Intelligent automobile part airtightness detection method based on intelligent sensing technology

Abstract

The invention relates to the technical field of automobile part detection, in particular to an automobile part airtightness intelligent detection method based on an intelligent sensing technology, which comprises the following steps: establishing a digital three-dimensional geometric model for an automobile part, and arranging pressure and flow sensing nodes at key sealing positions to form a multi-physical-quantity synchronous sensing array. A part is placed in a closed cavity, gas is injected according to a preset pressurization curve, and the dynamic pressure distribution sequence and the micro-flow fluctuation sequence of the inner and outer surfaces are synchronously collected. And constructing a surface stress evolution graph based on the pressure sequence, comparing and identifying stress abnormal regions region by region, constructing a leakage airflow vector field based on the flow sequence, matching and positioning leakage coordinates and dominant directions, and fusing various defect characteristics to generate a sealing integrity comprehensive diagnosis report. According to the method, local fine defects of the seal can be accurately captured, synchronous positioning of a leakage position and stress abnormity is realized, and refinement and diagnosis accuracy of airtightness detection are improved.

Description

Technical Field

[0001] This invention relates to the field of automotive parts testing technology, and in particular to an intelligent testing method for the airtightness of automotive parts based on intelligent sensing technology. Background Technology

[0002] Current methods for airtightness testing of automotive components often employ single-point testing using a single pressure or flow sensor. The sealing performance of the component is determined based on the overall pressure drop or flow rate change within the testing chamber. This approach fails to integrate sensor placement with the component's digital 3D geometric model, relying instead on scattered sensor units to collect single physical quantity data in an asynchronous, discrete acquisition mode. The existing testing process only provides numerical assessments of the overall pressure and flow parameters of the component, failing to construct dynamic stress evolution maps of the component surface based on pressure change data, nor to construct a leakage airflow vector field across the sealing boundary based on micro-flow fluctuation data.

[0003] Traditional detection methods cannot capture the details of local pressure changes and micro-flow fluctuations at critical sealing locations of components. Overall parameter determination can obscure the characteristic information of local sealing defects, making it impossible to distinguish between areas of abnormal stress concentration and areas of stress deficiency on the component surface. Existing technologies can only determine whether a component is leaking, but cannot locate the leak location spatially or determine the dominant flow direction. The analysis of sealing defects is limited in scope and lacks sufficient positioning accuracy.

[0004] It is necessary to rely on the digital three-dimensional geometric model of the parts to complete the arrayed deployment and synchronous data acquisition of multiple physical quantity sensing nodes at key sealing positions. It is also necessary to achieve multi-dimensional fusion diagnosis of sealing defects by comparing regional differences in stress evolution maps and matching the spatial shape and intensity of leakage airflow vector field, so as to make up for the shortcomings of existing detection in capturing local features and accurately locating defects. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose an intelligent detection method for the airtightness of automotive parts based on intelligent sensing technology.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent detection method for the airtightness of automotive parts based on intelligent sensing technology, comprising: A digital three-dimensional geometric model is established for automotive parts of a predetermined model, and multiple pressure sensing nodes and flow sensing nodes are deployed at key sealing locations based on the digital three-dimensional geometric model to form a multi-physical quantity synchronous sensing array. The automotive parts to be tested are placed in a sealed testing chamber, and testing gas is injected into the testing chamber through a pressurization pipeline according to a preset pressurization curve. At the same time, the dynamic pressure distribution sequence and micro-flow fluctuation sequence of the inner and outer surfaces of the automotive parts are collected by the multi-physical quantity synchronous sensing array. Based on the dynamic pressure distribution sequence, a surface stress evolution map of automotive parts during the pressurization process is constructed, and based on the micro-flow fluctuation sequence, a leakage airflow vector field passing through the sealing boundary of the parts is constructed. By comparing the surface stress evolution map with the pre-stored standard stress distribution template region by region, areas of abnormal stress concentration and areas of stress deficiency are identified. The spatial shape and intensity matching analysis of the leaked airflow vector field and the pre-stored standard leak field are performed to locate the spatial coordinates of the leak and the dominant leak direction. By integrating the stress anomaly concentration area, stress deficiency area, and the spatial coordinates of the leakage occurrence and the dominant leakage direction, a comprehensive defect diagnosis report on the sealing integrity of automotive components is generated.

[0007] As a further aspect of the present invention, a digital three-dimensional geometric model is established for a predetermined model of automotive parts, including: The surface contour of standard automotive parts is scanned from all angles using a 3D laser scanning device to collect high-density point cloud data. The point cloud registration algorithm is used to stitch together the point cloud data obtained from multiple scans to form a point cloud model that completely covers the surface of the automotive parts. The point cloud model is triangulated to generate a surface mesh model representing the three-dimensional shape of the automotive parts. Based on the surface mesh model and combined with the design drawings of automotive parts, the key sealing structures and the preset placement positions of sensors are geometrically reconstructed and feature-annotated. The surface mesh model with feature annotations, along with the preset sensor node coordinates and connection relationships, are integrated into the digital three-dimensional geometric model. As a further aspect of the present invention, a surface stress evolution map of automotive components during the pressurization process is constructed based on the dynamic pressure distribution sequence, including: From the dynamic pressure distribution sequence, the pressure readings of each pressure sensing node during the complete pressurization cycle are extracted in chronological order to form a multi-node pressure-time curve cluster with time as the horizontal axis and pressure as the vertical axis. The multi-node pressure-time curve cluster is synchronized and aligned to ensure that the data time base of all pressure sensing nodes is consistent. The coordinates of each pressure sensing node are mapped to the corresponding position in the digital three-dimensional geometric model, and the pressure value of each node at the same time is used as the surface stress characterization value of the coordinate point. Several characteristic time points are selected during the pressurization process. For each characteristic time point, the surface stress characterization values ​​of all pressure sensing nodes are spatially interpolated to generate a stress distribution cloud map covering the entire surface of the automotive parts at the characteristic time point. Multiple stress distribution cloud maps generated in chronological order are superimposed and dynamically evolved to construct a surface stress evolution map that reflects the distribution of pressure from injection to stabilization.

[0008] As a further aspect of the present invention, the multi-node pressure-time curve cluster is subjected to synchronous alignment processing, including: Obtain the timestamp of the opening of the pressure regulating valve in the pressurization pipeline, and define the timestamp as the pressurization start time; Identify the inflection point in the pressure-time curve of each pressure sensing node where the pressure first shows a significant increase; Calculate the time difference between the inflection point time and the pressurization start time for each node, and use it as the data transmission delay of the node; Based on the data transmission delay of each node, time shift compensation is performed on its pressure-time curve to align the pressure start-up time of all nodes with the pressurization start time.

[0009] As a further aspect of the present invention, a leakage airflow vector field crossing the sealing boundary of the component is constructed based on the micro-flow fluctuation sequence, including: From the micro-flow fluctuation sequence, the direction and magnitude of the gas flow velocity detected by each flow sensing node are analyzed. Based on the relative positions of all flow sensing nodes in space, a spatial network topology describing the connectivity of gas flow is constructed. By utilizing the spatial network topology and the flow velocity direction of each node, the gas flow path is traced and analyzed to predict the migration path of the gas inside the automotive parts. Based on the predicted gas migration path and the velocity data, the volumetric flow rate change of each segment along the gas migration path is calculated. On the sealed boundary of the digital three-dimensional geometric model, vector labels are made at locations where there are significant changes in volumetric flow rate. The vector labels include the location coordinates of the leak point, the leakage flow rate intensity, and the airflow direction. All vector labels are combined to form the leakage airflow vector field.

[0010] As a further aspect of the present invention, the surface stress evolution map is compared with a pre-stored standard stress distribution template region by region to identify regions of stress anomaly concentration and regions of stress deficiency, including: The standard stress distribution template corresponding to the model of the automotive part to be tested and the test pressure is invoked. The standard stress distribution template records the reference stress value range of each area on the surface of the part under qualified sealing conditions at each stage of pressurization. In the surface stress evolution spectrum, the same characteristic time point as the standard stress distribution template is selected, and the actual stress value at the same coordinate position is extracted; The actual stress value at each coordinate position at each characteristic time point is compared with the range of the reference stress value; A continuous spatial region where the actual stress value is consistently higher than the upper limit of the corresponding reference stress value range is defined as the stress anomaly concentration region. The continuous spatial region where the actual stress value is consistently lower than the lower limit of the corresponding reference stress value range is marked and defined as the stress deficiency region.

[0011] As a further aspect of the present invention, the spatial shape and intensity matching analysis of the leaked gas flow vector field and the pre-stored standard leak field is performed to locate the spatial coordinates of the leak and the dominant leak direction, including: The standard leakage field corresponding to the model of the automotive component to be tested and the test pressure is invoked. The standard leakage field describes the spatial distribution and intensity range of the expected leakage vector on the sealing boundary of the digital three-dimensional geometric model under permissible micro-leakage conditions. Each leakage vector in the leakage airflow vector field is compared with the expected leakage vector at the corresponding spatial location in the standard leakage field; When the intensity of the actual leakage vector at a certain coordinate location exceeds the upper limit of the intensity specified in the standard leakage field, that coordinate location is determined to be a potential leakage point; For all identified potential leak points, calculate the angle between the actual leak vector and the expected leak vector. Potential leak points whose directional angle exceeds a preset angle tolerance or whose leakage intensity exceeds the intensity limit are selected, their spatial coordinates are determined as the spatial coordinates of the leak occurrence, and the direction of their actual leakage vector is determined as the dominant leakage direction.

[0012] As a further aspect of the present invention, the stress anomaly concentration area, the stress deficiency area, and the spatial coordinates of the leakage occurrence and the dominant leakage direction are integrated to generate a comprehensive defect diagnosis report on the sealing integrity of automotive components, including: On the digital three-dimensional geometric model, the stress anomaly concentration area, the stress deficiency area, and the spatial coordinates of the leakage occurrence are highlighted and marked respectively. Analyze the spatial relationship between the stress anomaly concentration area, the stress deficiency area and the spatial coordinates; When the spatial coordinates of the leakage are located at or adjacent to the area of ​​abnormal stress concentration, the correlation record in the diagnostic report is "high pressure concentration leads to seal failure"; When the spatial coordinates of the leakage are located at or adjacent to the stress deficiency area, the diagnostic report will record the correlation as "insufficient stress leading to poor sealing". When the spatial coordinates of the leak are neither in the area of ​​abnormal stress concentration nor in the area of ​​stress deficiency, but are associated with the leak path indicated by the dominant leak direction, the correlation is recorded as "leak caused by assembly or material defects" in the diagnostic report. By summarizing all vector annotations, spatial relationships, and correlation records, a comprehensive defect diagnosis report is generated, which includes the defect location, defect type, and cause.

[0013] As a further aspect of the present invention, the method further includes a step of feedback of detection results and optimization of process parameters based on the comprehensive defect diagnosis report: Analyze the comprehensive defect diagnosis reports of the same batch of automotive parts to extract common defect types and defect location distribution characteristics; The common defect types and defect location distribution characteristics are correlated with the production process parameters of automotive parts, including glue injection pressure, bolt tightening torque sequence, and welding temperature profile. When a correlation is identified between a defect pattern and a process parameter value range, adjustment suggestions for the production process parameters are generated. The adjustment suggestions are fed back to the production line control system for adaptive optimization of the process parameters of automotive parts produced subsequently.

[0014] As a further aspect of the present invention, the method further includes a sensor array self-calibration and sealing reference establishment step performed before the injection of the detection gas: A stable reference negative pressure environment is established in the detection chamber, and the initial readings of all pressure sensing nodes and flow sensing nodes in the multi-physical quantity synchronous sensing array are recorded at this time. For standard defect-free samples of the automotive parts, a complete pressure testing process was performed to collect their standard dynamic pressure distribution sequence and standard micro-flow fluctuation sequence. Based on the standard dynamic pressure distribution sequence, the standard stress distribution template is constructed and stored in the database; Based on the standard micro-flow fluctuation sequence, the standard leakage field is constructed and stored in the database; The initial reading benchmark, standard stress distribution template, and standard leakage field will be used together as the comparison benchmark for subsequent testing of automotive parts of the same model.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are as follows: Based on the digital 3D geometric model of a predetermined automotive component, multiple pressure and flow sensing nodes are directionally deployed at key sealing locations to form a multi-physical quantity synchronous sensing array. This array synchronously collects dynamic pressure distribution sequences and micro-flow fluctuation sequences from the inner and outer surfaces of the automotive component. This allows for the acquisition of localized multi-physical quantity synchronous data at key sealing locations, abandoning the data acquisition mode of single-physical-quantity, single-point asynchronous acquisition. It fully preserves the dynamic changes in local pressure and micro-flow during component pressurization, accurately captures subtle parameter changes at the sealing location, and reconstructs the true physical quantity changes at the sealing location.

[0016] A surface stress evolution map is constructed based on a dynamic pressure distribution sequence and compared with a pre-stored standard stress distribution template for regional differences. A leakage airflow vector field is constructed based on a micro-flow fluctuation sequence and analyzed for spatial morphology and intensity matching with a pre-stored standard leakage field. A comprehensive defect diagnosis report is generated by integrating stress anomaly concentration areas, stress deficiency areas, leakage spatial coordinates, and the dominant leakage direction. This system can accurately distinguish between stress anomaly concentration areas and stress deficiency areas on component surfaces, pinpoint the specific spatial coordinates of the leakage, clarify the dominant flow direction of the leakage, refine the analysis dimensions of sealing defects, and form a complete diagnostic report covering stress characteristics and leakage characteristics, achieving a full-dimensional presentation of sealing defects from local feature identification to spatial location. Attached Figure Description

[0017] Figure 1 The flowchart is a process for the intelligent airtightness detection method for automotive parts based on intelligent sensing technology as described in this invention. Figure 2 A flowchart for constructing a surface stress evolution map; Figure 3 This is a diagram showing the flow rate changes along the leakage path. Figure 4 Diagram for comprehensive defect diagnosis of sealing integrity of automotive parts; Figure 5 An optimized analysis chart showing the relationship between bolt tightening torque and defect incidence. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Example See Figure 1 This invention provides an intelligent detection method for the airtightness of automotive parts based on intelligent sensing technology. The specific method is as follows: A digital 3D geometric model is established for automotive parts of a predetermined model. Based on this model, multiple pressure and flow sensing nodes are deployed at key sealing locations, forming a multi-physical quantity synchronous sensing array. During testing, the part to be tested is placed in a sealed testing chamber, and test gas is injected through a pressurization pipeline according to a preset pressurization curve. Simultaneously, the sensing array collects dynamic pressure distribution sequences and micro-flow fluctuation sequences on the inner and outer surfaces of the part. Based on the dynamic pressure distribution sequence, a surface stress evolution map is constructed during the pressurization process, and based on the micro-flow fluctuation sequence, a leakage airflow vector field crossing the sealing boundary is constructed. The surface stress evolution map is compared with a pre-stored standard stress distribution template region by region to identify areas of abnormal stress concentration and areas of stress deficiency. The leakage airflow vector field is analyzed for spatial morphology and intensity matching with a pre-stored standard leakage field to locate the spatial coordinates of the leakage and the dominant leakage direction. By integrating the areas of abnormal stress concentration, areas of stress deficiency, spatial coordinates, and dominant leakage direction, a comprehensive defect diagnosis report on the sealing integrity of the automotive part is generated.

[0020] In one embodiment of the present invention, a digital three-dimensional geometric model is established for a predetermined model of automotive parts. A three-dimensional laser scanning device is used to scan the surface contour of the standard sample automotive parts from all angles, acquiring high-density point cloud data. A point cloud registration algorithm is used to stitch together the point cloud data obtained from multiple scans, forming a point cloud model that completely covers the surface of the automotive parts. The point cloud model is then triangulated to generate a surface mesh model representing the three-dimensional shape of the automotive parts. Based on the surface mesh model and combined with the design drawings of the automotive parts, the key sealing structures and the preset placement positions of sensors are geometrically reconstructed and feature-annotated. The surface mesh model with feature annotations, along with the preset sensor node coordinates and connection relationships, are integrated into a digital three-dimensional geometric model.

[0021] In the specific implementation, an aluminum alloy cylinder head of a four-cylinder engine is used as an example of a predetermined automotive component to illustrate the process of establishing a digital three-dimensional geometric model. During the implementation, a high-precision line laser 3D scanning device is used to perform a full-angle scan of the surface contour of the standard sample engine cylinder head. The scanning device collects data from multiple perspectives around the cylinder head sample fixed on a rotary worktable. The scanning line spacing is set to 0.05 mm, and the coverage angle of a single scan is 60 degrees. Through six scans from different directions, high-density point cloud data covering the entire outer surface of the cylinder head, the inner wall of the water jacket channel, and the inner wall of the bolt holes are obtained. The initial point cloud set contains approximately 12 million three-dimensional coordinate points.

[0022] After the scanning was completed, the point cloud data obtained from the six scans were stitched together using a point cloud registration algorithm. Specifically, the iterative nearest-point algorithm was employed. This algorithm uses the point cloud data from the first scan as a reference and calculates the optimal rigid body transformation matrix between the point cloud data from each subsequent scan and the reference point cloud to achieve precise alignment. The objective function F of the iterative nearest-point algorithm minimizes the sum of squared distances between corresponding point pairs. The function expression is as follows:

[0023] in: Represents a rotation matrix. Represents the translation vector. Represents the points in the point cloud to be registered. Representing the benchmark point in the cloud and The corresponding nearest neighbor, To find the number of corresponding point pairs, the function is solved iteratively. Value minimization and The parameters are used to complete the spatial coordinate system of all six sets of point cloud data and seamlessly stitch them together to form a single dense point cloud model that completely covers all surfaces of the engine cylinder head.

[0024] The stitched complete point cloud model is triangulated using the Poisson surface reconstruction algorithm. The algorithm implicitly fits a smooth isosurface based on the normal vector information of the points in the point cloud data. Then, the isosurface is meshed to generate a closed surface mesh model composed of triangular facets. The generated surface mesh model contains approximately 6 million triangular facets, which can accurately characterize the complex three-dimensional shape of the engine cylinder head, the cooling water jacket cavity, and the features of each mounting boss. Based on the generated surface mesh model, combined with the two-dimensional design drawings and three-dimensional computer-aided design drawings of the engine cylinder head, the key sealing structures and the preset placement positions of sensors are geometrically reconstructed and feature-annotated on the model. In the specific implementation, the boundary loop of the cylinder head gasket sealing surface is manually outlined on the surface mesh model using computer-aided engineering software, and the parametric surface representation of the sealing surface is reconstructed. At the same time, similar geometric feature extraction and surface reconstruction are performed on the intake manifold mating surface, exhaust manifold mating surface, and camshaft bearing cover mating surface. On the reconstructed sealing surface, according to the installation coordinates of the pressure sensing nodes and flow sensing nodes predefined in the detection scheme, virtual sensor icons are created at the corresponding three-dimensional coordinate positions in the model, and attribute information is associated with each icon, including sensor type, number, range, and logical address in the sensing array. Understandably, the surface mesh model with feature annotations, along with the preset sensor node coordinates and connection relationships, will eventually be integrated into a digital three-dimensional geometric model of the engine cylinder head. This model is stored in a standard triangular mesh file format. The file not only contains vertex and face information, but also stores all the annotated sealing surface geometric parameters and sensor node attributes and topological connection relationships in the form of an additional annotation layer.

[0025] In one embodiment of the present invention, see [reference] Figure 2Based on a dynamic pressure distribution sequence, a surface stress evolution map of automotive components during the pressurization process is constructed. Pressure readings for each pressure sensing node within a complete pressurization cycle are extracted chronologically from the dynamic pressure distribution sequence, forming a multi-node pressure-time curve cluster with time as the horizontal axis and pressure as the vertical axis. The multi-node pressure-time curve cluster is synchronized and aligned. The timestamp of the opening of the pressure regulating valve in the pressurization pipeline is obtained and defined as the pressurization start time. The inflection point when the pressure first significantly increases in the pressure-time curve of each pressure sensing node is identified. The time difference between the inflection point and the pressurization start time of each node is calculated as the data transmission delay of the node. Time shift compensation is applied to the pressure-time curves of each node based on the data transmission delay, aligning the pressure start time of all nodes to the pressurization start time. The coordinates of each pressure sensing node are mapped to the corresponding position in the digital 3D geometric model, and the pressure value of each node at the same moment is used as the surface stress characterization value of the coordinate point. Several characteristic time points during the pressurization process are selected, and spatial interpolation is performed on the surface stress characterization values ​​of all pressure sensing nodes at each characteristic time point to generate a stress distribution cloud map covering the entire surface of the automotive component at that characteristic time point. Multiple stress distribution cloud maps generated in chronological order are then overlaid and dynamically evolved to construct a surface stress evolution map reflecting the entire process of pressure distribution from injection to stabilization.

[0026] In practical implementation, taking the engine cylinder head airtightness testing process as an example, this paper illustrates the method for constructing a surface stress evolution map. In this implementation, a total of 48 miniature pressure sensors are pre-embedded on the upper and lower surfaces, sides, and internal water jacket of the engine cylinder head, forming a sensing array. During the 20-second pressurization process of injecting nitrogen into the sealed testing chamber, the readings of all pressure sensing nodes are simultaneously recorded at a sampling frequency of 1000 Hz, forming a dynamic pressure distribution sequence. From the dynamic pressure distribution sequence, the pressure readings of each pressure sensing node within the complete pressurization cycle are extracted in chronological order, generating 48 pressure-time curves. These curves together form a multi-node pressure-time curve cluster with time as the horizontal axis and pressure as the vertical axis. Each curve corresponds to the data record of one sensing node from the start of pressure rise to its stable holding period.

[0027] The multi-node pressure-time curve cluster is synchronized and aligned. The data processing system obtains the precise electrical signal timestamp of the opening of the high-speed solenoid valve in the pressurization pipeline and uses this timestamp... Defined as the initiation time of pressurization. In some embodiments, the algorithm automatically identifies the inflection point in each pressure-time curve where the pressure value first exceeds the initial baseline pressure by 10%. Calculate the inflection point time of each pressure sensing node. With the start of pressurization Time difference Time difference This is defined as the data transmission delay of that pressure sensing node. The data transmission delay is based on the specific pressure sensing node. The corresponding pressure-time curve is compensated for by overall time shift, that is, the time axis of the entire curve is shifted to the left. This ensures that the pressure rise time of all 48 pressure-time curves is aligned with a uniform pressurization start time on the time axis. This eliminates time deviations caused by differences in sensor response and asynchrony in data links.

[0028] The coordinates of each pressure sensing node are mapped to the corresponding position in the digital 3D geometric model. The 3D spatial coordinates (x, y, z) of each pressure sensing node are pre-stored in the digital 3D geometric model. The pressure values ​​of each pressure sensing node at the same time are then time-aligned. The value is directly assigned to the surface stress characterization value at that coordinate point at that moment. Five characteristic time points were selected during the pressurization process: 1 second, 5 seconds, 10 seconds, 15 seconds, and 20 seconds after the start of pressurization. For each characteristic time point, spatial interpolation was performed on the surface stress values ​​of all pressure sensing nodes. This interpolation uses an inverse distance weighted interpolation algorithm, which calculates the stress values ​​at any other location on the model surface based on the known stress values ​​at the sensor coordinates. The formula for inverse distance weighted interpolation is:

[0029] in: It is the interpolated stress at the point (x,y) to be determined. It is the stress characterization value of the j-th known pressure sensing node. It is a weighting coefficient and , It is the two-dimensional planar projection distance from the point (x, y) to the j-th pressure sensing node. This is the number of nearest neighbor sensor nodes participating in the interpolation, set here. This calculation generates a continuous stress distribution cloud map covering the entire outer surface of the engine cylinder head at each characteristic time point. The cloud map uses different colors to represent the stress distribution from 0 MPa to 2.5 MPa. The five stress distribution cloud maps generated in chronological order are overlaid and dynamically evolved. The analysis process involves arranging the five cloud maps in chronological order and calculating the stress change rate of each pixel between adjacent cloud maps, constructing a surface stress evolution map reflecting the entire process of pressure distribution from injection to stabilization. This map is a multidimensional data set containing time, space, and stress intensity dimensions.

[0030] In one embodiment of the present invention, a leakage airflow vector field traversing the sealing boundary of a component is constructed based on a micro-flow fluctuation sequence. The direction and magnitude of the gas velocity detected by each flow sensing node are analyzed from the micro-flow fluctuation sequence. A spatial network topology describing the connectivity of gas flow is constructed based on the relative positions of all flow sensing nodes in space. The gas flow path is traced and analyzed using the spatial network topology and the velocity direction of each node to infer the migration path of the gas inside the automotive component. The volumetric flow rate change of each segment along the inferred gas migration path is calculated by combining the velocity magnitude data. Locations with significant volumetric flow rate changes are vector-marked on the sealing boundary of the digital three-dimensional geometric model. These vector markings include the location coordinates of the leak point, the leakage flow rate intensity, and the airflow direction. All vector markings are summarized to form a leakage airflow vector field. The surface stress evolution map is compared region-by-region with a pre-stored standard stress distribution template. A standard stress distribution template corresponding to the model and testing pressure of the automotive component to be tested is invoked. This template records the reference stress value range of each region on the surface of the component under qualified sealing conditions at each stage of pressurization. In the surface stress evolution map, select the same characteristic time points as the standard stress distribution template and extract the actual stress values ​​at the same coordinate positions. Compare the actual stress value at each coordinate position at each characteristic time point with the reference stress value range. Mark continuous spatial regions where the actual stress value is consistently higher than the upper limit of the corresponding reference stress value range, defining them as stress anomaly concentration regions. Mark continuous spatial regions where the actual stress value is consistently lower than the lower limit of the corresponding reference stress value range, defining them as stress deficiency regions.

[0031] In practical implementation, taking the micro-flow monitoring of the engine cylinder head during pressurization testing as an example, this paper illustrates the construction of the leakage airflow vector field and the stress difference comparison process. In this implementation, an array of 32 miniature thermal mass flow sensors is arranged around the sealing grooves and suspected leakage paths of the engine cylinder head. The sensors collect data at a frequency of 200 Hz to form a micro-flow fluctuation sequence. The direction and magnitude of the gas flow velocity detected by each flow sensing node are analyzed from the micro-flow fluctuation sequence. The data processing unit reads the installation azimuth angle of each sensor in three-dimensional space and, combined with the voltage signal output by the sensor and the calibration curve, converts the signal into a gas flow velocity value with positive and negative signs. A positive value indicates that the airflow is in the direction the sensor is pointing, and a negative value indicates the opposite direction. Simultaneously, the flow velocity is calculated based on the voltage amplitude, with the unit being standard liters per minute (SPM).

[0032] A spatial network topology describing gas flow connectivity is constructed based on the relative spatial positions of all flow sensor nodes. This topology is represented as a graph, where each flow sensor node is a vertex with its three-dimensional coordinates. If the straight-line distance between two flow sensor nodes is less than a preset connectivity threshold of 10 cm, an undirected edge is established between the two vertices, indicating that gas may migrate directly between these two locations. This generates a connectivity matrix describing the adjacency relationships between nodes. The algorithm uses the spatial network topology and the flow velocity direction of each node to trace and analyze the gas flow path. Starting from all nodes displaying negative flow velocity values ​​(i.e., airflow inflow), the algorithm searches along the network topology edges for adjacent nodes with the same flow velocity direction, gradually tracing until a node with a positive flow velocity value (i.e., airflow outflow) is found. This allows the algorithm to infer possible migration paths of gas inside the engine cylinder head. In some embodiments, the algorithm records all node sequences from potential leak inflow points to outflow points as candidate paths.

[0033] Based on the inferred gas migration path and velocity data, the volumetric flow rate change of each segment along the path is calculated. For a candidate path, starting from the inflow node, the flow rate difference between adjacent nodes is calculated along the node sequence. The formula for calculating the flow rate difference is:

[0034] in: Indicates the first path The node and the first Changes in volumetric flow rate between nodes Indicates the first The volumetric flow rate measured by the flow sensors at each node. This represents the volumetric flow rate of the previous node, and the calculation is performed for all adjacent node pairs. The segments with changes significantly greater than the background noise level of 0.01 standard liters per minute were marked as effective leakage path segments. Vector annotations were applied to locations with significant volumetric flow rate changes on the sealing boundaries of the digital 3D geometric model. On the boundary surfaces of the cylinder head model, such as the cylinder head gasket sealing surface and the intake manifold mounting surface, arrow vectors were generated from the inflow point coordinates to the outflow point coordinates based on the calculated endpoint coordinates of the effective leakage path segments. These vector annotations included the location coordinates of the leakage point, the leakage flow rate intensity, and the airflow direction. The leakage flow rate intensity was taken from all values ​​along the path. The average value of all vector annotations is used to form a leakage airflow vector field covering the entire sealing boundary. The vector field is superimposed on the 3D model in the form of a vector graphic layer.

[0035] The surface stress evolution map is compared region by region with a pre-stored standard stress distribution template. The standard stress distribution template, corresponding to the engine cylinder head model and testing pressure, is invoked. This template records the baseline stress value range for each region of the engine cylinder head surface under qualified sealing conditions at five characteristic time points during pressurization. The baseline stress value range is represented by minimum and maximum values. Five characteristic time points identical to those in the standard stress distribution template are selected from the surface stress evolution map. The actual stress values ​​at the same coordinate positions defined in the standard template are extracted from the map data. These coordinate positions correspond to 256 regular grid sampling points pre-defined in the digital 3D geometric model. The actual stress value at each coordinate position at each characteristic time point is compared with the baseline stress value range. This comparison determines whether the actual stress value falls within the baseline minimum and maximum value range for the corresponding position and time point. Continuous spatial regions where the actual stress value consistently exceeds the upper limit of the corresponding baseline stress value range are marked as stress anomaly concentration areas. Continuous spatial regions where the actual stress value consistently falls below the lower limit of the corresponding baseline stress value range are marked as stress deficiency areas. These marking operations highlight the continuous grid regions in different colors on the 3D model. Optionally, refer to Table 1 for an example of stress value comparison at four sampling points on the cylinder head gasket seal surface at the 10th second of pressurization.

[0036] Table 1: Comparison of Stress Values ​​at Sampling Points at Characteristic Time Points

[0037] In some embodiments, the algorithm for marking continuous spatial regions is based on the connectivity of three-dimensional grids. If the comparison results of adjacent grid points are both "above the upper limit" or both "below the lower limit", these grid points are merged into the same region, and the three-dimensional bounding box coordinates of the stress anomaly concentration region and the stress deficiency region are finally output.

[0038] See Figure 3 This is a flow rate change analysis diagram of a leakage path, used to analyze the flow rate variation pattern of gas in the leakage path and locate the effective leakage section. The flow rate change between nodes 3 and 9 all exceed the 0.01 SLM threshold, representing the main gas leakage migration path. The flow rate change at node 8 reaches 0.041 SLM, which is the leakage peak of the entire path and the most critical leakage risk point. The flow rate change between the starting node 1 and the ending node 11 is 0, indicating that the leakage is mainly concentrated in the middle section of the path, with the two ends being the boundary points for gas inflow / outflow. This directly locates the core leakage channel for gas migration inside the cylinder head, providing precise location information for subsequent seal repair. Quantifying the leakage intensity distribution helps determine the severity of the leakage and guides the priority of process optimization. Spatially correlating with stress anomaly areas allows for further analysis of the causal relationship of "stress concentration / deficiency → seal failure → leakage".

[0039] In one embodiment of the present invention, the leaking airflow vector field is matched with a pre-stored standard leak field in terms of spatial morphology and intensity to locate the spatial coordinates of the leak and the dominant leak direction. A standard leak field corresponding to the model and testing pressure of the automotive component to be tested is invoked. This standard leak field describes the spatial distribution and intensity range of the expected leak vector on the sealing boundary of the digital three-dimensional geometric model under permissible micro-leakage conditions. Each leak vector in the leaking airflow vector field is compared with the expected leak vector at the corresponding spatial location in the standard leak field. When the intensity of the actual leak vector at a certain coordinate location exceeds the upper limit of intensity specified in the standard leak field, that coordinate location is identified as a potential leak point. The angle between the actual leak vector and the expected leak vector is calculated for all identified potential leak points. Potential leak points whose angle exceeds a preset tolerance or whose leak intensity exceeds the upper limit are selected, their spatial coordinates are determined as the spatial coordinates of the leak, and the direction of their actual leak vector is determined as the dominant leak direction. A comprehensive defect diagnosis report on the sealing integrity of the automotive component is generated by integrating the stress anomaly concentration area, stress deficiency area, and the spatial coordinates of the leak on the digital three-dimensional geometric model. The stress anomaly concentration area, stress deficiency area, and the spatial coordinates of the leak are highlighted on the digital three-dimensional geometric model. Analyze the spatial relationships between areas of abnormal stress concentration, areas of stress deficiency, and spatial coordinates. When the spatial coordinates of a leak occur at or near an area of ​​abnormal stress concentration, the diagnostic report will record the correlation as "High pressure concentration leading to seal failure." When the spatial coordinates of a leak occur at or near an area of ​​stress deficiency, the diagnostic report will record the correlation as "Insufficient stress leading to poor sealing." When the spatial coordinates of a leak occur neither in an area of ​​abnormal stress concentration nor in an area of ​​stress deficiency, but are associated with a leak path indicating the dominant leak direction, the diagnostic report will record the correlation as "Assembly or material defects causing leakage." Summarize all vector annotations, spatial relationships, and correlation records to generate a comprehensive defect diagnostic report including defect location, defect type, and cause.

[0040] In practical implementation, taking the leakage analysis stage of engine cylinder head airtightness testing as an example, this paper illustrates the matching analysis of the leakage airflow vector field and the standard leakage field, as well as the generation process of a comprehensive defect diagnosis report. In the practical implementation, the system calls a pre-stored standard leakage field corresponding to the currently tested engine cylinder head model and the 1.5 MPa testing pressure. The standard leakage field is stored in the form of database entries, describing the spatial distribution and intensity range of the expected leakage vector on the sealing boundary of each digital three-dimensional geometric model of the cylinder head under permissible micro-leakage conditions. The expected leakage intensity range for the middle of the cylinder head intake side mating surface is set to [0.001, 0.005] standard liters per minute, while the cylinder head gasket area is expected to be in an approximately zero-leakage state, with an upper limit of 0.0005 standard liters per minute.

[0041] Each leakage vector in the currently detected leakage airflow vector field is compared with the expected leakage vector at the corresponding spatial location in the standard leakage field. The comparison process involves traversing all marked vector points in the leakage airflow vector field and querying the standard leakage field database for the expected leakage intensity range at the same three-dimensional coordinate location. , ] and the unit vector of the expected leakage direction The intensity of the actual leakage vector at a certain coordinate location. Exceeding the upper limit of the strength specified in the standard leakage field When this occurs, the system identifies this coordinate location as a potential leak point. In some embodiments, for all identified potential leak points, the actual leak vector direction unit vector is calculated. Unit vector relative to the expected leakage vector direction The angle between the directions , direction angle Calculated using the vector dot product formula:

[0042] in: It is the inverse cosine function. This represents the vector dot product operation. It is the unit vector in the actual leakage direction. This is the unit vector representing the expected leakage direction. The angle between the directions is then selected. Exceeding the preset angle tolerance by 15 degrees or the actual leakage intensity Exceeding the intensity limit The system identifies potential leak points and determines their spatial coordinates as the final leak locations, while also identifying the direction of the actual leak vector corresponding to these points as the dominant leak direction. Optionally, Table 2 illustrates the matching analysis process and results for four coordinate points on the exhaust side of the engine cylinder head.

[0043] Table 2: Example Table of Matching Analysis between Leakage Gas Flow Vector Field and Standard Leakage Field

[0044] By integrating the identified stress concentration and stress deficiency areas, along with the spatial coordinates and dominant leakage direction determined in the previous steps, a comprehensive defect diagnosis report on the engine cylinder head seal integrity is generated. In practice, stress concentration and stress deficiency areas, as well as the spatial coordinates of the leakage occurrence, are highlighted on the digital 3D geometric model of the engine cylinder head using red, blue, and flashing yellow markers, respectively. The spatial relationship between the stress concentration and stress deficiency areas and the leakage point's spatial coordinates is analyzed, and the Euclidean distance from each leakage point's coordinates to the nearest stress concentration and stress deficiency area boundaries is calculated. When the leakage point's spatial coordinates are located at or adjacent to a stress concentration area, the report records the correlation as "high pressure concentration leading to seal failure," with "adjacent" defined as a distance of less than 5 mm. When the leakage point's spatial coordinates are located at or adjacent to a stress deficiency area, the report records the correlation as "insufficient stress leading to poor sealing." When the spatial coordinates of the leak point are neither in a stress concentration area nor a stress deficiency area, but are associated with the leak path indicated by the dominant leak direction, the correlation record in the diagnostic report is "leakage caused by assembly or material defects." The correlation is determined by the reverse extension line of the leak vector direction passing through a known assembly joint or material porosity marking area. All highlighted vector markings, spatial relationship analysis results, and correlation records are summarized, and a comprehensive defect diagnostic report document containing the three-dimensional coordinates of the defect location, a description of the defect type, and a hypothetical cause is generated according to a preset report template.

[0045] See Figure 4 This is a comprehensive defect diagnosis diagram for the sealing integrity of automotive components, used to visually display the spatial relationship between cylinder head surface defects, stress anomalies, and leak points. The background contour lines represent the severity of defects; the redder the color, the more severe the defect, and the bluer the color, the less severe the defect. Red squares represent areas of concentrated stress anomalies, where the actual stress consistently exceeds the standard upper limit, indicating a high risk of seal failure. Blue triangles represent areas of insufficient stress, where the actual stress consistently falls below the standard lower limit, indicating insufficient seal clamping force. Yellow stars represent leak points, located by the flow vector field to pinpoint the actual gas leak location. The areas of concentrated stress anomalies (red squares) are all distributed in areas of high defect severity, closely matching the spatial location of the large red high-defect area in the upper right corner. Leak points (yellow stars) are located on the upper yellow contour lines, adjacent to the edge of the areas of concentrated stress anomalies, consistent with the defect cause logic of "high pressure concentration leading to seal failure." Areas of insufficient stress (blue triangles) are concentrated in the transition zone between the left blue contour lines and the right red high-defect area, indicating uneven distribution of seal clamping force in these areas, a potential cause of poor sealing.

[0046] In one embodiment of the present invention, based on the feedback of detection results from the comprehensive defect diagnosis report and the process parameter optimization steps, common defect types and defect location distribution characteristics are extracted from the comprehensive defect diagnosis reports of the same batch of automotive parts. The common defect types and defect location distribution characteristics are correlated with the production process parameters of the automotive parts, including injection pressure, bolt tightening torque sequence, and welding temperature curve. When a correlation is identified between a defect pattern and the value range of a process parameter, adjustment suggestions for that production process parameter are generated. These adjustment suggestions are fed back to the production line control system for adaptive optimization of the process parameters of subsequent automotive parts. Before injecting the detection gas, a sensor array self-calibration and sealing benchmark establishment step is performed. A stable reference negative pressure environment is established in the detection chamber, and the initial reading benchmarks of all pressure and flow sensing nodes in the multi-physical quantity synchronous sensing array are recorded. A complete pressure testing process is performed on standard defect-free samples of the automotive parts to collect their standard dynamic pressure distribution sequence and standard micro-flow fluctuation sequence. A standard stress distribution template is constructed based on the standard dynamic pressure distribution sequence and stored in a database. A standard leakage field is constructed based on the standard micro-flow fluctuation sequence and stored in a database. The initial reading benchmark, standard stress distribution template, and standard leakage field will be used together as the comparison benchmark for subsequent testing of automotive parts of the same model.

[0047] In the specific implementation, for the mass production inspection of automotive engine cylinder heads, the steps of feedback on inspection results and optimization of process parameters based on comprehensive defect diagnosis reports are described. Comprehensive defect diagnosis reports of 100 engine cylinder head parts in the same batch are statistically analyzed to extract common defect types and defect location distribution characteristics. In the specific implementation, commonality analysis shows that 63 cylinder head comprehensive defect diagnosis reports contain records of "insufficient stress leading to poor sealing", and the spatial coordinates of 58 of these defect points are concentrated in the cylinder head gasket sealing surface area between the third and fourth cylinders. In addition, records of "assembly or material defects leading to leakage" frequently appear around the upper left bolt hole on the intake manifold mounting surface in the comprehensive defect diagnosis reports of another 25 cylinder heads.

[0048] The extracted common defect types and defect location distribution characteristics are correlated with the manufacturing process parameters of automotive parts. These parameters include injection pressure, bolt tightening torque sequence, and welding temperature profile. Essentially, the process parameter logs for these 100 engine cylinder heads are retrieved from the Manufacturing Execution System and aligned with the comprehensive defect diagnosis report for each cylinder head in terms of time sequence and serial number. When a correlation is identified between a defect pattern and the value range of a process parameter, adjustment suggestions for the manufacturing process parameters are generated. In some embodiments, the algorithm calculates the statistical correlation between the defect incidence rate and the process parameter values. For cases where the defect "insufficient stress leading to poor sealing" is concentrated in the area between cylinders three and four, analysis reveals that the third-stage setting value of the bolt tightening torque in this area... It shows a negative correlation with the defect incidence rate, and the correlation strength is... From the formula:

[0049] in: The sample size is 100 here. It is the first The actual value of the tightening torque in the third stage of the cylinder head. This is the average value of this torque for this batch. It is the first Defect markings for each cylinder head (1 for defects, 0 for no defects). It is the average defect incidence rate, calculated as follows. A value of -0.72 indicates a low level. The value is strongly correlated with a higher defect rate, thus generating an adjustment recommendation for the "third-step tightening torque of the bolt connection between the third and fourth cylinders," suggesting an increase in the torque value from the current 85 N·m to 92 N·m. This adjustment recommendation is understood to be fed back to the production line control system for adaptive optimization of the process parameters of subsequent automotive parts production. The adjustment recommendation is issued to the tightening gun controller in a structured command format via an industrial communication protocol, and the controller automatically updates the torque parameters at the corresponding workstation in the subsequent production sequence.

[0050] Before injecting the detection gas, a self-calibration and sealing baseline establishment procedure for the sensor array was performed to establish a stable reference negative pressure environment within the detection chamber. The pressure within the detection chamber was then evacuated and stabilized at -50 kPa using a vacuum pump. The readings of all pressure and flow sensing nodes in the multi-physical quantity synchronous sensing array were recorded at this point. All pressure sensing node readings were within the range of -49.8 to -50.2 kPa, and all flow sensing node readings were within the range of -0.002 to 0.002 standard liters per minute. These readings were recorded as the initial reading baseline. A complete pressure testing process was performed on standard defect-free samples of automotive parts. These standard defect-free samples were engine cylinder heads that had undergone rigorous offline testing to confirm the absence of any leaks. Standard dynamic pressure distribution sequences and standard micro-flow fluctuation sequences were collected at a standard test pressure of 1.5 MPa. Based on the standard dynamic pressure distribution sequence, a standard stress distribution template was constructed and stored in a database. The template recorded the pressure value range of the standard sample at five characteristic time points and 256 preset coordinate points. Based on a standard micro-flow fluctuation sequence, a standard leakage field is constructed and stored in a database. The standard leakage field records the amplitude and direction range of the background flow fluctuation of the standard sample at each sealing boundary coordinate point. In some embodiments, the initial reading benchmark, the standard stress distribution template, and the standard leakage field are used together as the comparison benchmark for subsequent testing of cylinder head components of the same model. Before each test, the system automatically loads the benchmark data set of the corresponding model. The real-time data collected during the test, after deducting the initial reading benchmark, is then compared and analyzed with the standard stress distribution template and the standard leakage field.

[0051] See Figure 5 This is an optimization analysis chart showing the relationship between bolt tightening torque and defect incidence rate. It quantifies the impact of bolt tightening torque on the sealing defect incidence rate, providing a basis for process adjustments on the production line. The current process defect incidence rate is 85N. The torque reaches its peak value (approximately 0.40) at point m, indicating that the current torque setting is precisely at the point of highest defect risk. The torque deviates from 85N. At m, the defect incidence rate drops rapidly, 80N At m, approximately 0.08, 95N The value is approximately 0.05 at m. It is recommended that the peak process defect rate be shifted to 92N. The value is around m (approximately 0.28), and the peak amplitude decreases significantly. This is within the recommended torque range of 92 N. At m, the defect incidence rate is approximately 0.27, which is significantly lower than that of the current process at 85N. 0.40 at m. Torque exceeds 92N. After m, the defect incidence rate continued to decline, 95N The torque is approximately 0.09 m, close to the optimal level of current technology. The torque is 85 N. m increased to 92N m can reduce the defect incidence rate from 0.40 to about 0.27, a reduction of about 32.5%.

[0052] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for intelligent airtightness testing of automotive parts based on intelligent sensing technology, characterized in that, The method includes: A digital three-dimensional geometric model is established for automotive parts of a predetermined model, and multiple pressure sensing nodes and flow sensing nodes are deployed at key sealing locations based on the digital three-dimensional geometric model to form a multi-physical quantity synchronous sensing array. The automotive parts to be tested are placed in a sealed testing chamber, and testing gas is injected into the testing chamber through a pressurization pipeline according to a preset pressurization curve. At the same time, the dynamic pressure distribution sequence and micro-flow fluctuation sequence of the inner and outer surfaces of the automotive parts are collected by the multi-physical quantity synchronous sensing array. Based on the dynamic pressure distribution sequence, a surface stress evolution map of automotive parts during the pressurization process is constructed, and based on the micro-flow fluctuation sequence, a leakage airflow vector field passing through the sealing boundary of the parts is constructed. By comparing the surface stress evolution map with the pre-stored standard stress distribution template region by region, areas of abnormal stress concentration and areas of stress deficiency are identified. The spatial shape and intensity matching analysis of the leaked airflow vector field and the pre-stored standard leak field are performed to locate the spatial coordinates of the leak and the dominant leak direction. By integrating the stress anomaly concentration area, stress deficiency area, and the spatial coordinates of the leakage occurrence and the dominant leakage direction, a comprehensive defect diagnosis report on the sealing integrity of automotive components is generated.

2. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 1, characterized in that, The process of establishing a digital three-dimensional geometric model for a predetermined model of automotive parts includes: The surface contour of standard automotive parts is scanned from all angles using a 3D laser scanning device to collect high-density point cloud data. The point cloud registration algorithm is used to stitch together the point cloud data obtained from multiple scans to form a point cloud model that completely covers the surface of the automotive parts. The point cloud model is triangulated to generate a surface mesh model representing the three-dimensional shape of the automotive parts. Based on the surface mesh model and combined with the design drawings of automotive parts, the key sealing structures and the preset placement positions of sensors are geometrically reconstructed and feature-annotated. The surface mesh model with feature annotations, along with the preset sensor node coordinates and connection relationships, are integrated into the digital three-dimensional geometric model.

3. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 2, characterized in that, The construction of the surface stress evolution map of automotive components during the pressurization process based on the dynamic pressure distribution sequence includes: From the dynamic pressure distribution sequence, the pressure readings of each pressure sensing node during the complete pressurization cycle are extracted in chronological order to form a multi-node pressure-time curve cluster with time as the horizontal axis and pressure as the vertical axis. The multi-node pressure-time curve cluster is synchronized and aligned to ensure that the data time base of all pressure sensing nodes is consistent. The coordinates of each pressure sensing node are mapped to the corresponding position in the digital three-dimensional geometric model, and the pressure value of each node at the same time is used as the surface stress characterization value of the coordinate point. Several characteristic time points are selected during the pressurization process. For each characteristic time point, the surface stress characterization values ​​of all pressure sensing nodes are spatially interpolated to generate a stress distribution cloud map covering the entire surface of the automotive parts at the characteristic time point. Multiple stress distribution cloud maps generated in chronological order are superimposed and dynamically evolved to construct a surface stress evolution map that reflects the distribution of pressure from injection to stabilization.

4. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 3, characterized in that, The process of synchronizing and aligning the multi-node pressure-time curve cluster includes: Obtain the timestamp of the opening of the pressure regulating valve in the pressurization pipeline, and define the timestamp as the pressurization start time; Identify the inflection point in the pressure-time curve of each pressure sensing node where the pressure first shows a significant increase; Calculate the time difference between the inflection point time and the pressurization start time for each node, and use it as the data transmission delay of the node; Based on the data transmission delay of each node, time shift compensation is performed on its pressure-time curve to align the pressure start-up time of all nodes with the pressurization start time.

5. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 4, characterized in that, The construction of the leakage airflow vector field across the sealing boundary of the component based on the micro-flow fluctuation sequence includes: From the micro-flow fluctuation sequence, the direction and magnitude of the gas flow velocity detected by each flow sensing node are analyzed. Based on the relative positions of all flow sensing nodes in space, a spatial network topology describing the connectivity of gas flow is constructed. By utilizing the spatial network topology and the flow velocity direction of each node, the gas flow path is traced and analyzed to predict the migration path of the gas inside the automotive parts. Based on the predicted gas migration path and the velocity data, the volumetric flow rate change of each segment along the gas migration path is calculated. On the sealed boundary of the digital three-dimensional geometric model, vector labels are made at locations where there are significant changes in volumetric flow rate. The vector labels include the location coordinates of the leak point, the leakage flow rate intensity, and the airflow direction. All vector labels are combined to form the leakage airflow vector field.

6. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 5, characterized in that, The step of comparing the surface stress evolution map with a pre-stored standard stress distribution template region by region to identify areas of abnormal stress concentration and areas of stress deficiency includes: The standard stress distribution template corresponding to the model of the automotive part to be tested and the test pressure is invoked. The standard stress distribution template records the reference stress value range of each area on the surface of the automotive part at each stage of pressurization under qualified sealing conditions. In the surface stress evolution spectrum, the same characteristic time point as the standard stress distribution template is selected, and the actual stress value at the same coordinate position is extracted; The actual stress value at each coordinate position at each characteristic time point is compared with the range of the reference stress value; A continuous spatial region where the actual stress value is consistently higher than the upper limit of the corresponding reference stress value range is defined as the stress anomaly concentration region. The continuous spatial region where the actual stress value is consistently lower than the lower limit of the corresponding reference stress value range is marked and defined as the stress deficiency region.

7. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 6, characterized in that, The step of performing spatial morphology and intensity matching analysis between the leaked gas flow vector field and a pre-stored standard leak field to locate the spatial coordinates of the leak and the dominant leak direction includes: The standard leakage field corresponding to the model of the automotive component to be tested and the test pressure is invoked. The standard leakage field describes the spatial distribution and intensity range of the expected leakage vector on the sealing boundary of the digital three-dimensional geometric model under permissible micro-leakage conditions. Each leakage vector in the leakage airflow vector field is compared with the expected leakage vector at the corresponding spatial location in the standard leakage field; When the intensity of the actual leakage vector at a certain coordinate location exceeds the upper limit of the intensity specified in the standard leakage field, that coordinate location is determined to be a potential leakage point; For all identified potential leak points, calculate the angle between the actual leak vector and the expected leak vector. Potential leak points whose directional angle exceeds a preset angle tolerance or whose leakage intensity exceeds the intensity limit are selected, their spatial coordinates are determined as the spatial coordinates of the leak occurrence, and the direction of their actual leakage vector is determined as the dominant leakage direction.

8. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 7, characterized in that, The system integrates the stress anomaly concentration area, stress deficiency area, and the spatial coordinates of the leakage occurrence and the dominant leakage direction to generate a comprehensive defect diagnosis report on the sealing integrity of automotive components, including: On the digital three-dimensional geometric model, the stress anomaly concentration area, the stress deficiency area, and the spatial coordinates of the leakage occurrence are highlighted and marked respectively. Analyze the spatial relationship between the stress anomaly concentration area, the stress deficiency area and the spatial coordinates; When the spatial coordinates of the leak occur are located at or adjacent to the area of ​​abnormal stress concentration, the correlation record in the diagnostic report is "high pressure concentration leads to seal failure"; When the spatial coordinates of the leakage are located at or adjacent to the stress deficiency area, the diagnostic report will record the correlation as "insufficient stress leading to poor sealing"; When the spatial coordinates of the leak are neither in the area of ​​abnormal stress concentration nor in the area of ​​stress deficiency, but are associated with the leak path indicated by the dominant leak direction, the correlation is recorded in the diagnostic report as "leak caused by assembly or material defects". By summarizing all vector annotations, spatial relationships, and correlation records, a comprehensive defect diagnosis report is generated, which includes the defect location, defect type, and cause.

9. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 8, characterized in that, The method also includes a step of feedback of detection results and optimization of process parameters based on the comprehensive defect diagnosis report: Analyze the comprehensive defect diagnosis reports of the same batch of automotive parts to extract common defect types and defect location distribution characteristics; The common defect types and defect location distribution characteristics are correlated with the production process parameters of automotive parts, including glue injection pressure, bolt tightening torque sequence, and welding temperature profile. When a correlation is identified between a defect pattern and a process parameter value range, adjustment suggestions for the production process parameters are generated. The adjustment suggestions are fed back to the production line control system for adaptive optimization of the process parameters of automotive parts produced subsequently.

10. The intelligent detection method for airtightness of automotive parts based on intelligent sensing technology according to claim 9, characterized in that, The method further includes a sensor array self-calibration and sealing baseline establishment step performed before the injection of the detection gas: A stable reference negative pressure environment is established in the detection chamber, and the initial readings of all pressure sensing nodes and flow sensing nodes in the multi-physical quantity synchronous sensing array are recorded at this time. For standard defect-free samples of the automotive parts, a complete pressure testing process was performed to collect their standard dynamic pressure distribution sequence and standard micro-flow fluctuation sequence. Based on the standard dynamic pressure distribution sequence, the standard stress distribution template is constructed and stored in the database; Based on the standard micro-flow fluctuation sequence, the standard leakage field is constructed and stored in the database; The initial reading benchmark, standard stress distribution template, and standard leakage field will be used together as the comparison benchmark for subsequent testing of automotive parts of the same model.

Images