magnitude of the feature values;

[0065] A verification matrix adapted to the core verification object is created based on the assembly fault knowledge graph, so as to simulate the corresponding target deviation data based on the verification matrix and the core verification object;

[0066] In the fault propagation path, a target bottom-level node that matches the target deviation data is selected, and the target bottom-level node is set as the root cause of the torque deviation.

[0067] The third aspect of the present invention proposes:

[0068] A computer includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the commercial vehicle assembly inspection method as described above.

[0069] The fourth aspect of the present invention proposes:

[0070] A readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the commercial vehicle assembly inspection method as described above.

[0071] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0072] Figure 1 A flowchart of the commercial vehicle assembly inspection method provided in the first embodiment of the present invention;

[0073] Figure 2 This is a structural block diagram of a commercial vehicle assembly and inspection system provided in the third embodiment of the present invention.

[0074] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0075] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0076] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0077] 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 herein in the description of the invention 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.

[0078] Please see Figure 1 The image shows a commercial vehicle assembly inspection method provided in the first embodiment of the present invention. The commercial vehicle assembly inspection method provided in this embodiment can objectively and effectively detect whether the bolts are properly assembled, thereby improving the assembly efficiency of commercial vehicles.

[0079] Specifically, this embodiment provides:

[0080] A commercial vehicle assembly inspection method, specifically including the following steps:

[0081] Step S10: Based on the assembly station coordinates, the torque-rotation angle curve of the bolt assembly, the ultrasonic echo signal of the bolt, and the vibration frequency of the tool are collected by a preset distributed sensing unit and input into a preset feature separation algorithm. The preset feature separation algorithm is used to construct the target mapping relationship between the torque-rotation angle curve and the ultrasonic echo signal of the bolt to generate the corresponding full-cycle torque feature vector.

[0082] Among them, it should be noted that the first step focuses on the comprehensiveness of data collection: based on the coordinates of the assembly station, distributed sensing units are deployed to synchronously collect three types of core data. Specifically, the torque-rotation angle curve (which directly reflects the matching relationship between the assembly force and the rotation of the bolt and is the core basis for the pre-tightening force to meet the standard), the ultrasonic echo signal of the bolt (which penetrates the inside of the bolt to detect hidden defects such as thread slipping and cracks), and the vibration frequency of the tool (which reflects the operating stability of the assembly tool, and abnormal vibration will cause fluctuations in torque output). The data is input into a preset feature separation algorithm, and the core purpose is to break the limitations of single data: by constructing the target mapping relationship between the torque-rotation angle curve and the ultrasonic echo signal of the bolt (such as "the torque is normal but the ultrasonic echo is abnormal, indicating that there are defects inside the bolt"), a full-cycle torque feature vector is generated. Specifically, this vector integrates three-dimensional information of "force-shape-quality" to provide a comprehensive basis for subsequent judgment. This is for the convenience of subsequent processing.

[0083] Step S20: Extract the dynamic torque threshold range adapted to the current bolt characteristics from the preset assembly parameter database, and determine whether the bolt is assembled normally according to the dynamic torque threshold range and the full-cycle torque feature vector.

[0084] Among them, it should be noted that the second step achieves precise judgment: The bolt specifications of commercial vehicles are diverse (ranging from M12 to M24), and the material differences are large (carbon steel, alloy steel). A unified threshold cannot adapt to all scenarios. Therefore, a dynamic torque threshold range adapted to the current bolt characteristics is extracted from the preset assembly parameter database (for example, the threshold for alloy steel M16 bolts is 280 - 320 N·m, and for carbon steel it is 220 - 260 N·m). By comparing the full-cycle torque feature vector with the threshold range, it is determined whether the bolt is assembled normally. Specifically, if the vector falls within the range and the characteristics are stable, it indicates that the assembly is qualified; if it exceeds the range or the characteristics mutate, it is marked as abnormal. This is for the convenience of subsequent processing.

[0085] Step S30: If it is determined that the bolt is not assembled normally according to the dynamic torque threshold range and the full-cycle torque feature vector, a process backtracking is performed, and the root cause of the torque deviation is located by combining fault tree analysis.

[0086] Among them, it should be noted that if it is judged that the assembly is abnormal, directly adjusting the parameters is likely to cause repeated failures. Therefore, it is necessary to first perform a process backtracking (restore the full process data of the assembly), and combine fault tree analysis (taking "torque deviation" as the top event and decomposing intermediate events such as "tool failure", "bolt defect", and "improper operation") to locate the root cause. This is for the convenience of subsequent processing.

[0087] Step S40: Generate corresponding assembly parameter optimization instructions according to the root cause of the torque deviation, and feedback them to the execution system to complete the normal assembly of the bolt correspondingly.

[0088] It's important to note that the assembly parameter optimization instructions generated based on the root cause (e.g., for deviations caused by tool vibration, the instruction is "reduce tool speed and calibrate torque sensor") are fed back to the execution system to complete the corrective assembly, ensuring that the problem is addressed at its root rather than merely masked. This facilitates subsequent processing.

[0089] Second Embodiment

[0090] Furthermore, the step of constructing the target mapping relationship between the torque-rotation angle curve and the bolt ultrasonic echo signal through the preset feature separation algorithm to generate the corresponding full-cycle torque feature vector includes:

[0091] The timestamps of the torque-rotation curve and the acquisition time of the ultrasonic echo signal of the bolt are aligned by a dynamic time warping algorithm to construct a corresponding target mapping relationship, and the effective vibration features in the vibration frequency of the tool are extracted by a variational mode decomposition algorithm.

[0092] Based on the target mapping relationship, dynamic correlation coefficients between torque parameters and effective vibration characteristics under different assembly stages are established to construct the corresponding correlation coefficient matrix.

[0093] Based on the correlation coefficient matrix, the torque-angle curves within the entire assembly cycle are subjected to feature enhancement processing, and bolt material and specification parameters are introduced as constraints to generate the full-cycle torque feature vector.

[0094] It's important to note that the first step involves data time alignment and noise filtering: There may be a time difference between the acquisition devices for the torque-angle curve and the ultrasonic echo signal (e.g., the torque sensor responds faster than the ultrasonic probe). Dynamic Time Warping (DTW) is used to align the timestamps of both, constructing a target mapping relationship of "torque at the same moment - the ultrasonic echo signal of the bolt," specifically avoiding association errors caused by time misalignment. Simultaneously, the tool vibration frequency contains interference signals such as motor noise and tool collisions. Variational Mode Decomposition (VMD) is used to separate these signals according to the frequency scale, extracting effective vibration features strongly correlated with torque changes (e.g., the stable vibration frequency of the tool during the tightening phase), and eliminating irrelevant noise.

[0095] The second step is to establish a phased correlation model: Bolt assembly is divided into three stages: pre-tightening (torque rises slowly), yielding (torque plateau period), and tightening (torque rises rapidly to its peak). The correlation between torque and vibration is completely different in each stage (e.g., vibration is small and stable in the pre-tightening stage, while vibration increases with increasing torque in the tightening stage). Based on the target mapping relationship, the dynamic correlation coefficients between torque parameters and effective vibration characteristics in different stages are calculated (e.g., correlation coefficient 0.2 in the pre-tightening stage and 0.8 in the tightening stage), and a correlation coefficient matrix is ​​constructed. Specifically, the matrix clearly presents the parameter correlation strength in each stage, providing a phased basis for subsequent feature extraction.

[0096] The third step involves feature enhancement and vector generation: Based on the correlation coefficient matrix, the torque-angle curve for the entire cycle is enhanced (e.g., a high correlation coefficient during the tightening stage amplifies the torque characteristics of that stage), highlighting information strongly related to assembly quality. Simultaneously, bolt material (affecting torque bearing capacity) and specifications (affecting torque transmission efficiency) are introduced as constraints. Specifically, for example, the torque characteristics of high-strength bolts must conform to the pattern of "high peak value and rapid rise." These constraints prevent the misjudgment of features suitable for ordinary bolts as abnormal. Finally, the enhanced features and constraints are fused to generate a full-cycle torque feature vector. This vector encompasses the overall data patterns and is adapted to the current bolt characteristics, facilitating subsequent processing.

[0097] Furthermore, the step of performing feature enhancement processing on the torque-angle curve throughout the entire assembly cycle based on the correlation coefficient matrix, and introducing bolt material and specification parameters as constraints to generate the full-cycle torque feature vector includes:

[0098] Based on the correlation coefficient matrix, density peak clustering is used to locate abnormal fluctuation segments in the torque-angle curve. An adaptive weighted algorithm is then used to correct outliers. The inverse of the correlation coefficient matrix is ​​used as the weight to dynamically gain the corrected torque-angle curve and output target feature data.

[0099] The core parameters of the bolt are extracted, and the constraint factors corresponding to the core parameters are screened out through grey relational analysis. The target feature data and the constraint factors are then fused across dimensions to generate a corresponding fused feature set.

[0100] The fused feature set is quantized to generate the full-cycle torque feature vector.

[0101] It should be noted that the first step, correcting anomalies and feature gain, involves the following: Regions in the correlation coefficient matrix exhibiting "abrupt changes in correlation coefficients" correspond to anomalous fluctuations in the torque-angle curve (e.g., tool jamming causing a sudden drop in torque, where the correlation coefficient between vibration characteristics and torque plummets from 0.8 to 0.1). Density peak clustering is used to quickly locate these fluctuation segments. An adaptive weighted algorithm corrects outliers (e.g., smoothing out sudden torque drops according to their preceding and following trends, rather than directly removing them) to ensure data continuity. The inverse of the correlation coefficient matrix is ​​used as the weight (regions with low correlation coefficients have higher weights, amplifying their features to identify potential problems), and dynamic gain is applied to the corrected data to output target feature data. Specifically, this data eliminates anomalous interference while retaining key potential information.

[0102] The second step is to achieve cross-dimensional feature fusion: extract the core parameters of the bolt (material strength, thread precision, surface treatment method), and use grey relational analysis to filter out constraint factors strongly correlated with these parameters (e.g., material strength corresponds to "peak torque constraint," and thread precision corresponds to "torque rise rate constraint"). The target feature data (data dimension) and constraint factors (physical dimension) are then fused across dimensions. Specifically, for example, the data feature "peak torque 180 N·m" combined with the constraint factor "material strength grade 8.8" forms the fused feature "peak torque 180 N·m (meets the requirements of grade 8.8 bolts)," generating a fused feature set. This fusion avoids misjudgments of "data being qualified but not conforming to the physical characteristics of the bolt," improving the actual adaptability of the features.

[0103] The third step is vector quantization: Qualitative features (such as "surface galvanization") in the fused feature set are converted into quantitative data through encoding. Quantitative features (such as torque values) are standardized (eliminating the influence of dimensions), ultimately generating a full-cycle torque feature vector with uniform dimensions and standardized values. Specifically, this vector can be directly input into the subsequent judgment model to ensure the accuracy and efficiency of the calculation, facilitating subsequent processing.

[0104] Furthermore, the step of determining whether the bolt is properly assembled based on the dynamic torque threshold range and the full-cycle torque feature vector includes:

[0105] A Mahalanobis distance algorithm is used to construct an adaptation network between the full-cycle torque feature vector and the dynamic torque threshold interval. Based on the adaptation network, the full-cycle torque feature vector is projected onto the two-dimensional evaluation plane corresponding to the dynamic torque threshold interval through feature dimension mapping, so as to output the minimum distance value between the feature point and the threshold boundary.

[0106] Combining the three stages of bolt assembly—preload, yield, and tightening—the full-cycle torque feature vector is decomposed into several stage sub-vectors, and the matching degree between each stage sub-vector and the standard stage feature template is calculated.

[0107] If the minimum distance value exceeds the preset distance threshold or the matching degree of any of the stage sub-vectors is lower than the preset matching degree threshold, it is marked as a preliminary anomaly. The preliminary anomaly result is then used to initiate multi-source data cross-validation to determine whether the bolt is properly assembled.

[0108] It should be noted that the first step is to construct an adaptation evaluation model: the full-cycle torque feature vector is multi-dimensional data (including torque at each stage, vibration correlation coefficients, etc.). A Mahalanobis distance algorithm is used to construct an adaptation network between this vector and the dynamic torque threshold range. Specifically, Mahalanobis distance can eliminate the influence of correlations between features (such as the linear correlation between torque and vibration), and is more suitable for distance calculation of multi-dimensional data than Euclidean distance. The high-dimensional vector is projected onto a two-dimensional evaluation plane through feature dimension mapping, outputting the minimum distance value between the feature point and the threshold boundary. Specifically, the smaller the distance value, the closer the assembly is to a normal state; if the distance value exceeds a preset threshold (e.g., corresponding to a torque deviation of ±5%), it is marked as a potential anomaly.

[0109] The second step involves phased matching verification: the three stages of bolt assembly (preload, yielding, and tightening) have clearly defined standard feature templates (e.g., the standard template for the yielding stage is "torque fluctuation ≤3%, duration 2-3 seconds"). The full-cycle torque feature vector is decomposed into three stage sub-vectors, and the matching degree of each sub-vector with the standard template is calculated. Specifically, for example, if the matching degree for the preload stage is 95% (normal), but the matching degree for the yielding stage is 60% (abnormal), it indicates a problem in the yielding stage (e.g., uneven bolt material leading to premature yielding). This phased decomposition can accurately pinpoint the specific stage where the anomaly occurs, avoiding the problem of "ambiguous overall judgment".

[0110] The third step involves initiating multi-source cross-validation: Distance values ​​or matching degrees in a single dimension may be affected by sensor fluctuations (such as temporary interference from the ultrasonic probe causing feature anomalies). Therefore, cross-validation is initiated for preliminary anomaly results. Specifically, this involves combining defect detection results from the ultrasonic echo signal and stability analysis of tool vibration to comprehensively determine whether it is a genuine anomaly. For example, "abnormal torque characteristics but normal ultrasonic signal and stable vibration" may be a false anomaly caused by torque sensor drift. If all three are abnormal, it is determined to be a genuine assembly problem, ensuring the reliability of the judgment results for subsequent processing.

[0111] Furthermore, the step of initiating multi-source data cross-validation of the preliminary abnormal results to determine whether the bolts are properly assembled includes:

[0112] The target assembly stage corresponding to the preliminary abnormal results is detected, and a multi-source verification feature set is constructed based on the abnormality type of the target assembly stage;

[0113] The ultrasonic resonant frequency deviation, vibration phase abrupt change value and torque characteristic anomaly parameters in the multi-source verification feature set are extracted, and converted into corresponding fuzzy confidence scores through fuzzy inference. The conditional probability of the multi-source verification feature set is calculated using a Bayesian network.

[0114] The fuzzy confidence and the conditional probability are fused to obtain a multi-dimensional anomaly confidence, and the bolt assembly status is determined based on the magnitude of the multi-dimensional anomaly confidence.

[0115] It should be noted that the first step focuses on the target verification scope: first, locate the target assembly stage corresponding to the initial anomaly (such as anomaly in the tightening stage), and construct a multi-source verification feature set according to the anomaly type (such as low torque peak). Specifically, it includes the ultrasonic resonant frequency (reflecting the internal stress of the bolt), the vibration phase change value (reflecting the compatibility between the tool and the bolt), and the torque characteristic anomaly parameters (reflecting the assembly force) of that stage, so as to avoid inefficiency caused by an excessively large verification scope.

[0116] The second step is to quantify abnormal features: features such as ultrasonic resonant frequency deviation (e.g., standard value 50kHz, measured 48kHz) and vibration phase change value (e.g., change of 15°) are qualitative descriptions. They are converted into fuzzy confidence levels through fuzzy reasoning (e.g., a deviation of 2kHz corresponds to a confidence level of 0.8 for "insufficient bolt stress"). Bayesian networks are used to calculate the conditional probabilities of multi-source features (e.g., when "torque is low and resonant frequency is low", the probability of assembly abnormality is 92%). Specifically, Bayesian networks are good at handling the probabilistic fusion of multi-source information and can effectively integrate the complementary information of various features.

[0117] The third step generates and judges multi-dimensional confidence scores: Fuzzy confidence scores (reflecting the degree of anomaly of a single feature) are weighted and fused with conditional probabilities (reflecting the reliability of the correlation between features) to obtain multi-dimensional anomaly confidence scores (e.g., a comprehensive confidence score of 0.93). A confidence score threshold is set (e.g., 0.8). If the score exceeds the threshold, it is judged as an assembly abnormality; if it is below the threshold, it is judged as a false anomaly and sensor calibration is triggered. Specifically, this quantitative judgment method avoids the subjectivity of "empirical judgment" and ensures the standardization and accuracy of anomaly judgment, facilitating subsequent processing.

[0118] Furthermore, the step of performing process backtracking and combining fault tree analysis to locate the root cause of torque deviation includes:

[0119] The entire process data chain of the assembly is traced through the preset distributed sensing unit, and the corresponding abnormal features contained in the entire process data chain are extracted.

[0120] Triggered by the aforementioned abnormal features, the assembly fault knowledge graph is invoked, and a corresponding fault tree is generated based on entity association rules;

[0121] Based on the fault tree, a corresponding fault propagation path is generated, and the fault propagation path is verified across domains to output the root cause of the torque deviation.

[0122] It should be noted that the first step is to trace the entire data chain: by associating the timestamps of the distributed sensing units, the data trajectory of the entire assembly process is restored. Specifically, from "tool start-pre-tightening start-yield-tightening end", torque, ultrasonic and vibration data are completely traced, and abnormal features are extracted (such as a sudden change in vibration phase when the tool is started, followed by a slow increase in torque). These features are direct clues for root cause location.

[0123] The second step is to construct a fault tree model: using the extracted abnormal features as triggering conditions (such as "slow torque increase"), the assembly fault knowledge graph is invoked. Specifically, this graph contains historical fault cases (such as "tool speed is too high, resulting in low torque transmission efficiency" and "bolt thread damage, resulting in high resistance") and entity association rules (such as "abnormal vibration → tool problem" and "ultrasonic abnormality → bolt problem") to generate a fault tree: the top event is "torque deviation", the intermediate events are "tool failure", "bolt defect" and "improper operating parameters", and the bottom events are the specific fault causes (such as tool torque sensor inaccuracy, bolt stripping, and excessively high speed setting).

[0124] The third step is to verify the fault propagation path: The fault tree may generate multiple potential propagation paths (such as "torque deviation → tool failure → sensor misalignment" or "torque deviation → bolt defect → stripped threads"). Cross-domain verification (combining tool maintenance records and bolt quality inspection reports) is used to filter out the true path. Specifically, for example, if the tool maintenance record shows "sensor not calibrated for 3 days" and the bolt quality inspection report shows no abnormalities, then the "sensor misalignment" path is the true path, and the corresponding root cause of the output torque deviation is "tool torque sensor misalignment," providing a clear direction for subsequent parameter optimization and facilitating subsequent processing.

[0125] Furthermore, the step of performing cross-domain verification on the fault propagation path to output the root cause of the torque deviation includes:

[0126] Extract the feature values ​​corresponding to each underlying node in the fault propagation path, and select the core verification objects based on the magnitude of the feature values;

[0127] A verification matrix adapted to the core verification object is created based on the assembly fault knowledge graph, so as to simulate the corresponding target deviation data based on the verification matrix and the core verification object;

[0128] In the fault propagation path, a target bottom-level node that matches the target deviation data is selected, and the target bottom-level node is set as the root cause of the torque deviation.

[0129] It should be noted that the first step is to screen the core verification objects: there may be multiple underlying nodes in the fault propagation path (such as "sensor inaccuracy", "excessive speed", "voltage fluctuation"). Extract the feature values ​​of each node (such as the feature value of sensor inaccuracy is "torque measurement error of 10%", and the feature value of excessive speed is "actual speed is 15% higher than the set speed"). Screen out the core verification objects with the most significant feature values ​​(such as sensor error of 10% as the core object). Specifically, focusing on the core objects can reduce the verification complexity and improve efficiency.

[0130] The second step is to construct and simulate a verification model: Based on the assembly fault knowledge graph, a verification matrix adapted to the core verification object is created. Specifically, the matrix contains the correspondence between "fault cause - simulated deviation data - verification index" (e.g., "sensor misalignment" corresponds to the simulated deviation data of "torque measurement value is 10% lower than the actual value", and the verification index is "matching degree between ultrasonic stress and torque"). The core verification object is substituted into the matrix to simulate and generate target deviation data (e.g., after simulating sensor misalignment, the matching degree between the torque curve and ultrasonic stress drops from 90% to 65%).

[0131] The third step is to match and locate the root cause: Among the bottom-level nodes of the fault propagation path, the target bottom-level node whose "actual deviation data corresponding to its characteristic value is consistent with the simulated target deviation data" is selected. Specifically, for example, if the actual deviation data of "sensor misalignment" (torque error 10%, matching degree 65%) is completely consistent with the simulated data, while the actual deviation data of "excessive speed" (torque error 3%, matching degree 85%) is inconsistent with the simulated data, then "sensor misalignment" is set as the root cause of torque deviation. This "simulation-matching" verification method ensures the uniqueness and accuracy of root cause location, providing a reliable basis for subsequent parameter optimization and facilitating subsequent processing.

[0132] Please see Figure 2 The third embodiment of the present invention provides:

[0133] A commercial vehicle assembly inspection system, wherein the system includes:

[0134] The acquisition module is used to acquire the torque-rotation curve of bolt assembly, bolt ultrasonic echo signal and tool vibration frequency based on the assembly station coordinates through a preset distributed sensing unit, so as to input them into a preset feature separation algorithm, and construct the target mapping relationship between the torque-rotation curve and the bolt ultrasonic echo signal through the preset feature separation algorithm to generate the corresponding full-cycle torque feature vector.

[0135] The extraction module is used to extract the dynamic torque threshold range that is suitable for the current bolt characteristics from the preset assembly parameter database, and to determine whether the bolt is assembled normally based on the dynamic torque threshold range and the full-cycle torque feature vector.

[0136] The analysis module is used to perform process backtracking and locate the root cause of torque deviation by combining fault tree analysis if it is determined that the bolt assembly is abnormal based on the dynamic torque threshold range and the full-cycle torque feature vector.

[0137] The generation module is used to generate corresponding assembly parameter optimization instructions based on the root cause of the torque deviation, and feed them back to the execution system to complete the normal assembly of the bolts.

[0138] Furthermore, the acquisition module is specifically used for:

[0139] The timestamps of the torque-rotation curve and the acquisition time of the ultrasonic echo signal of the bolt are aligned by a dynamic time warping algorithm to construct a corresponding target mapping relationship, and the effective vibration features in the vibration frequency of the tool are extracted by a variational mode decomposition algorithm.

[0140] Based on the target mapping relationship, dynamic correlation coefficients between torque parameters and effective vibration characteristics under different assembly stages are established to construct the corresponding correlation coefficient matrix.

[0141] Based on the correlation coefficient matrix, the torque-angle curves within the entire assembly cycle are subjected to feature enhancement processing, and bolt material and specification parameters are introduced as constraints to generate the full-cycle torque feature vector.

[0142] Furthermore, the acquisition module is specifically used for:

[0143] Based on the correlation coefficient matrix, density peak clustering is used to locate abnormal fluctuation segments in the torque-angle curve. An adaptive weighted algorithm is then used to correct outliers. The inverse of the correlation coefficient matrix is ​​used as the weight to dynamically gain the corrected torque-angle curve and output target feature data.

[0144] The core parameters of the bolt are extracted, and the constraint factors corresponding to the core parameters are screened out through grey relational analysis. The target feature data and the constraint factors are then fused across dimensions to generate a corresponding fused feature set.

[0145] The fused feature set is quantized to generate the full-cycle torque feature vector.

[0146] Furthermore, the extraction module is specifically used for:

[0147] A Mahalanobis distance algorithm is used to construct an adaptation network between the full-cycle torque feature vector and the dynamic torque threshold interval. Based on the adaptation network, the full-cycle torque feature vector is projected onto the two-dimensional evaluation plane corresponding to the dynamic torque threshold interval through feature dimension mapping, so as to output the minimum distance value between the feature point and the threshold boundary.

[0148] Combining the three stages of bolt assembly—preload, yield, and tightening—the full-cycle torque feature vector is decomposed into several stage sub-vectors, and the matching degree between each stage sub-vector and the standard stage feature template is calculated.

[0149] If the minimum distance value exceeds the preset distance threshold or the matching degree of any of the stage sub-vectors is lower than the preset matching degree threshold, it is marked as a preliminary anomaly. The preliminary anomaly result is then used to initiate multi-source data cross-validation to determine whether the bolt is properly assembled.

[0150] Furthermore, the extraction module is specifically used for:

[0151] The target assembly stage corresponding to the preliminary abnormal results is detected, and a multi-source verification feature set is constructed based on the abnormality type of the target assembly stage;

[0152] The ultrasonic resonant frequency deviation, vibration phase abrupt change value and torque characteristic anomaly parameters in the multi-source verification feature set are extracted, and converted into corresponding fuzzy confidence scores through fuzzy inference. The conditional probability of the multi-source verification feature set is calculated using a Bayesian network.

[0153] The fuzzy confidence and the conditional probability are fused to obtain a multi-dimensional anomaly confidence, and the bolt assembly status is determined based on the magnitude of the multi-dimensional anomaly confidence.

[0154] Furthermore, the analysis module is specifically used for:

[0155] The entire process data chain of the assembly is traced through the preset distributed sensing unit, and the corresponding abnormal features contained in the entire process data chain are extracted.

[0156] Triggered by the aforementioned abnormal features, the assembly fault knowledge graph is invoked, and a corresponding fault tree is generated based on entity association rules;

[0157] Based on the fault tree, a corresponding fault propagation path is generated, and the fault propagation path is verified across domains to output the root cause of the torque deviation.

[0158] Furthermore, the analysis module is specifically used for:

[0159] Extract the feature values ​​corresponding to each underlying node in the fault propagation path, and select the core verification objects based on the magnitude of the feature values;

[0160] A verification matrix adapted to the core verification object is created based on the assembly fault knowledge graph, so as to simulate the corresponding target deviation data based on the verification matrix and the core verification object;

[0161] In the fault propagation path, a target bottom-level node that matches the target deviation data is selected, and the target bottom-level node is set as the root cause of the torque deviation.

[0162] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the commercial vehicle assembly inspection method as described above.

[0163] The fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the commercial vehicle assembly inspection method as described above.

[0164] In summary, the commercial vehicle assembly inspection method and system provided in the above embodiments of the present invention can objectively and effectively detect whether the bolts are properly assembled, thereby improving the assembly efficiency of commercial vehicles.

[0165] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0166] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0167] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0168] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0169] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0170] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A method for inspecting the assembly of commercial vehicles, characterized in that, The method includes: Based on the assembly station coordinates, a preset distributed sensing unit collects the torque-rotation angle curve of bolt assembly, bolt ultrasonic echo signal, and tool vibration frequency. These data are then input into a preset feature separation algorithm, which constructs a target mapping relationship between the torque-rotation angle curve and the bolt ultrasonic echo signal to generate a corresponding full-cycle torque feature vector. The steps for generating the full-cycle torque feature vector include: The timestamps of the torque-rotation curve and the acquisition time of the ultrasonic echo signal of the bolt are aligned by a dynamic time warping algorithm to construct a corresponding target mapping relationship, and the effective vibration features in the vibration frequency of the tool are extracted by a variational mode decomposition algorithm. Based on the target mapping relationship, dynamic correlation coefficients between torque parameters and effective vibration characteristics under different assembly stages are established to construct the corresponding correlation coefficient matrix. Based on the correlation coefficient matrix, the torque-angle curves within the entire assembly cycle are subjected to feature enhancement processing, and bolt material and specification parameters are introduced as constraints to generate the full-cycle torque feature vector accordingly. Extract the dynamic torque threshold range that is compatible with the current bolt characteristics from the preset assembly parameter database, and determine whether the bolt is properly assembled based on the dynamic torque threshold range and the full-cycle torque feature vector. If the bolt assembly is determined to be abnormal based on the dynamic torque threshold range and the full-cycle torque feature vector, then process backtracking is performed, and the root cause of the torque deviation is located by combining fault tree analysis. Based on the root cause of the torque deviation, corresponding assembly parameter optimization instructions are generated and fed back to the execution system to complete the normal assembly of the bolts.

2. The commercial vehicle assembly inspection method according to claim 1, characterized in that, The step of performing feature enhancement processing on the torque-angle curve throughout the entire assembly cycle based on the correlation coefficient matrix, and introducing bolt material and specification parameters as constraints to generate the full-cycle torque feature vector includes: Based on the correlation coefficient matrix, density peak clustering is used to locate abnormal fluctuation segments in the torque-angle curve. An adaptive weighted algorithm is then used to correct outliers. The inverse of the correlation coefficient matrix is ​​used as the weight to dynamically gain the corrected torque-angle curve and output target feature data. The core parameters of the bolt are extracted, and the constraint factors corresponding to the core parameters are screened out through grey relational analysis. The target feature data and the constraint factors are then fused across dimensions to generate a corresponding fused feature set. The fused feature set is quantized to generate the full-cycle torque feature vector.

3. The commercial vehicle assembly inspection method according to claim 1, characterized in that, The step of determining whether the bolt is properly assembled based on the dynamic torque threshold range and the full-cycle torque feature vector includes: A Mahalanobis distance algorithm is used to construct an adaptation network between the full-cycle torque feature vector and the dynamic torque threshold interval. Based on the adaptation network, the full-cycle torque feature vector is projected onto the two-dimensional evaluation plane corresponding to the dynamic torque threshold interval through feature dimension mapping, so as to output the minimum distance value between the feature point and the threshold boundary. Combining the three stages of bolt assembly—preload, yield, and tightening—the full-cycle torque feature vector is decomposed into several stage sub-vectors, and the matching degree between each stage sub-vector and the standard stage feature template is calculated. If the minimum distance value exceeds the preset distance threshold or the matching degree of any of the stage sub-vectors is lower than the preset matching degree threshold, it is marked as a preliminary anomaly. The preliminary anomaly result is then used to initiate multi-source data cross-validation to determine whether the bolt is properly assembled.

4. The commercial vehicle assembly inspection method according to claim 3, characterized in that, The step of initiating multi-source data cross-validation of the preliminary abnormal results to determine whether the bolts are properly assembled includes: The target assembly stage corresponding to the preliminary abnormal results is detected, and a multi-source verification feature set is constructed based on the abnormality type of the target assembly stage; The ultrasonic resonant frequency deviation, vibration phase abrupt change value and torque characteristic anomaly parameters in the multi-source verification feature set are extracted, and converted into corresponding fuzzy confidence scores through fuzzy inference. The conditional probability of the multi-source verification feature set is calculated using a Bayesian network. The fuzzy confidence and the conditional probability are fused to obtain a multi-dimensional anomaly confidence, and the bolt assembly status is determined based on the magnitude of the multi-dimensional anomaly confidence.

5. The commercial vehicle assembly inspection method according to claim 1, characterized in that, The steps of performing process backtracking and combining fault tree analysis to locate the root cause of torque deviation include: The entire process data chain of the assembly is traced through the preset distributed sensing unit, and the corresponding abnormal features contained in the entire process data chain are extracted. Triggered by the aforementioned abnormal features, the assembly fault knowledge graph is invoked, and a corresponding fault tree is generated based on entity association rules; Based on the fault tree, a corresponding fault propagation path is generated, and the fault propagation path is verified across domains to output the root cause of the torque deviation.

6. The commercial vehicle assembly inspection method according to claim 5, characterized in that, The step of performing cross-domain verification of the fault propagation path to output the root cause of the torque deviation includes: Extract the feature values ​​corresponding to each underlying node in the fault propagation path, and select the core verification objects based on the magnitude of the feature values; A verification matrix adapted to the core verification object is created based on the assembly fault knowledge graph, so as to simulate the corresponding target deviation data based on the verification matrix and the core verification object; In the fault propagation path, a target bottom-level node that matches the target deviation data is selected, and the target bottom-level node is set as the root cause of the torque deviation.

7. A commercial vehicle assembly inspection system, characterized in that, The system includes: The acquisition module is used to acquire the torque-rotation angle curve of bolt assembly, bolt ultrasonic echo signal, and tool vibration frequency based on the assembly station coordinates through a preset distributed sensing unit. These data are then input into a preset feature separation algorithm, which constructs a target mapping relationship between the torque-rotation angle curve and the bolt ultrasonic echo signal to generate a corresponding full-cycle torque feature vector. The step of generating the full-cycle torque feature vector includes: The timestamps of the torque-rotation curve and the acquisition time of the ultrasonic echo signal of the bolt are aligned by a dynamic time warping algorithm to construct a corresponding target mapping relationship, and the effective vibration features in the vibration frequency of the tool are extracted by a variational mode decomposition algorithm. Based on the target mapping relationship, dynamic correlation coefficients between torque parameters and effective vibration characteristics under different assembly stages are established to construct the corresponding correlation coefficient matrix. Based on the correlation coefficient matrix, the torque-angle curves within the entire assembly cycle are subjected to feature enhancement processing, and bolt material and specification parameters are introduced as constraints to generate the full-cycle torque feature vector accordingly. The extraction module is used to extract the dynamic torque threshold range that is suitable for the current bolt characteristics from the preset assembly parameter database, and to determine whether the bolt is assembled normally based on the dynamic torque threshold range and the full-cycle torque feature vector. The analysis module is used to perform process backtracking and locate the root cause of torque deviation by combining fault tree analysis if it is determined that the bolt assembly is abnormal based on the dynamic torque threshold range and the full-cycle torque feature vector. The generation module is used to generate corresponding assembly parameter optimization instructions based on the root cause of the torque deviation, and feed them back to the execution system to complete the normal assembly of the bolts.

8. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the commercial vehicle assembly inspection method as described in any one of claims 1 to 6.

9. A readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the commercial vehicle assembly inspection method as described in any one of claims 1 to 6.

Images