Full-automatic robot stacking and servo press-fitting forming process of phosphoric acid fuel cell stack

Abstract

The invention relates to the technical field of fuel cell manufacturing, and discloses a full-automatic robot stacking and servo press-fitting forming process of a phosphoric acid fuel cell stack, and the process comprises the following steps: executing base material initialization and quantitative acid coating; a visual detection unit is used for extracting coating features and calculating a coating distribution density index; a robot execution unit is controlled through multi-stage visual feedback to complete assembly initial positioning and visual servo fine positioning; when the stacking reaches a preset period, the central control unit calculates a corrected target displacement according to a coating distribution density index, and drives the servo press-fitting module to execute self-adaptive press-fitting based on feedforward compensation; and repeating the steps until the whole pile is finished. According to the method, the mapping relation between the optical characteristics of the coating and the press-fitting displacement is established, so that the microscopic difference of the coating thickness is effectively compensated, the fluctuation of the contact resistance is eliminated, the closed-loop monitoring of the whole-flow data is realized, and the assembly precision and the performance consistency of the galvanic pile are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of fuel cell manufacturing technology, specifically to a fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks. Background Technology

[0002] Phosphoric acid fuel cell stacks are typically composed of hundreds of bipolar plates and membrane electrode assemblies stacked alternately. The quality of this assembly directly determines the stack's output power and lifespan. During automated assembly, the phosphoric acid coating on the component surface not only acts as a proton transporter for the electrolyte but also fills the microscopic gaps at the contact surfaces to reduce contact resistance.

[0003] Existing assembly processes often employ fixed displacement parameters or constant pressure settings during the press-fitting stage. This control mode ignores the microscopic thickness differences and density fluctuations inherent in the chemical coating itself. When the coating is locally thicker or has a higher density, press-fitting with a preset fixed displacement can lead to excessive local contact stress, potentially damaging the porous microstructure of the component. Conversely, when the coating is thinner, the same press-fitting displacement cannot establish sufficient contact pressure, resulting in increased interlayer contact resistance and consequently affecting the overall voltage consistency and electrochemical performance of the fuel cell stack.

[0004] The manufacturing of fuel cell stacks involves the continuous vertical stacking of numerous flexible or brittle sheet components, requiring extremely high spatial alignment precision. Traditional industrial robot gripping operations rely primarily on taught coordinates or mechanical limits for positioning, lacking dynamic correction mechanisms for random deviations in incoming material positions and accumulated errors in robotic arm movements. As the number of stacked layers increases, minute single-layer alignment errors accumulate and amplify layer by layer, causing the geometric center of the fuel cell stack to shift or tilt, reducing the structural stability and packaging reliability of the finished product.

[0005] Existing automated production lines have limitations in data interaction and process monitoring. There is often a lack of data linkage between the upper-level production management system and the lower-level actuators based on individual characteristics, making it difficult to achieve real-time binding of component physicochemical data with the final pressing process results. During production, if abnormal fluctuations in pressing force or displacement occur, the system often cannot quickly trigger the underlying safety interlock logic based on real-time feedback data, leading to defective products flowing into subsequent processes. Furthermore, due to the lack of a full lifecycle process data chain, it is difficult to accurately trace and analyze the causes of finished product failures. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks. This process solves the problem that existing press-fitting processes use fixed displacement or pressure parameters, which cannot adapt to the thickness differences and distribution density fluctuations of the phosphoric acid coating at the microscale, resulting in poor consistency of interlayer contact resistance of the stack components or damage to the microstructure due to local overpressure.

[0007] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks, which solves the technical problems of contact resistance fluctuations and poor stacking height consistency caused by uneven coating distribution and component size tolerances during stack assembly.

[0008] This invention provides a fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks. This process is based on a fully automated robotic stacking and servo press-fitting system comprising a central control unit, a robotic execution unit, a vision inspection unit, and coating and weighing units. The process includes the following steps: Step S100, Substrate Initialization and Weight Detection: The central control unit controls the electronic balance to record the substrate mass of the component before coating. Step S200, Quantitative Acid Coating and Coating Amount Verification: The coating and weighing unit coats the component surface with phosphate solution, and the central control unit verifies the actual net weight of the coating based on the difference between the mass after coating and the mass of the substrate.

[0009] Step S300, Visual Inspection and Feature Quantization of Coating Quality: The central control unit uses the visual inspection unit to acquire coating images, extract effective coating areas, and calculate the coating distribution density index to characterize the coating distribution trend. Step S400, Component Grasping and Initial Positioning: The visual inspection unit identifies the initial position deviation of the component, and the robot execution unit grasps the component based on the initial position deviation.

[0010] Step S500, Visual Servo Precision Positioning: The robot execution unit moves the component to the hovering position. The central control unit calculates the precision positioning deviation vector based on the bottom features of the component obtained by the visual inspection unit, and controls the robot execution unit to fine-tune the stacking so that the center of the component is aligned with the center of the stack.

[0011] Step S600, Adaptive servo press-fitting based on feedforward compensation: When the number of stacked layers reaches the preset number of press-fitting cycles, the central control unit retrieves the coating distribution density index corresponding to all components within the number of press-fitting cycles and calculates and generates a corrected target displacement. The servo press-fitting module of the robot execution unit applies vertical surface pressure to the stacked body according to the corrected target displacement.

[0012] Step S700, Complete stack cyclic assembly: After the pressing is completed, the central control unit determines whether the current total number of layers has reached the total number of layers of the stack; if the total number of layers of the stack has not been reached, steps S100 to S600 are repeated.

[0013] In this embodiment of the invention, the specific process for calculating the coating distribution density index is as follows: The central control unit performs Gaussian filtering for noise reduction and threshold segmentation on the coating image to extract the effective coating area. The central control unit calculates the sum of the gray values ​​of all pixels within the effective coating area, divides the sum of gray values ​​by the total pixel area of ​​the theoretically coated area, and obtains the coating distribution density index. The central control unit binds the coating distribution density index to the unique identification code of the component and stores it in the production data queue.

[0014] In this embodiment of the invention, step S300 further includes a process for determining the qualification of the coating: the central control unit divides the effective coating area into multiple sub-grids and calculates the local variance of pixel grayscale in each sub-grid to determine uniformity. The central control unit calculates the ratio of the total number of pixels in the effective coating area to the total number of pixels in the theoretical coating area to determine coverage. When the uniformity or coverage is lower than a preset threshold, the central control unit issues a rejection command.

[0015] In this embodiment of the invention, visual servoing precision positioning is achieved as follows: The robot execution unit suspends the component above the upward-looking lens of the visual inspection unit, and the visual inspection unit captures an image of the bottom of the component. The central control unit calculates the image domain deviation of the component's actual pose relative to the theoretical stacked pose. The central control unit uses calibration parameters to convert the image domain deviation into motion compensation commands in the robot's base coordinate system. The robot execution unit drives the end effector to perform pose fine-tuning based on the motion compensation commands.

[0016] In this embodiment of the invention, the calculation of the corrected target displacement is based on the fluctuation of coating distribution data. The central control unit calculates the arithmetic mean of the coating distribution density index of all components within the current pressing cycle to obtain the cycle-averaged density index. The central control unit calculates the difference between the cycle-averaged density index and the reference density index, multiplies this difference by a displacement compensation coefficient, and adds the product to the nominal pressing displacement to obtain the corrected target displacement. The displacement compensation coefficient characterizes the amount of physical displacement adjustment corresponding to a unit density index deviation.

[0017] In this embodiment of the invention, the servo pressing module controls the pressing quality through pressure feedback. The servo pressing module drives the rigid pressure head to press down until the actual displacement equals the corrected target displacement. During the pressure holding phase, the central control unit monitors the actual contact pressure value through a force sensor. When the actual contact pressure value exceeds the preset target pressure range, the central control unit generates an alarm signal and terminates the pressing action.

[0018] In this embodiment of the invention, during the cycle of step S700, the robot execution unit is reset, the coating and weighing unit applies an acidic coating to the center position of the back side of the stack, and the central control unit records the timestamp of the coating operation to monitor the electrolyte exposure time.

[0019] In this embodiment of the invention, the initial positioning of the component is achieved by capturing an image of the loading position using a top-down view lens of the vision inspection unit. The central control unit uses a feature matching algorithm to identify the component model and an edge detection algorithm to calculate the initial coordinate deviation of the component center on the conveyor belt plane. The robot execution unit adjusts the gripping coordinates based on the initial coordinate deviation.

[0020] In this embodiment of the invention, the lower-level computer of the central control unit collects pressure and displacement data from the servo pressing module in real time. When an abnormality is detected in the pressure or displacement data, the lower-level computer cuts off the enable signal of the servo driver. After completing the pressing cycle, the lower-level computer sends the process result data back to the upper-level computer, which then merges and stores the process result data with the coating distribution density index.

[0021] In this embodiment of the invention, the robot execution unit performs gripping and pressing via a composite end effector. The robot execution unit drives the composite end effector to descend, and the vacuum adsorption module first contacts the stack. As the composite end effector continues to descend, the vacuum adsorption module, under pressure, retracts via an elastic floating mechanism, causing the rigid pressure head of the servo pressing module to protrude and contact the stack, thereby applying pressure to the stack.

[0022] This invention provides a fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks. It offers the following advantages: 1. This invention achieves the technical effect of dynamically correcting the servo press-fit displacement based on the actual thickness trend of the coating on the component surface by extracting the coating distribution density index and using it as a feedforward variable for the press-fit parameters. This method effectively compensates for the differences in physicochemical characteristics at the material level, reduces the inconsistency in contact resistance caused by coating thickness fluctuations, and improves the performance stability of the fuel cell stack after assembly.

[0023] 2. This invention utilizes a multi-level visual feedback system composed of a second and a third imaging device, combined with calibration parameters, to convert image domain pose deviations into motion compensation quantities for multi-joint industrial robots. This precise positioning logic eliminates component loading deviations and cumulative errors in robotic arm movement, ensuring the spatial positioning accuracy of multi-layer thin-film components during the stacking process.

[0024] 3. The central control unit of this invention collects feedback data from force sensors and position detection sensors in real time via a lower-level computer, and works with a higher-level computer to index and bind production data packages and process results. This control logic, based on dual closed-loop monitoring of pressure and displacement, provides a complete unit quality traceability file, effectively identifying and preventing abnormal states during the assembly process. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0026] 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.

[0027] Please see the appendix Figure 1 This invention provides a fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks.

[0028] This invention provides a fully automated robotic stacking and servo pressing system for phosphoric acid fuel cell stacks, including a robot execution unit, a vision inspection unit, a coating and weighing unit, and a central control unit.

[0029] The robot actuator comprises a multi-jointed industrial robot. The multi-jointed industrial robot has six degrees of freedom, and a composite end effector is rigidly connected to its end flange. The composite end effector integrates a vacuum adsorption module and a servo press-fit module. The vacuum adsorption module is connected to a vacuum generator via an air circuit and is mounted on the composite end effector via a flexible floating mechanism, allowing the vacuum adsorption module to retract under axial pressure, thereby exposing the rigid pressure head of the servo press-fit module.

[0030] The servo press-fit module includes a servo electric cylinder and a transmission mechanism. A force sensor is connected in series in the force transmission path of the servo press-fit module. The force sensor is used to monitor the contact pressure value in real time during the press-fit process. The composite end effector is equipped with a position detection sensor, which is used to provide real-time feedback on the absolute displacement of the composite end effector in the Z-axis direction. The position detection sensor uses a high-precision grating ruler or absolute encoder with a resolution at the micrometer level.

[0031] The visual inspection unit includes a first imaging device, a second imaging device, a third imaging device, and a fourth imaging device. The first imaging device is mounted directly above the coating station, with its optical axis perpendicular to the coating plane. The first imaging device is used to capture grayscale images of the coated component surface. Grayscale image The data is transmitted to the central control unit for calculating the coating distribution density index. .

[0032] The second imaging device is positioned above the component loading area. Its field of view covers the component's gripping position, and it is used to acquire the initial positional deviation of the component on the conveyor belt plane. The third imaging device is fixedly mounted on a bracket on the side of the stacking platform, with its lens set vertically upwards. When the multi-joint industrial robot's gripping component hovers above the third imaging device, the device captures precise positioning feature points on the bottom of the component, which are used to calculate the deviation vector between the component's actual gripping pose and the theoretical stacking pose. The fourth imaging device is mounted above the stacking platform to acquire image information of the upper surface of the stack.

[0033] The coating and weighing unit includes an automatic dispensing machine and an electronic balance. The electronic balance is connected to the central control unit and is used to read the substrate mass of the component before and after coating. and quality after coating The central control unit is based on the quality of the substrate. and quality after coating Calculate the difference in net coating weight The dispensing nozzle of the automatic dispensing machine is driven by a multi-axis module. The automatic dispensing machine is used to apply a quantitative amount of phosphate solution to the surface of components.

[0034] The central control unit adopts an architecture that includes a host computer and slave computers. The host computer is connected to the first imaging device, the second imaging device, and the third imaging device. The host computer is equipped with an image processing module for performing Otsu thresholding and edge detection, and calculating the coating distribution density index. The lower-level computer connects to the multi-joint industrial robot, servo press-fit module, force sensor, and position detection sensor via industrial real-time Ethernet. The central control unit operates based on the coating distribution density index calculated by the upper-level computer. Generate the corrected press-fit target displacement It also controls the servo press-fit module to perform actions.

[0035] For the electrical connections and communication protocols between the robot execution unit, vision inspection unit, coating and weighing unit and central control unit, those skilled in the art can implement them using standard industrial specifications, which will not be elaborated here.

[0036] An embodiment of the present invention provides an automated assembly method for a phosphoric acid fuel cell stack, comprising the following steps: Step S100: Perform substrate initialization and weight detection. The cathode or anode assembly is transported to the weighing station. The central control unit controls the electronic balance to perform a tare and zeroing operation, and records the substrate weight before coating. .

[0037] Step S200: Perform quantitative acid coating and coating amount verification. An automatic dispensing machine coats the component surface with a phosphate solution, and the central control unit controls the coating amount. After coating is completed, the central control unit acquires the post-coating quality of the component. And according to the formula Calculate the actual net weight of coating If the actual net weight of the coating Within the preset tolerance range, the component proceeds to the next process.

[0038] Step S300: Perform visual inspection and feature quantization of the coating quality. The host computer acquires the coating image using the first imaging device and performs noise reduction and segmentation processing. The host computer determines whether the uniformity and coverage of the coating are qualified, and calculates the coating distribution density index for qualified components. Coating distribution density index It is used to characterize the distribution trend of the coating in the microscopic dimension and is stored in the production data queue.

[0039] Step S400: Perform component grasping and initial positioning. The second imaging device identifies the component model and initial position. The host computer calculates the initial deviation vector of the component relative to the theoretical grasping position. The multi-jointed industrial robot adjusts its grasping posture based on the initial deviation vector to pick up components.

[0040] Step S500: Perform visual servo precision positioning. The multi-joint industrial robot moves the component to a position above the third imaging device and hovers it. The host computer calculates the deviation vector between the actual pose and the theoretical stacking pose based on the visual features of the component's bottom. This information is then converted into motion compensation commands in the robot's base coordinate system. The multi-joint industrial robot performs fine-tuning movements to precisely place components on the stacking platform.

[0041] Step S600: Perform adaptive servo press-fitting based on feedforward compensation. The central control unit monitors the number of stacked layers. When the number of stacked layers reaches the preset number of press-fitting cycles... At that time, the central control unit retrieves the coating distribution density index of all components within this cycle. The average value is then calculated. Based on this average value and a preset compensation model, the central control unit calculates and corrects the target displacement. The servo press-fit module drives the rigid press head to adjust the target displacement. Vertical pressure is applied to the stack to compensate for differences in coating thickness.

[0042] Step S700: Perform complete stack cyclic assembly and process iteration. After pressing, an automatic dispensing machine applies a coating to the back of the sub-stack. The central control unit determines whether the current total number of layers has reached the total number of layers for the stack. If the condition is not met, the system will cycle through steps S100 to S600; if the condition is met, the system will perform full-pile compaction and terminate the process.

[0043] Step S800 involves data interaction and full-process closed-loop control. Step S800 is integrated throughout the preceding steps, with the host computer and slave computer exchanging data via a real-time communication link. The host computer is responsible for image processing and coating distribution density index. and correct target displacement The lower-level computer is responsible for the calculation and distribution of data; the lower-level computer is responsible for the motion control of the multi-joint industrial robot, the force / position closed-loop control of the servo press module, and the abnormal safety interlock, and transmits the process data back to the upper-level computer for full life cycle traceability.

[0044] The central control unit performs the coating quality visual inspection and feature quantification described in step S300. Step S300 specifically includes sub-steps S310 to S350.

[0045] Step S310: Image acquisition and preprocessing are performed. Driven by a trigger signal, the first imaging device, under illumination from a highly uniform coaxial light source, captures an image of the component surface located at the inspection station, obtaining the original grayscale image. The host computer reads the original grayscale image. .

[0046] Step S311, the host computer processes the original grayscale image. Perform Gaussian filtering. The host computer uses a pre-calibrated Gaussian kernel function and the original grayscale image... Perform convolution operations to generate a smoothed image after denoising. .

[0047] Step S320: Perform coating region segmentation and extraction. The host computer uses the Otsu's algorithm (maximum inter-class variance method) to process the smoothed image. Perform global threshold optimization. The algorithm calculates the optimal grayscale threshold. This maximizes the inter-class variance between the foreground and background.

[0048] Step S321, the host computer bases the grayscale threshold on the optimal grayscale threshold. Generate a binarized mask image If pixel grayscale value Greater than or equal to the optimal grayscale threshold If the coating area is positive, it is marked as 1; otherwise, it is marked as 0. The host computer defines the effective coating area. For binarized mask image The set of all pixels marked as 1.

[0049] Step S330, perform the qualification determination of the coating quality. Step S330 specifically includes the uniformity determination and the coverage determination.

[0050] Step S331, the host computer divides the effective coating area into rectangular sub-grids. The host computer calculates the local variance of the pixel grayscale within each rectangular sub-grid and calculates the mean value of the local variances of all rectangular sub-grids, denoted as the overall uniformity index . If the overall uniformity index exceeds the preset non-uniformity threshold, it is determined that the coating uniformity is unqualified.

[0051] Step S332, the host computer counts the total number of pixels within the effective coating area . The host computer uses the formula to calculate the coverage percentage , where is the number of pixels corresponding to the theoretical coating area calculated according to the component design drawing. If the coverage percentage is less than 80%, it is determined that the coating coverage is unqualified. For the components determined to be unqualified, the central control unit issues a rejection instruction.

[0052] Step S340, perform the calculation and binding of the coating distribution density index. For the components determined to be qualified in Step S330, the host computer calculates the coating distribution density index . This calculation is based on the gray-scale - thickness linear mapping principle under coaxial illumination, that is, the local thickness of the phosphate coating is positively correlated with the gray-scale integral value of the pixels at that place.

[0053] Step S341, the host computer calculates the coating distribution density index according to the following formula: In the formula, represents the pixel coordinates in the smoothed image ; is the effective coating area extracted in Step S321; is the pixel gray-scale value at this coordinate; is the total pixel area of the theoretical coating area.

[0054] Step S350, the central control unit generates a production data packet, which includes the unique identification code of the component, the actual coating net weight and the coating distribution density index . The central control unit pushes the production data packet into the first-in, first-out queue of the production execution system. When the component flows to the press-fitting station, the subsequent servo control algorithm calls the coating distribution density index in the production data packet It is used to calculate the displacement compensation amount of press fitting.

[0055] The specific mathematical principles of Gaussian filtering, Otsu thresholding, and variance calculation involved in the above steps are well-known techniques in the field of image processing and will not be elaborated here.

[0056] The central control unit executes steps S400 and S500 in the overall process flow, and eliminates component material arrival position deviation and cumulative error of robotic arm movement through a multi-level visual feedback mechanism.

[0057] The central control unit performs the component grasping and initial positioning described in step S400, which specifically includes sub-steps S410 to S430.

[0058] Step S410: Perform image acquisition and feature recognition at the loading position. When the photoelectric sensor detects that the component has arrived at the loading position, the second imaging device captures a top-view image of the component. The host computer preprocesses the top-view image and extracts key feature points from the image using a scale-invariant feature transform algorithm or an accelerated robust feature algorithm. The host computer performs feature matching between the key feature points and pre-stored standard templates in the database to verify whether the component model is consistent with the current production work order.

[0059] Step S420: Perform initial grasping deviation calculation. The host computer identifies two diagonal positioning holes or edge corners on the component, and calculates the actual center coordinates of the component in the conveyor belt plane coordinate system using a sub-pixel edge detection algorithm. and rotation angle The host computer reads the preset theoretical capture coordinates. And calculate the initial deviation vector. .

[0060] Step S430: Perform a grasping action based on deviation compensation. The central control unit will input the initial deviation vector. Send to the robot controller. The robot controller will then send the initial deviation vector. The coordinates of the gripping point, superimposed on the teaching coordinates, are used to drive the multi-joint industrial robot to move to the corrected position via inverse kinematics calculation. The vacuum adsorption module of the composite end effector is activated, picking up the component and lifting it to a safe height.

[0061] The central control unit performs the visual servo fine positioning described in step S500, which specifically includes sub-steps S510 to S550.

[0062] Step S510: Perform upward-looking imaging and hovering alignment. The multi-jointed industrial robot carrying the components moves to the preset hovering coordinates directly above the third imaging device. At this moment, the bottom surface of the component is directly facing the lens of the third imaging device.

[0063] Step S520: Perform precise feature extraction and deviation calculation. The third imaging device captures a grayscale image of the bottom of the component. The host computer extracts the component's edge features or pre-made alignment marks from the grayscale image and calculates the component's actual pose in the image coordinate system. ,in For pixel coordinates, This is the rotation angle within the image plane. The host computer will then display the actual pose. Alignment with theoretical centering pose Compare and generate an image domain deviation vector. : Step S530: Perform coordinate system transformation and motion compensation calculation. The host computer uses a pre-calibrated hand-eye transformation matrix to convert the image domain deviation vector... Converted into a motion compensation matrix for the robot's end effector.

[0064] Step S531: The host computer calculates the corrected robot target pose matrix according to the following formula. : In the formula, The current hovering position of the multi-joint industrial robot in the base coordinate system Homogeneous transformation matrix; This is the rotation calibration matrix of the third imaging device coordinate system relative to the robot's base coordinate system; For image domain bias vector The constructed deviation transformation matrix.

[0065] Step S532, Deviation Transformation Matrix The specific definitions are as follows: In the formula, and These are the horizontal and vertical scaling factors from the image pixel coordinate system to the physical space coordinate system, respectively. The scaling factors are determined by the calibration parameters of the third imaging device. , , The image domain deviation component is calculated in step S520.

[0066] Step S540: Perform six-degree-of-freedom fine-tuning motion. The central control unit will calculate the target pose matrix. This is converted into joint angle commands for a multi-joint industrial robot. The multi-joint industrial robot drives the servo motors of each joint, causing the composite end effector to move the component in translation and rotation in space until the theoretical axis of the component center coincides with the stack center.

[0067] Step S550: Perform the vertical stacking action. After completing precise positioning, the multi-jointed industrial robot maintains its horizontal posture and descends vertically along the Z-axis. When the Z-axis height fed back by the position detection sensor reaches the preset placement height, the vacuum adsorption module releases the components.

[0068] The central control unit executes step S600 in the overall process flow, whereby it uses the coating distribution density index as a feedforward variable to calculate the press-fit displacement parameters. The central control unit performs the adaptive sequential press-fit based on feedforward compensation as described in step S600, which specifically includes sub-steps S610 to S640.

[0069] Step S610: Perform pressing cycle determination and data aggregation. The central control unit has an internal stacking layer counter. Each time the multi-jointed industrial robot completes the release action of one layer of components, the stacking layer counter is updated. Perform the accumulation operation. The central control unit determines the current stacking layer counter. Is the value equal to the preset number of pressing cycles? .like Less than The central control unit skips the pressing action. If equal The central control unit triggers the press-fit command and locks the most recently added data in the production data queue. Group data. Central control unit extracts. The corresponding coating distribution density index in the data set (in Values ​​range from 1 to ), and calculate Coating distribution density index The arithmetic mean, denoted as the periodic average density index. .

[0070] Step S620: Perform the correction calculation for the target press-fit displacement. The central control unit reads the reference density index stored in the pre-stored memory. and nominal press-fit displacement Nominal press-fit displacement To ensure that the Z-axis coordinate value corresponds to the predetermined contact pressure for a standard thickness component, this embodiment sets the Z-axis coordinate value to increase as the servo pressure head descends. The central control unit uses a preset linear compensation model to calculate the corrected target displacement for this pressing action. .

[0071] Step S621, correct the target displacement The calculation formula is as follows: In the formula, This is the displacement compensation coefficient, expressed in millimeters per grayscale unit. Displacement compensation coefficient The K value, obtained through process experiment calibration, characterizes the displacement adjustment required for a unit density index deviation.

[0072] Step S630: Perform multi-stage servo pressing action. The servo pressing module receives the corrected target displacement. And drive the servo electric cylinder to move.

[0073] The servo-electric cylinder first approaches the surface of the stack at a first velocity; At a preset distance from the theoretical contact surface, the servo electric cylinder switches to a second speed, which is less than the first speed. The servo electric cylinder then presses down at a constant speed at the second speed until the actual displacement is detected by the position detection sensor. Equal to correcting the target displacement ; The servo electric cylinder maintains its current position and enters the pressure holding stage.

[0074] Step S640: Perform pressure monitoring and reset. During the pressure holding phase, the central control unit continuously collects the actual contact pressure value fed back by the force sensor. The central control unit compares the actual contact pressure value. With respect to the preset target pressure range If the actual contact pressure value Within the target pressure range, the central control unit records the press-fit data. If the actual contact pressure value... If the pressure exceeds the target range, the central control unit generates an alarm signal. The preset pressure holding time is then reached. Then, the servo pressing module drives the rigid pressing head to retract at the third speed, while the central control unit activates the stacking layer counter. Reset to zero.

[0075] The central control unit performs the whole-stack cyclic assembly described in step S700. Step S700 includes sub-steps S710 to S730.

[0076] Step S710: Perform press-fit reset and interlayer interface detection. The servo press-fit module completes the pressure holding action and drives the rigid press head to retract to a safe height. The multi-joint industrial robot moves to the avoidance position. The host computer uses the fourth imaging device to take a picture of the uppermost surface of the current stacked component assembly. The host computer receives the image and executes the defect detection algorithm to confirm that there are no foreign objects or crushing defects on the surface.

[0077] Step S720: Perform coating on the back of the stack. The automatic dispensing machine moves above the stacking platform.

[0078] In step S721, the automatic dispensing machine performs quantitative coating on the central area of ​​the uppermost surface of the stacked component assembly, and the central control unit controls the coating target amount to be 500mg.

[0079] Step S722: After the coating is completed, the central control unit records the timestamp of this coating operation to monitor the electrolyte exposure time.

[0080] Step S730: Perform the entire stack cycle judgment and termination. The central control unit reads the preset total number of stack layers. The central control unit internally maintains a global total layer counter. .

[0081] Step S731: The central control unit compares the global total number of layers counter. With the total number of layers in the fuel cell stack .

[0082] like Less than The central control unit will use the stacking layer counter described in step S610. Reset to zero and jump the control program back to step S100 to start the substrate initialization and front-side loading process of the next layer component.

[0083] like equal The central control unit controls the multi-joint industrial robot to perform the stacking and compaction action and sends a stacking completion signal.

[0084] The central control unit performs the data interaction and full-process closed-loop control described in step S800. Step S800 specifically includes sub-steps S810 to S840.

[0085] Step S810: Perform data acquisition and feature binding. The host computer receives data streams from the first imaging device, the second imaging device, and the third imaging device through the image acquisition interface. The host computer performs algorithm processing on the data streams to generate process parameters. The process parameters include: the initial deviation vector obtained by the second imaging device. The precise positioning deviation vector obtained by the third imaging device and the coating distribution density index acquired and calculated by the first imaging device. The host computer reads the substrate mass output by the electronic balance via a serial communication interface. With the quality after coating The host computer establishes a data structure in memory, indexes and binds process parameters with the unique identifier of the current component, forming a production data package to be processed.

[0086] Step S820: Perform data communication and command issuance. A communication link is established between the host computer and the slave computer via the ADS communication protocol or TCP / IP socket.

[0087] During the component capture phase, the host computer will use the initial deviation vector. Write the specified register address to the lower-level machine. The lower-level machine reads the data from the register address and converts it into the base coordinate system offset of the multi-joint industrial robot.

[0088] During the press-fitting preparation stage, the host computer retrieves the coating distribution density index of all components within the corresponding period according to the logic described in step S610. And calculate the periodic average density index. The host computer uses the calculation formula described in step S621 to generate the corrected target displacement. And will correct the target displacement It is sent directly to the lower-level machine as the final position command for the servo press-fitting action.

[0089] Step S830: Perform real-time motion control and low-level closed-loop feedback. The lower-level computer is connected to the robot controller, the servo driver of the servo press module, and the force sensor via a real-time industrial Ethernet bus.

[0090] During servo press fitting, the lower-level computer sends position control words to the servo driver at millisecond intervals. The lower-level computer reads the actual contact pressure value fed back by the force sensor via the bus. The lower-level machine internally runs a PID control algorithm to determine the actual contact pressure value. The force is compared with the preset target pressure, and the output torque of the servo drive is adjusted in real time to form a force control inner loop.

[0091] The actual displacement fed back by the position detection sensor As a feedback signal of the position loop, it ensures that the rigid pressure head of the servo press-fit module accurately reaches the corrected target displacement. This forms a location-controlled outer ring.

[0092] Step S840: Perform data recording and exception handling. After completing the stacking and pressing operations for each cycle, the lower-level computer transmits the final pressing height, maximum contact pressure, and holding time data for the current production cycle back to the upper-level computer. The upper-level computer then compares the transmitted data with the coating distribution density index recorded in step S810. The data is merged to generate a complete quality profile for each individual component and stored in the industrial database. If the lower-level machine detects the actual contact pressure value in step S830... or actual displacement If an over-limit anomaly occurs, the lower-level computer immediately cuts off the enable signal of the servo driver, stops the movement of the multi-joint industrial robot and the servo pressing module, and sends a fault code to the upper-level computer.

[0093] The specific message format definition and PID control parameter tuning of the above communication protocol can be configured by those skilled in the art based on the specific hardware selection manual. This is well-known technology in the field and will not be elaborated here.

[0094] 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 fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks, executed by a fully automated robotic stacking and servo press-fitting system, the system comprising a central control unit, a robotic execution unit, a vision inspection unit, and a coating and weighing unit; the process includes the following steps: Step S100, Substrate initialization and weight detection: The central control unit controls the electronic balance recording component to record the substrate quality before coating; Step S200, Quantitative acid coating and coating amount verification: The coating and weighing unit apply phosphate solution to the surface of the component, and the central control unit verifies the actual net weight of the coating based on the difference between the mass after coating and the mass of the substrate. Step S300, Visual Inspection and Feature Quantization of Coating Quality: The central control unit uses the visual inspection unit to acquire coating images, extract effective coating areas, and calculate the coating distribution density index to characterize the micro-distribution trend of the coating. Step S400, Component Grasping and Initial Positioning: The visual detection unit identifies the initial position deviation of the component, and the robot execution unit grasps the component based on the initial position deviation; Step S500, Visual Servo Precision Positioning: The robot execution unit moves the component to the hovering position, and the central control unit calculates the precision positioning deviation vector based on the bottom features of the component obtained by the visual detection unit, and controls the robot execution unit to perform fine-tuning stacking; Step S600, Adaptive servo press-fitting based on feedforward compensation: When the number of stacked layers reaches the preset number of press-fitting cycles, the central control unit retrieves the coating distribution density index corresponding to all components within the number of press-fitting cycles and calculates and generates a corrected target displacement; the servo press-fitting module of the robot execution unit applies vertical surface pressure to the stacked body according to the corrected target displacement; Step S700, Complete Stack Cyclic Assembly: After the pressing is completed, the central control unit determines whether the current total number of layers has reached the total number of layers of the stack; if the total number of layers of the stack has not been reached, steps S100 to S600 are executed cyclically.

2. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claim 1, characterized in that, In step S300, the specific method for calculating the coating distribution density index is as follows: The central control unit performs Gaussian filtering for noise reduction and threshold segmentation on the coating image to extract the effective coating area. The central control unit calculates the sum of gray values ​​of all pixels in the effective coating area, divides the sum of gray values ​​by the total area of ​​pixels in the theoretical coating area, and obtains the coating distribution density index. The central control unit binds the coating distribution density index with the unique identification code of the component and stores it in the production data queue.

3. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claim 2, characterized in that, Step S300 also includes a coating conformity assessment step: The central control unit divides the effective coating area into multiple sub-grids and calculates the local variance of pixel grayscale in each sub-grid to determine uniformity. The central control unit calculates the ratio of the total number of pixels in the effective coating area to the total number of pixels in the theoretical coating area to determine coverage. When the uniformity or the coverage is lower than a preset threshold, the central control unit issues a rejection command.

4. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claim 1, characterized in that, In step S500, the specific execution process of the visual servo precision positioning is as follows: The robot execution unit suspends the component above the upward-looking lens of the vision detection unit; the vision detection unit captures an image of the bottom of the component, and the central control unit calculates the image domain deviation of the actual pose of the component relative to the theoretical stacking pose; the central control unit uses the hand-eye calibration matrix and the deviation transformation matrix to convert the image domain deviation into motion compensation commands in the robot base coordinate system; the robot execution unit drives the end effector to perform six-degree-of-freedom fine adjustments according to the motion compensation commands until the center of the component coincides with the center of the stack.

5. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claim 1, characterized in that, In step S600, the specific logic for calculating and generating the corrected target displacement is as follows: The central control unit calculates the arithmetic mean of the coating distribution density index of all components within the current cycle, denoted as the cycle average density index; the central control unit calculates the difference between the cycle average density index and the density index of the standard reference sample; the central control unit multiplies the difference by a preset displacement compensation coefficient, and adds the product to the nominal press-fit displacement to obtain the corrected target displacement; wherein, the displacement compensation coefficient is used to characterize the amount of physical displacement adjustment required per unit density index deviation.

6. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claim 5, characterized in that, Step S600 also includes a pressure monitoring step: the servo press module drives the rigid press head to press down at a preset speed until the actual displacement fed back by the position detection sensor is equal to the corrected target displacement; During the pressure holding phase, the central control unit monitors the actual contact pressure value in real time through a force sensor; when the actual contact pressure value exceeds the preset target pressure range, the central control unit generates an alarm signal and terminates the operation.

7. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claim 1, characterized in that, In step S700, before the cyclic execution steps S100 to S600, the following steps are also included: the robot execution unit is reset; the coating and weighing unit applies an acid coating to the center position of the back side of the stack; and the central control unit records the timestamp of the coating operation to monitor the electrolyte exposure time.

8. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claim 4, characterized in that, In step S400, the specific method for component capture and initial positioning is as follows: The visual inspection unit captures images of the loading position using its top-view lens, and the central control unit uses a feature matching algorithm to identify the component model. The central control unit calculates the initial coordinate deviation of the component center on the conveyor belt plane using a sub-pixel edge detection algorithm. The robot execution unit adjusts the gripping point based on the initial coordinate deviation and uses a vacuum adsorption module to pick up the component.

9. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claims 1-8, characterized in that, The process also includes a closed-loop control step throughout the entire process: The lower-level computer of the central control unit collects the pressure data and displacement data of the servo pressing module in real time; when the pressure data or displacement data is detected to be abnormal, the lower-level computer cuts off the enable signal of the servo driver and sends a fault code to the upper-level computer of the central control unit. After each pressing cycle is completed, the lower computer sends the process result data back to the upper computer, which then merges and stores the process result data with the coating distribution density index.

10. The fully automated robotic stacking and servo press-fitting process for phosphoric acid fuel cell stacks according to claims 1-8, characterized in that, In step S600, the servo press-fit module performs the press-fit action, including the following mechanical engagement steps: The robot execution unit drives the composite end effector to descend, and the vacuum adsorption module on the composite end effector first contacts the stack; As the composite end effector continues to descend, the vacuum adsorption module retracts under the action of axial reaction force through the elastic floating mechanism; the retraction action of the elastic floating mechanism causes the rigid pressure head of the servo press module to protrude from the vacuum adsorption module, and the rigid pressure head directly applies the vertical surface pressure to the stack.

Images