A pretreatment method for metal bipolar plates in water-cooled hydrogen fuel cells

Abstract

This invention relates to the field of metal bipolar plate detection technology, a core component of hydrogen fuel cell stacks. It discloses a preprocessing method for water-cooled hydrogen fuel cell metal bipolar plates, including constructing a sealed topological mesh model and distinguishing between sensitive and non-sensitive topological regions; acquiring defect feature vectors using a vision component; classifying bipolar plates according to material properties; marking non-metallic soft defects as components to be compensated; monitoring contact stiffness and performing height-based flexible torque compensation for the components to be compensated using a servo electric compressor; further analyzing associated risk markers; establishing directional pressure difference vectors on both sides of the target boundary for tightened airtightness testing; and finally fusing pressure decay data, defect height, and peak contact stiffness to generate a quality judgment result. This invention achieves accurate identification, adaptive repair, and high-sensitivity detection of bipolar plate defects through visual topological mapping and multiphysics collaboration, effectively improving product yield and sealing reliability.

Description

Technical Field

[0001] This invention relates to the field of metal bipolar plate testing technology, a core component of hydrogen fuel cell stacks, specifically a pretreatment method for metal bipolar plates in water-cooled hydrogen fuel cells. Background Technology

[0002] Metal bipolar plates are a core component of hydrogen fuel cell stacks, involving complex processes such as precision stamping, welding, and coating. With the large-scale development of the hydrogen energy industry, ensuring the airtightness and surface quality of bipolar plates efficiently is crucial. Therefore, automated pretreatment technology integrating appearance defect identification and airtightness testing has become an indispensable part of the manufacturing production line.

[0003] Existing technologies typically employ a sequential workflow: first, surface scratches, foreign objects, and other defects are identified through automated optical inspection, and then the material is transported to the airtightness testing station. Airtightness testing often uses direct pressure or differential pressure methods, filling each cavity of the bipolar plate with gas at a constant pressure, and then monitoring the pressure decay curve to determine the overall sealing performance. This is currently the mainstream quality control method.

[0004] Although existing technologies have achieved automated screening, there are still shortcomings: First, existing visual inspection lacks a topological understanding of the flow field structure, often misjudging non-functional defects located in non-sealing critical areas as scrap, resulting in wasted costs; Second, the detection logic has difficulty in distinguishing the material properties of defects, causing soft adhesive overflow that could be eliminated through compression compensation to be directly scrapped, while hard foreign objects that are not removed are very likely to crush precision molds during compression; In addition, traditional full-cavity pressure equalization testing uses a fixed mode and fails to establish a directional extreme physical field environment for potential weak points discovered by vision, causing minor leaks after mechanical compensation to be easily masked by background noise, making it difficult to meet stringent sealing reliability requirements. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a pretreatment method for metal bipolar plates in water-cooled hydrogen fuel cells, which solves the problems of high false rejection rates caused by the lack of defect topology analysis and material identification, as well as insufficient sensitivity of airtightness testing to minute defects in existing technologies.

[0006] To achieve the above objectives, the present invention provides a pretreatment method for a water-cooled hydrogen fuel cell metal bipolar plate, comprising the following steps:

[0007] S1. Establish a global physical coordinate system and construct a sealed topology mesh model based on the digital design drawings of the bipolar plate, dividing the bipolar plate into sensitive topology regions and non-sensitive topology regions. S2, use the appearance inspection component to obtain the defect feature vector, determine the material properties based on the defect feature vector, classify the bipolar plate into standard parts, critical risk parts or parts to be compensated, and generate associated risk tags for the parts to be compensated. S3, based on the defect feature vector and the associated risk marker, estimate the test time and schedule the bipolar plate to the target airtightness detection component; S4, start the servo electric press, monitor the contact stiffness in real time during the pressing process, and perform flexible torque compensation based on the defect height for the part to be compensated; S5, analyze the associated risk marker, control the multi-channel gas valve island to establish a directional pressure difference vector on both sides of the associated target boundary, and collect the pressure decay data of the high-pressure side cavity within a set time. S6. The pressure attenuation data, the maximum vertical height in the defect feature vector, and the peak contact stiffness during the pressing process are integrated to generate the final quality judgment result based on the multi-dimensional parameter fusion judgment model.

[0008] Preferably, in step S1, constructing the sealed topology mesh model includes: Extract the geometric paths of the first isolation boundary between the hydrogen chamber and the air chamber, the second isolation boundary between the hydrogen chamber and the cooling water chamber, the third isolation boundary between the air chamber and the cooling water chamber, and the external sealing boundary; Calculate the shortest Euclidean distance from any coordinate point on the plane to any path in the set of boundary paths; Set a sealing strip width tolerance threshold. If the shortest Euclidean distance is less than the sealing strip width tolerance threshold, it is determined that the coordinate point is located in a sensitive topology area. If the shortest Euclidean distance is greater than or equal to the sealing strip width tolerance threshold, the coordinate point is determined to be located in a non-sensitive topology area.

[0009] Preferably, in step S2, the defect feature vector includes defect type, planar size, maximum vertical height, and optical reflectivity index; The classification includes: if the maximum vertical height is greater than zero and the optical reflectivity index is greater than a preset metal material reflectivity threshold, it is marked as a critical risk component; If the maximum vertical height is greater than zero and the optical reflectivity index is less than or equal to the reflectivity threshold of the metal material, it is marked as a component to be compensated. If the maximum vertical height is less than or equal to zero, it is marked as a standard part.

[0010] Preferably, in step S2, generating the associated risk marker includes: Map the defect coordinates of the component to be compensated to the sealing topology mesh model; If the defect coordinates are located in the sensitive topology region, traverse the set of boundary paths and select the boundary path with the smallest Euclidean distance as the associated target boundary. Generate the associated risk tag that includes the associated target boundary type.

[0011] Preferably, in step S3, the estimated test time includes: For bipolar plates with no defects or defects in non-sensitive areas, a standard test cycle is set. For bipolar plates with associated risk markers or markers to be compensated, the total time is set to include the time spent on servo stiffness detection and the extended time of the tightening test. The scheduling includes: real-time monitoring of the task queues of multiple parallel airtightness testing components, calculating the remaining waiting time of the current queue of each airtightness testing component, and selecting the component with the smallest sum of the remaining waiting time and the expected test time as the target component.

[0012] Preferably, in step S4, the target pressure compensation value is the sum of the mold closing pressure and the correction term under standard operating conditions; The correction term is determined by the product of the pressure compensation coefficient, the maximum vertical height value among all defects, and the elastic modulus of the sealing strip. The flexible torque compensation is used to eliminate the influence of non-metallic soft protrusions on the parallelism of the seal.

[0013] Preferably, in step S4, the flexible torque compensation includes: calculating a target pressure compensation value and controlling the servo electric press to load to the target pressure compensation value; the target pressure compensation value is the sum of the mold closing pressure under standard working conditions and the correction term; the correction term is determined by the product of the pressure compensation coefficient, the maximum vertical height value among all defects, and the elastic modulus of the sealing strip.

[0014] Preferably, in step S5, establishing the directional pressure difference vector includes: If the associated target boundary is the first isolation boundary, the hydrogen chamber is set to a tightened test high pressure, and the air chamber is set to a vacuum negative pressure; If the associated target boundary is the second isolation boundary, the hydrogen chamber is set as the tightened test high pressure, and the cooling water chamber is set as the vacuum negative pressure; If the associated target boundary is the third isolation boundary, the air cavity is set to the tightened test high pressure, and the cooling water cavity is set to the vacuum negative pressure; The directional pressure difference vector is obtained by subtracting the vacuum negative pressure from the tightened test high pressure.

[0015] Preferably, in step S6, the multi-dimensional parameter fusion judgment model adopts a dynamic leakage rate threshold; The calculation method of the dynamic leakage rate threshold is as follows: obtain the nominal leakage rate threshold of the standard part; calculate the difference of one minus the adjustment term; calculate the product of the nominal leakage rate threshold and the difference; In the calculation method, the adjustment term is determined by the product of the risk weighting coefficient and the height ratio; The height ratio is the ratio of the maximum vertical height value to the upper limit of the maximum defect height allowed by the process.

[0016] Preferably, in S6, the generation of the final quality determination result includes: Compare the measured pressure decay rate with the dynamic leakage rate threshold, and combine the peak contact stiffness; Only when the pressure decay rate is less than or equal to the dynamic leakage rate threshold and the peak contact stiffness is less than the preset safety stiffness threshold, determine that the bipolar plate is qualified; If it is determined to be qualified, write the defect location information and the target pressure compensation value into the product electronic label.

[0017] The present invention provides a pretreatment method for a water-cooled hydrogen fuel cell metal bipolar plate. It has the following beneficial effects: 1. By constructing a sealed topology grid model, the bipolar plate is divided into a sensitive topology area and a non-sensitive topology area. This partition mapping mechanism can effectively filter out surface defects in non-sealing key areas such as the flow channel ridge, avoiding misjudgment and scrapping caused by non-functional appearance defects; at the same time, the system only generates risk marks for defects located on the sensitive sealing boundary and triggers subsequent enhanced tests, thereby greatly reducing unnecessary over-detection while ensuring the sealing quality and improving the passing rate of the production line.

[0018] 2. By obtaining the optical reflectivity index and defect height through a vision sensor, the present invention realizes the precise classification of the defect material properties. For identified non-metal soft protrusions such as overflow glue, the system performs flexible torque compensation based on the defect height through a servo electric press, and uses the compression margin of the sealing gasket to eliminate the influence of small protrusions on the sealing performance, thereby saving some repairable products and reducing production costs; for metal hard protrusions, the system directly triggers shutdown protection, effectively preventing hard foreign objects from crushing precision molds or piercing the bipolar plate during the pressing process.

[0019] 3. This invention replaces the traditional full-cavity pressure equalization test with a directional differential pressure vector loading strategy that correlates with the target boundary. By applying high pressure on one side of the potentially risky isolation boundary and a vacuum negative pressure on the other side, the system establishes a maximized pressure gradient at the defect location. This physical amplification effect significantly increases the gas flow rate through tiny penetrating defects or interface gaps, making it easier for pressure sensors to detect even mechanically compensated minor leaks. This improves the ability to detect critical defects and ensures the sealing reliability of the bipolar plate under complex operating conditions. Attached Figure Description

[0020] Figure 1 This is a flowchart of the bipolar plate adaptive preprocessing method of the present invention; Figure 2 This is a schematic diagram of the construction of the sealed topology mesh model of the present invention; Figure 3 This is a flowchart of the multidimensional defect vectorization extraction and material safety preliminary screening process of the present invention; Figure 4 This is a flowchart of the dynamic workstation scheduling process based on task load according to the present invention; Figure 5 This is a flowchart of the adaptive servo pressing process based on contact stiffness feedback of the present invention. Figure 6 This is a flowchart of the differential pressure vector directional loading process based on topological boundaries according to the present invention; Figure 7 This is a flowchart of the multi-dimensional parameter fusion determination model of the present invention. Detailed Implementation

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] See attached document Figure 1 , Figure 1 This is a flowchart of an adaptive preprocessing method for bipolar plates based on visual topology mapping and multiphysics field collaboration according to an embodiment of the present invention. The present invention provides a preprocessing method for metal bipolar plates in water-cooled hydrogen fuel cells, which operates on a fully automated inspection line including a feeding assembly, an appearance inspection assembly, a dual-arm robotic arm, an airtightness inspection assembly, and a sorting and unloading assembly.

[0023] The fully automated inspection line achieves data interaction and collaborative control between various workstations through a central control system. The appearance inspection component is equipped with a vision sensor capable of depth measurement to acquire three-dimensional topographic data of the bipolar plate surface. The airtightness inspection component is equipped with a servo electric compressor and a multi-channel gas valve island. The servo electric compressor has dual closed-loop control capabilities for displacement and pressure, and the multi-channel gas valve island is independently connected to the hydrogen chamber, air chamber, and cooling water chamber of the bipolar plate. This pretreatment method includes the following steps: S1. The system selects the center of the first locating pin hole on the bipolar plate as the origin of the coordinate system, and the line connecting the centers of the first and second locating pin holes as the reference axis to establish a global physical coordinate system. Under this global physical coordinate system, the system reads the digital design drawing of the bipolar plate and extracts the geometric paths of the isolation boundaries between the hydrogen chamber and the air chamber, the hydrogen chamber and the cooling water chamber, the air chamber and the cooling water chamber, and the outer sealing boundary. The system defines the area where the above boundary paths are located as the sensitive topology region and the remaining flow channel region as the non-sensitive topology region, generating a sealing topology mesh model. This step ensures that the visual inspection data is strictly consistent with the mechanical positioning reference of the subsequent airtightness testing fixture.

[0024] S2. The visual inspection component scans the bipolar plate to identify surface defects. The system extracts the type, planar dimensions, height relative to the reference plane, and surface optical reflectivity features of each defect to generate a defect feature vector. The system determines the material properties of the defect based on the optical reflectivity features. If the defect height is greater than zero and the material property is determined to be a high-hardness metal, the system marks the bipolar plate as a critical risk component and directly terminates the subsequent pressing process. If the defect height is greater than zero and the material property is determined to be a low-hardness non-metal, the system marks the bipolar plate as a component requiring compensation. The system further maps the coordinates of the defect points to a sealed topology mesh model. If the defect point is located within a sensitive topology region, the system generates an associated risk marker and records the corresponding boundary type.

[0025] S3. Based on the defect feature vector and risk markers generated in step S2, the control system calculates the estimated testing time required for the current bipolar plate in the airtightness testing component. For bipolar plates without defects or with defects in non-sensitive areas, the standard testing time is set; for bipolar plates with risk markers or markers to be compensated, a long-cycle testing time including stiffness detection and tightening tests is set. The system monitors the task queue lengths of airtightness testing components A, B, and C in real time, and allocates the bipolar plates to the shortest current waiting time or a preset dedicated testing station according to the estimated testing time, and performs the transfer via a dual-arm robotic arm.

[0026] S4. After the bipolar plate is transferred to the designated airtightness detection component and mechanically positioned, the servo electric press executes its actions based on the height data in the defect feature vector. The servo electric press first moves to the contact critical point in position control mode, then switches to stiffness detection mode, moving downwards at a preset low speed and monitoring the rate of change of pressure with displacement in real time, i.e., the contact stiffness. If the contact stiffness exceeds the preset safety threshold, the servo electric press performs an emergency stop and reports an error. If the contact stiffness is within the safe range, the servo electric press calculates the target pressure compensation value based on the defect height and the preset elastic modulus of the sealing material, and performs the final pressing and pressure holding.

[0027] S5. After compression is complete, the gas path control system reads the associated risk markers. If a risk marker pointing to a specific isolation boundary exists, the system controls the multi-channel gas path valve island to apply differentiated pressures to each cavity. The system fills the cavity on one side of the isolation boundary with high-pressure test gas, while simultaneously controlling the cavity on the other side of the isolation boundary to connect to the atmosphere or evacuate to a vacuum, establishing the maximum pressure gradient vector on both sides of the physical boundary where the defect is located. For the remaining unrelated cavities, the system fills them with intermediate support pressure. Sensors collect pressure decay data of the high-pressure side cavity within a set time.

[0028] S6. The system acquires the pressure decay rate from airtightness testing, the defect size from visual inspection, the topological risk coefficient of the defect location, and the peak stiffness during the pressing process. The system evaluates the data based on a preset multi-dimensional parameter fusion judgment model. If the measured pressure decay rate and peak contact stiffness both meet the dynamic judgment criteria, the bipolar plate is deemed qualified, and the robot arm sorts it into the qualified product bin; if the quality index is less than the preset qualified threshold, the bipolar plate is deemed unqualified, and the robot arm sorts it into the unqualified product bin.

[0029] See attached document Figure 2 , Figure 2 This is a schematic diagram illustrating the construction of a sealed topology mesh model according to an embodiment of the present invention. In this embodiment, constructing a unified sealed topology model based on physical benchmarks includes the following sub-steps: S101, the center of the first locating pin hole on the metal bipolar plate is selected as the origin of the coordinate system. The line connecting the geometric centers of the first and second locating pin holes is selected as the positive direction of the X-axis. The normal direction perpendicular to the bipolar plate plane is selected as the positive direction of the Z-axis. A three-dimensional Cartesian coordinate system is established as the global physical coordinate system based on the above definitions. The appearance inspection component, airtightness inspection component, and robotic gripping position in the fully automated inspection line all adopt physical positioning references consistent with the global physical coordinate system. For the identification and center fitting algorithm of the locating pin holes, conventional image processing methods such as least squares fitting based on edge gradients are used.

[0030] S102, read the CAD design file of the bipolar plate, and the geometric features of the flow channel and sealing groove defined in the analytical formula. In the global physical coordinate system, extract the geometric centerline data of four key sealing paths: the first isolation boundary between the hydrogen chamber and the air chamber, the second isolation boundary between the hydrogen chamber and the cooling water chamber, the third isolation boundary between the air chamber and the cooling water chamber, and the outermost sealing boundary of the bipolar plate. Each of the above boundary paths is represented by a series of discrete coordinate points: ; In the formula, Indicates the first Boundary paths; Represents the coordinates of discrete points on the path; This represents the total number of discrete points along the path.

[0031] S103. Based on the extracted boundary path set, a sealed topology mesh model covering the entire plane of the bipolar plate is constructed. The boundary path set includes the first isolation boundary, the second isolation boundary, the third isolation boundary, and the outer sealing boundary. For any point on the plane, the shortest Euclidean distance from it to any path in the boundary path set is calculated. A sealing strip width tolerance threshold is set, which is determined based on 50% to 120% of the sealing strip width in the bipolar plate design drawings.

[0032] Define mapping function Binarize the region: ; In the formula, Represents coordinate points The region attribute value, Represents the set of boundary paths; This represents the sealing strip width tolerance threshold. When the region attribute value is 1, the point coordinates are determined to be located in a sensitive topology region, indicating that this region directly relates to the sealing isolation performance between different media cavities. When the region attribute value is 0, the point coordinates are determined to be located in a non-sensitive topology region, indicating that this region is a flow channel ridge or flow field region. This sealing topology mesh model is stored in the control system's database in digital mesh form, serving as the location mapping benchmark for visual defect risk assessment and airtight differential pressure loading in subsequent steps.

[0033] See attached document Figure 3 , Figure 3 This is a flowchart of multidimensional defect vectorization extraction and material safety preliminary screening according to an embodiment of the present invention. In this embodiment, performing multidimensional defect vectorization extraction and material safety preliminary screening includes the following sub-steps: S201 utilizes a 2D line scan camera and a 3D depth sensor integrated into the appearance inspection component to perform a full-frame scan of the bipolar plate. The 2D line scan camera acquires a grayscale image of the surface under coaxial lighting conditions, while the 3D depth sensor acquires surface height point cloud data. Defect regions on the bipolar plate surface are identified, and the multidimensional physical properties of these defect regions are extracted to construct a defect feature vector. ; In the formula, Indicates the first The defect feature vector of each defect; The defect type classification code includes scratches, foreign matter, excess glue, and dents; This represents the projected area of ​​the defect in the two-dimensional image. This indicates the maximum vertical height of the defect relative to the bipolar plate reference plane; a convex shape is a positive value, and a concave shape is a negative value. Represents the planar coordinates of the defect's geometric center in the global physical coordinate system. ; The optical reflectivity index represents the optical reflectivity of the defective surface, which is the ratio of the average gray value of the defective region to the average gray value of the standard highly reflective metal sample. For specific image segmentation and feature extraction algorithms, a connected component analysis method based on threshold segmentation is employed.

[0034] S202, to distinguish between hard metallic foreign objects and soft non-metallic residues, a preset reflectivity threshold for metallic materials is established. This reflectivity threshold is derived based on the grayscale response calibration of a standard metallic sample under the same light source environment. Based on the maximum vertical height and optical reflectivity index in the defect feature vector, a safety screening judgment logic is executed: ; In the formula, Indicates that for the first The results of a security screening for one defect; This indicates the reflectivity threshold of metallic materials; This indicates a life-threatening situation, corresponding to a highly reflective, protruding, hard metallic foreign object. This indicates a state requiring compensation, corresponding to diffuse reflection and protruding non-metallic soft foreign matter or excess adhesive; This indicates a safe passage status, corresponding to a pit or no raised defect.

[0035] S203, for defects identified as requiring compensation, their planar coordinates are read and substituted into the sealed topology mesh model for retrieval. Calculate the topology risk marker: ; In the formula, Indicates topology risk markers; This represents the region attribute value defined in step S1; The coordinates of the defect are in the plane.

[0036] When the topology risk flag value is 1, the defect is determined to be located in a sensitive topology region. At this point, the set of boundary paths is traversed. Calculate the planar coordinates of all paths. The Euclidean distance to each boundary path is calculated, and the boundary path with the smallest distance is selected as the associated target boundary. When the topology risk flag value is 0, the defect is determined to be located in a non-sensitive topology region. This step determines whether the defect is located in a critical area affecting sealing performance by calculating the spatial relationship between the defect coordinates and the sealing topology mesh model.

[0037] S204: Summarize the safety screening results of all defects. If any defect's safety screening result is a critical risk state, generate a rejection instruction and control the robotic arm to move the bipolar plate into the scrap bin. If all defects are in a safe pass state, generate a comprehensive inspection instruction marked as a standard test. If at least one defect is in a compensation state and has no critical risk state, generate a comprehensive inspection instruction marked as adaptive compensation, containing the defect feature vector and associated target boundary, and send this adaptive preprocessing instruction to the central control system to schedule the airtightness detection component.

[0038] See attached document Figure 4 , Figure 4 This is a flowchart of dynamic workstation scheduling based on task load according to an embodiment of the present invention. In this embodiment, dynamic workstation scheduling based on task complexity includes the following sub-steps: S301, analyze defect attributes and estimate the time required for airtightness testing.

[0039] The central control system receives the comprehensive testing instruction package generated in step S2. It reads the safety screening results and topology risk markers from the instruction package. Based on the defect status of the bipolar plate, it calculates the estimated testing time required for the bipolar plate in the airtightness testing component.

[0040] Establish a model for calculating the estimated test time: ; In the formula, Indicates the estimated test time; This indicates the standard test cycle, which is the total standard inflation, voltage stabilization, and testing time for electrode plates with no defects or defects in non-sensitive areas. This indicates the servo stiffness detection time, which corresponds to the additional time required for the servo press to perform low-speed stiffness detection. This indicates that when the stringent testing is extended, it corresponds to the additional time required for long-term high-pressure differential leakage monitoring of the risky components. Indicates the first Topological risk markers for each defect; Indicates the first The results of a security screening for defects.

[0041] The fully automated inspection line is equipped with three parallel airtightness testing components, labeled as station A, station B, and station C. The control system maintains the task queue status of each station in real time. For any station... Calculate the remaining waiting time for the current queue: ; In the formula, Indicates workstation The remaining waiting time; Indicates workstation The number of bipolar plates currently being tested; This indicates the remaining test time for the bipolar plate currently being tested. The remaining test time is obtained by subtracting the running time from the estimated test time of the bipolar plate. Indicates workstation The number of bipolar plates allocated but not yet tested in the cache queue; Indicates the first in the cache queue Estimated testing time for each bipolar plate.

[0042] S302, to achieve dynamic balance of production line cycle time, the optimal target workstation is selected based on the principle of minimum waiting time. For the bipolar plate currently to be assigned, the total completion time after assigning it to each workstation is calculated, and the workstation with the minimum total completion time is selected as the target workstation: ; In the formula, Indicates the selected target workstation index; This indicates the estimated testing time for the bipolar plates currently awaiting allocation. Once the target workstation is determined, the control system locks the allocation result, updates the corresponding workstation's buffer queue information, and sends a handling command to the dual-arm robotic arm. The robotic arm responds to the command, picking up the bipolar plates from the appearance inspection component and placing them into the positioning fixture at the target workstation. This scheduling strategy ensures that long-time-consuming high-risk parts and short-time-consuming standard parts are evenly distributed across multiple workstations, avoiding production line stoppages caused by congestion at a single workstation.

[0043] See attached document Figure 5 , Figure 5 This is a flowchart of adaptive servo pressing based on contact stiffness feedback according to an embodiment of the present invention. In this embodiment, performing adaptive servo pressing based on contact stiffness feedback includes the following sub-steps: S401, after the bipolar plate is transferred to the airtightness detection component and positioned, the servo electric compressor is started. The compressor first enters a rapid approach mode, moving downwards at a first speed to a preset safe distance from the surface of the bipolar plate. Subsequently, the compressor switches to a stiffness detection mode, slowly contacting the bipolar plate at a second speed, where the second speed is much smaller than the first speed. In stiffness detection mode, the force sensor and displacement grating ruler inside the compressor provide real-time feedback of pressure and displacement data at a high-frequency sampling rate.

[0044] S402, the control system calculates the contact stiffness of the interface between the pressure head and the bipolar plate in real time, which is defined as the differential of pressure with respect to displacement: The control system calculates the contact stiffness of the interface between the pressure head and the bipolar plate in real time, which is defined as the differential of pressure with respect to displacement: ; In the formula, Indicates contact stiffness; This represents the pressure data at the current moment; This represents the displacement data at the current moment. This indicates the sampling period. A safety stiffness threshold is set, based on the elastic modulus and structural stiffness of the normal sealing material of the bipolar plate. During the pressing process, if the contact stiffness is detected to be greater than the safety stiffness threshold, it indicates that the pressure head has come into contact with an incompressible hard foreign object or that the bipolar plate placement is severely misaligned. In this case, the servo press immediately executes an emergency stop and retraction action and issues a mechanical interference alarm.

[0045] S403, if the contact stiffness remains below the safety stiffness threshold throughout the entire contact process, it indicates a smooth contact process. In this case, if the current bipolar plate is marked as the component to be compensated, the control system reads the maximum vertical height from the defect feature vector. To eliminate the influence of non-metallic soft protrusions on parallelism, the target pressure compensation value is calculated: ; In the formula, This indicates the corrected final mold closing pressure; This indicates the mold closing pressure under standard operating conditions. This represents the pressure compensation coefficient, with a value ranging from 0.8 to 1.2, used to correct the nonlinear effect of the protrusion on the contact surface pressure; This represents the maximum vertical height value among all defects; This indicates the elastic modulus of the sealing strip. The servo press continues to apply and maintain pressure based on the calculated target pressure compensation value, ensuring that all sealing cavities reach the predetermined sealing specific pressure, providing stable mechanical boundary conditions for subsequent sealing tests.

[0046] See attached document Figure 6 , Figure 6This is a flowchart of differential pressure vector directional loading based on topological boundaries according to an embodiment of the present invention. In this embodiment, performing differential pressure vector directional loading based on topological boundaries includes the following sub-steps: The S501 airtightness testing component is equipped with a multi-channel gas path control matrix, which can independently control the pressure state between the hydrogen chamber, air chamber, cooling water chamber, and the external environment. The control system reads the associated target boundary information from the adaptive preprocessing instruction. If the associated target boundary is empty, it indicates that the bipolar plate is a standard component. According to the standard test logic, the multi-channel gas path valve island is controlled to perform standard sequence tests, establishing a preset standard pressure gradient between each chamber, such as hydrogen chamber > air chamber > cooling water chamber, to detect the sealing performance between chambers and to the outside. At the same time, the same test pressure is filled into the three chambers.

[0047] If a valid value exists at the associated target boundary, it indicates that the bipolar plate is a risky component. The system dynamically reconstructs the valve island's opening and closing state based on the type of the associated target boundary to establish the maximum pressure gradient on both sides of the target boundary. The specific gas path configuration logic is as follows: 1. If the associated target boundary is the first isolation boundary, the hydrogen chamber pressure is set to the tightened test high pressure, the air chamber pressure is set to vacuum negative pressure, and the cooling water chamber pressure is set to equilibrium pressure. 2. If the associated target boundary is the second isolation boundary between the hydrogen chamber and the cooling water chamber, the hydrogen chamber pressure is set to the tightened test high pressure, the cooling water chamber pressure is set to vacuum negative pressure, and the air chamber pressure is set to equilibrium pressure. 3. If the associated target boundary is the third isolation boundary between the air chamber and the cooling water chamber, the air chamber pressure is set to the tightened test high pressure, the cooling water chamber pressure is set to vacuum negative pressure, and the hydrogen chamber pressure is set to equilibrium pressure. 4. If the associated target boundary is an external sealing boundary, the pressures of the hydrogen chamber, air chamber, and cooling water chamber are all set to the tightened test high pressure, while the external environment remains at atmospheric pressure.

[0048] The tightened test high pressure is set to 1.2 to 1.5 times the rated working pressure of the bipolar plate; the vacuum negative pressure is set to -50 kPa to -80 kPa; and the balancing pressure is set to standard atmospheric pressure. The three pressures satisfy the following relationship: tightened test high pressure > balancing pressure > vacuum negative pressure.

[0049] S502, according to the above configuration logic, the gas path control unit drives the proportional valve and vacuum generator to operate. This creates a directional pressure differential vector on both sides of the associated target boundary. ; In the formula, Represents the directional pressure difference vector; This indicates that the high voltage test has been tightened. This indicates a vacuum negative pressure. The direction of this directional pressure difference vector is perpendicular to the sealing zone area where there is a potential defect. The system maintains this pressure state for a preset stabilization time to eliminate thermal effects caused by adiabatic compression.

[0050] S503, after entering the testing phase, the gas source valve is closed, and the system enters a pressure-holding state. A high-precision differential pressure sensor monitors the pressure changes in the high-pressure side cavity in real time. Due to the application of superimposed positive and negative ultimate pressure differences on both sides of the associated target boundary, the gas flow velocity at potential minute leak points will be physically amplified. Calculate the pressure decay rate: ; In the formula, Indicates the pressure decay rate; This indicates the pressure monitoring value at the start of the test cycle; This indicates the pressure monitoring value at the end of the test cycle; Indicates the start time of the test cycle; This indicates the end time of the test cycle. The pressure decay rate will serve as the core input parameter for the final quality assessment. Through the aforementioned directional loading strategy, the physical amplification effect of the pressure difference is utilized to improve the detection sensitivity for seal failures caused by minute through-hole defects or soft foreign objects.

[0051] See attached document Figure 7 , Figure 7 This is a flowchart of a multi-dimensional parameter fusion determination model according to an embodiment of the present invention. In this embodiment, performing multi-dimensional parameter fusion determination includes the following sub-steps: S601, the control system retrieves key process parameters generated by the bipolar plate in each process from the database. The extracted data includes: the maximum vertical height value among all defects calculated in step S4, the peak contact stiffness feedback from the servo press in step S4, and the pressure attenuation rate calculated in step S5. The system performs time-series alignment on the above data to construct a multi-dimensional state vector for final quality determination.

[0052] S602, to ensure the reliability of the seal after compensation, the system does not use a single fixed leakage rate threshold, but dynamically adjusts the judgment criteria based on the severity of defects on the bipolar plate surface. The greater the height of the defect, the higher the potential risk to the sealing structure, and therefore the more stringent the requirements for airtightness. A dynamic leakage rate threshold calculation model is established: ; In the formula, Indicates the dynamic leakage rate threshold; This indicates the nominal leakage rate threshold for standard components, which is set based on the rated power of the fuel cell stack and the allowable hydrogen loss. This represents the risk weighting coefficient, ranging from 0.2 to 0.5, and is used to adjust the severity of the judgment. This represents the maximum vertical height value among all defects; This indicates the maximum allowable defect height, which is set based on the geometric constraints of the bipolar plate flow channel ridge height and the amount of sealing compression.

[0053] S603, the system compares the measured pressure decay rate with the dynamic leakage rate threshold, and performs a logical AND operation on the peak contact stiffness to generate the final quality judgment result: ; In the formula, Indicates the quality assessment result; Indicates the pressure decay rate; Indicates the dynamic leakage rate threshold; This indicates the peak contact stiffness, which is the maximum value of the contact stiffness monitored during the pressing process; This indicates the safety stiffness threshold set in step S4.

[0054] If the quality assessment result is OK, the system generates a qualified product signal and writes the defect location information and compensation pressure value of the bipolar plate into the product's electronic tag to guide the fine-tuning of the preload force during subsequent fuel cell assembly. If the quality assessment result is NG, the system generates a scrap signal and controls the robotic arm to move the bipolar plate into the scrap channel. This judgment logic ensures that mechanically compensated risky components must meet stricter airtightness indicators before leaving the factory, thus guaranteeing the effectiveness of the compensation measures at the physical level.

[0055] This embodiment is applied to an automated batch testing production line for metal bipolar plates in hydrogen fuel cells. The object to be tested is a stamped SUS316L stainless steel bipolar plate with a single plate thickness of 0.1 mm, a flow channel depth of 0.4 mm, and external dimensions of 400 mm × 150 mm. The system configuration includes a high-precision 3D profile measuring instrument, a six-axis collaborative robotic arm, a three-station parallel airtightness testing bench, and a central control unit.

[0056] The system establishes a physical coordinate system by visually locating the first and second positioning holes. The CAD model of this bipolar plate is imported, and the sealing groove paths of the hydrogen chamber, air chamber, and cooling water chamber are marked as sensitive topology areas, while the flow channel ridge and collection area are marked as non-sensitive topology areas. A corresponding bitmap mesh model is generated from the database, with a mesh resolution set to 0.1mm × 0.1mm.

[0057] The visual inspection component performs a full-frame scan of the bipolar plate. The inspection algorithm outputs a list of defects, including their location coordinates, maximum height value, and optical reflectivity.

[0058] For defects with high reflectivity and high prominence, they are identified as metallic foreign objects or stamping burrs and marked as fatal risks.

[0059] For defects with low reflectivity and high prominence, they are identified as sealant residue or fiber dust and marked as requiring compensation.

[0060] The defect coordinates are compared with the topology mesh, and the subsequent tightening test logic is triggered only when the defect is located in a sensitive topology area.

[0061] For plates marked as requiring compensation, they are moved to an available airtight workstation. A servo-electric press descends to contact the plate surface. The system sets the contact stiffness safety threshold to 2000 N / mm. If the stiffness change rate during contact does not exceed the limit, additional pressure is calculated based on the defect height. The standard mold closing pressure is set to 5000 N. If the defect height is 0.03 mm, the calculated compensation value is 5000 + 1.2 × 0.03 × Eseal, and the actual loading pressure is adjusted to the range of 5150 N to 5200 N to compact soft foreign objects.

[0062] For risk markings involving hydrogen / air isolation boundaries, the valve island operates as follows: the hydrogen side is charged to 150 kPa, and the air side is evacuated to -80 kPa, establishing a 230 kPa cross-boundary pressure difference. The pressure holding time is set to 10 seconds.

[0063] The pressure decay rate is collected, and the leakage rate threshold is dynamically adjusted based on the defect height. For defect-free standard parts, the threshold is set to 15 Pa / s; for risky parts with soft defects of 0.05 mm, the threshold is automatically tightened to 12 Pa / s.

[0064] Ten typical sample data points were selected and recorded during the trial operation of the production line. The data cover different defect types, locations, and handling results.

[0065] Table 1: Record of Bipolar Plate Pretreatment Process Parameters and Quality Judgment Data

[0066] Based on the data in Table 1, and combined with the technical principles of this invention and actual test performance, the following conclusions are drawn: Analysis of sample B12 data revealed that for a soft adhesive overflow defect with a height of 0.05 mm, the pressure decay rate measured in the control group was as high as 28.6 Pa / s, leading to a system failure. This is because the traditional fixed clamping force cannot fully compress the soft protrusion, resulting in a microscopic gap between the sealing ring and the bipolar plate. In contrast, the example group, through adaptive servo pressing in step S4, applied targeted compensating pressure based on the defect feature vector, slightly increasing the peak contact stiffness to 9,350 N / mm, indicating a tighter pressing. The final measured pressure decay rate was 14.2 Pa / s, falling within the acceptable range. This result demonstrates that this method, through multi-physics field collaborative control, effectively solves the problem of false rejection of good products due to non-fatal soft defects.

[0067] Sample C07 simulated a working condition with 0.08mm hard metal particles. In the control group, the contact stiffness spiked to 18,540 N / mm during pressing, far exceeding the normal range, but due to the lack of a feedback mechanism, the system still forced the test to complete. Although its pressure decay rate of 10.5 Pa / s was acceptable, this constituted a false seal and could easily cause permanent damage to expensive airtight tooling. Data from the example group clearly showed that when the contact stiffness exceeded the set safety threshold, the servo press triggered protection logic, and the process status was recorded as process termination. This mechanism physically blocked the flow of critically dangerous components, avoiding unnecessary losses and equipment damage in subsequent processes.

[0068] Sample D22 contained a very small through-crack. Under the standard test pressure of 100 kPa in the control group, the pressure decay rate was 18.2 Pa / s, lower than the fixed threshold of 20 Pa / s, leading to missed detection. The example group utilized the differential pressure vector directional loading technique based on topological boundaries in step S5 to establish a superposition field of high-pressure positive pressure and vacuum negative pressure on both sides of the defect. The increase in physical pressure difference caused a non-linear increase in the gas flow velocity through the microcrack, amplifying the measured pressure decay rate to 35.8 Pa / s, thus accurately identifying it as non-compliant by the system. This confirms that the directional loading strategy based on sensitive topological regions can significantly improve the ability to capture weak signals.

[0069] In summary, this invention achieves spatial mapping between visual features and mechanical actions by constructing a sealed topological mesh model. Utilizing contact stiffness feedback and multi-physics collaborative loading, it enables differentiated processing of different types of defects while ensuring equipment safety, effectively reducing the false alarm rate of soft defects and significantly improving the detection rate of minute, hidden defects.

[0070] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A pretreatment method for metal bipolar plates in a water-cooled hydrogen fuel cell, characterized in that, Includes the following steps: S1. Establish a global physical coordinate system and construct a sealed topology mesh model based on the digital design drawings of the bipolar plate, dividing the bipolar plate into sensitive topology regions and non-sensitive topology regions. S2, use the appearance inspection component to obtain the defect feature vector, determine the material properties based on the defect feature vector, classify the bipolar plate into standard parts, critical risk parts or parts to be compensated, and generate associated risk tags for the parts to be compensated. S3, based on the defect feature vector and the associated risk marker, estimate the test time and schedule the bipolar plate to the target airtightness detection component; S4, start the servo electric press, monitor the contact stiffness in real time during the pressing process, and perform flexible torque compensation based on the defect height for the part to be compensated; S5, analyze the associated risk marker, control the multi-channel gas valve island to establish a directional pressure difference vector on both sides of the associated target boundary, and collect the pressure decay data of the high-pressure side cavity within a set time. S6. The pressure attenuation data, the maximum vertical height in the defect feature vector, and the peak contact stiffness during the pressing process are integrated to generate the final quality judgment result based on the multi-dimensional parameter fusion judgment model.

2. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 1, characterized in that, In S1, constructing the sealed topology mesh model includes: Extract the geometric paths of the first isolation boundary between the hydrogen chamber and the air chamber, the second isolation boundary between the hydrogen chamber and the cooling water chamber, the third isolation boundary between the air chamber and the cooling water chamber, and the external sealing boundary; Calculate the shortest Euclidean distance from any coordinate point on the plane to any path in the set of boundary paths; Set a sealing strip width tolerance threshold. If the shortest Euclidean distance is less than the sealing strip width tolerance threshold, it is determined that the coordinate point is located in a sensitive topology area. If the shortest Euclidean distance is greater than or equal to the sealing strip width tolerance threshold, the coordinate point is determined to be located in a non-sensitive topology area.

3. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 1, characterized in that, In S2, the defect feature vector includes defect type, planar size, maximum vertical height, and optical reflectivity index; The classification includes: if the maximum vertical height is greater than zero and the optical reflectivity index is greater than a preset metal material reflectivity threshold, it is marked as a critical risk component; If the maximum vertical height is greater than zero and the optical reflectivity index is less than or equal to the reflectivity threshold of the metal material, it is marked as a component to be compensated. If the maximum vertical height is less than or equal to zero, it is marked as a standard part.

4. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 3, characterized in that, In step S2, generating associated risk markers includes: Map the defect coordinates of the component to be compensated to the sealing topology mesh model; If the defect coordinates are located in the sensitive topology region, traverse the set of boundary paths and select the boundary path with the smallest Euclidean distance as the associated target boundary. Generate the associated risk tag that includes the associated target boundary type.

5. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 1, characterized in that, In S3, the estimated test time includes: For bipolar plates with no defects or defects in non-sensitive areas, a standard test cycle is set. For bipolar plates with associated risk markers or markers to be compensated, the total time is set to include the time spent on servo stiffness detection and the extended time of the tightening test. The scheduling includes: real-time monitoring of the task queues of multiple parallel airtight detection components, calculating the remaining waiting time of the current queues of each airtight detection component, and selecting the component with the minimum cumulative value of the remaining waiting time and the estimated test time consumption as the target component.

6. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 5, characterized in that, In S4, the target pressure compensation value is the cumulative value of the mold clamping pressure under standard working conditions and the correction term; The correction term is determined by the product of the pressure compensation coefficient, the maximum vertical height value among all defects, and the elastic modulus of the sealing rubber strip; The flexible torque compensation is used to eliminate the influence of non-metal soft protrusions on the sealing parallelism.

7. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 6, characterized in that, In S4, the flexible torque compensation includes: calculating the target pressure compensation value, and controlling the servo electric press to load to the target pressure compensation value; the target pressure compensation value is the cumulative value of the mold clamping pressure under standard working conditions and the correction term; the correction term is determined by the product of the pressure compensation coefficient, the maximum vertical height value among all defects, and the elastic modulus of the sealing rubber strip.

8. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 1, characterized in that, In S5, the establishment of the directional pressure difference vector includes: If the associated target boundary is the first isolation boundary, set the hydrogen chamber to the severe test high pressure and set the air chamber to the vacuum negative pressure; If the associated target boundary is the second isolation boundary, set the hydrogen chamber to the severe test high pressure and set the cooling water chamber to the vacuum negative pressure; If the associated target boundary is the third isolation boundary, set the air chamber to the severe test high pressure and set the cooling water chamber to the vacuum negative pressure; The directional pressure difference vector is obtained by subtracting the vacuum negative pressure from the severe test high pressure.

9. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 1, characterized in that, In S6, the multi-dimensional parameter fusion decision model uses a dynamic leakage rate threshold; The calculation method of the dynamic leakage rate threshold is: obtaining the nominal leakage rate threshold of the standard part; Calculating the difference of one minus the adjustment term; Calculating the product of the nominal leakage rate threshold and the difference; In the calculation method, the adjustment term is determined by the product of the risk weighting coefficient and the height ratio; The height ratio is the ratio of the maximum vertical height value to the upper limit of the maximum defect height allowed by the process.

10. The pretreatment method for the metal bipolar plate of a water-cooled hydrogen fuel cell according to claim 9, characterized in that, In S6, the generation of the final quality decision result includes: Comparing the measured pressure decay rate with the dynamic leakage rate threshold and combining the peak contact stiffness; Only when the pressure decay rate is less than or equal to the dynamic leakage rate threshold and the peak contact stiffness is less than the preset safety stiffness threshold, it is determined that the bipolar plate is qualified; If it is determined to be qualified, write the defect position information and the target pressure compensation value into the product electronic label.

Images