Digital assembling system and method for fuel cell stack

Abstract

The invention relates to a digital assembly system and method for a fuel cell stack, the system is composed of a physical space, a digital space and an interaction layer for information interaction between the physical space and the digital space, the physical space comprises a physical layer, and the digital space comprises a functional layer, a model layer and a data layer. Data of an actual galvanic pile assembly process in a physical layer are collected in real time and uploaded to a digital space through an interaction layer; the digital space establishes digital models corresponding to fuel cell stack components and assembly modules in the physical space, and data layer information is processed, so that efficient prediction and real-time updating of the stack assembly state are realized; calling a result of the digital model by a functional layer in the digital space, and formulating a galvanic pile assembly strategy; and the physical space receives the assembly strategy from the digital space through the interaction layer to complete stack assembly. Compared with the prior art, the scientificity and intellectualization of stack assembly strategy formulation are improved, the performance of the fuel cell is improved, and the cost loss is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of fuel cells, and in particular relates to a digital assembly system and an assembly method for a fuel cell stack. Background Art

[0002] In recent years, fuel cells, as a device that directly converts the chemical energy of fuel and oxidant into electrical energy, have received increasing attention from the international and social communities and are considered to be the preferred clean and efficient power generation technology in the 21st century. Among them, proton exchange membrane fuel cells (PEMFC) have been widely used in many fields such as transportation, aerospace, and power storage due to their advantages such as high energy density, high energy conversion efficiency, clean and environmentally friendly, and low operating temperature.

[0003] PEMFC is mainly composed of bipolar plates, membrane electrode assemblies, sealing elements and end plates assembled in a stacked manner; in view of the stacked structural characteristics, in the battery assembly component link, it is necessary to consider the design characteristics of key components and the influence of manufacturing errors; in the battery assembly link, it is necessary to consider the influence of assembly process and assembly errors; in the battery working link, it is necessary to consider the influence of operating conditions; the entire assembly process has the characteristics of long process flow, multi-physical field coupling, and large number of parameters; the quality of assembly will have a direct impact on battery performance and service life. At present, the assembly methods of battery stacks mainly include manual assembly, semi-automatic equipment assembly and fully automatic equipment assembly. With the introduction of auxiliary mechanical equipment, the assembly speed of batteries has been significantly improved, but there are still problems such as: insufficient utilization of assembly process data, dependence on experience in determining assembly process, inability to dynamically display assembly process, inability to evaluate assembly quality in real time, and real-time adjustment of assembly process. These problems directly limit the further improvement of battery assembly quality.

[0004] After a literature survey of the prior art, it was found that the Chinese invention patent CN111463466B discloses an adaptive fuel cell assembly structure with composite functions. The structure uses limit strips to position parts, and uses adjustable end plates and buffer end plates to adjust the battery thickness and assembly pressure respectively, thereby improving the assembly efficiency and the uniformity of the stack assembly force during the assembly process. The Chinese invention patent application CN108539240A discloses an assembly device and method for a fuel cell stack. The scheme uses a press to press the fuel cell stack, and then bends the steel strip and welds it to the stack through a bending mechanism and a welding module, thereby greatly improving the efficiency of manual assembly of the fuel cell stack. The Chinese invention patent CN114927740B discloses a fuel cell stack and its compression assembly method. The scheme uses tensioning components and straps to reduce the thickness of the fuel cell stack without affecting the output power of the stack, thereby improving the stack accuracy.

[0005] Chinese invention patent CN112836391B discloses a fuel cell stack assembly method based on model-driven. Based on the model-driven method, a stack component processing quality evaluation model, a stack module assembly quality evaluation model, and a stack overall assembly quality evaluation model are established for the stack assembly performance requirements. The stack components and stack modules are graded, and the optimal assembly sequence of the stack components and stack modules in the total stack is obtained. The continuous assembly of the fuel cell stack is realized by combining a multi-station assembly production line. This technology can obtain the optimal arrangement scheme of components and stack modules based on the comprehensive evaluation index of the stack module and the comprehensive evaluation index of the stack, and realize the optimization of the assembly quality of the fuel cell stack product. However, because the information is transmitted one-way between the modules, it is impossible to realize the dynamic and real-time display of information during the assembly process. Therefore, the performance monitoring and improvement of the fuel cell needs to be further improved.

[0006] Chinese invention patent CN117217031B discloses an intelligent production method and system for fuel cell stacks, including: determining first production data of each component of the stack to be produced according to overall performance requirement data and stack performance requirement data; obtaining various performance data, attribute data and historical working data of the sample fuel cell system and determining the first matching relationship model and the first influence relationship model between each sample component of the sample stack accordingly; modifying the first production data according to the first matching relationship model and the first influence relationship model to obtain second production data; producing each component of the stack to be produced according to the second production data; assembling each component into a stack and testing it. Although this technology can intelligently control the production of the stack, and fully evaluate the performance matching relationship and work influence relationship between each component to produce each component of the stack more finely, thereby obtaining a high-quality stack, this technology does not fully explore the physical characteristics of the stack components and related process parameters during the assembly of the fuel cell stack, and cannot evaluate in real time, the assembly process cannot be adjusted in real time, and it is impossible to perform real-time performance optimization analysis and strategy modification formulation for the entire process flow of stack assembly based on the main assembly parameters.

[0007] The above-mentioned disclosed fuel cell stack assembly devices and schemes have not yet analyzed the mechanism of the effects of assembly component processing errors, component stacking sequence, and stack press displacement on the stack assembly quality. They cannot fully utilize the data during the assembly process, cannot evaluate the assembly quality in real time, and cannot adjust the assembly process in real time. There is a problem of battery performance loss due to errors in the assembly process and repeated debugging. Summary of the invention

[0008] The purpose of the present invention is to provide a digital assembly system and assembly method for fuel cell stacks in order to overcome the defects of the above-mentioned prior art, which makes full use of the data in the assembly process to construct an assembly digital model, realizes the integration and management of the model based on the unified modeling method, and realizes the real-time interaction between the actual process of fuel cell assembly and the digital model of fuel cell assembly through geometry and data reconstruction, so that the digital model of fuel cell stack assembly can accurately reflect the multi-physical field state of the stack during the actual space assembly process. The digital model uses the data of the system to not only accurately reproduce the assembly process of the fuel cell in the digital space, but also efficiently predict the assembly results and optimize the stack assembly process based on this, and provide guidance for the actual stack assembly process, thereby improving the quality and efficiency of the stack assembly, improving the performance of the fuel cell, and reducing the cost loss of the battery due to assembly process errors and repeated debugging.

[0009] The purpose of the present invention can be achieved through the following technical solutions: The present invention proposes a digital assembly system for fuel cell stacks, including physical space, digital space and information interaction between the two; wherein the physical space, namely the physical layer, corresponds to the entire process flow of actual stack assembly, including the screening and stacking of parts, the pressing and testing of the stack, and the devices and materials involved in the process flow; the digital space includes a functional layer, a model layer and a data layer; wherein the functional modules of the functional layer correspond one-to-one to the process flow in the physical layer, and the functional layer analyzes and optimizes the actual stack assembly process by calling the specific model in the model layer; the model layer includes A data model is obtained after processing the data stored in the data layer, a multi-physical field behavior analysis model for predicting the assembly state of the fuel cell stack based on the data model, and a service model for guiding the assembly of the fuel cell stack based on the multi-physical field behavior analysis model, and the integration and management of the model are realized through a unified modeling method, and the visual interaction of the model is realized by using geometric reconstruction and data reconstruction methods; the data layer stores and processes all information of the entire assembly process; the information interaction between the physical space and the digital space is completed through the interaction layer; each layer works together to realize the assembly guidance of the entire process of the fuel cell stack; based on the system to guide the assembly of the fuel cell stack, the specific assembly method includes the following steps:

[0010] S1: Parts screening: Use a variety of parts measurement tools to obtain actual processing information such as parts geometry and surface morphology and establish a corresponding parts data model. Then, based on the parts data model, evaluate the parts processing quality, select qualified parts in turn and grade them. Based on the parts data model, call the stacking analysis module, press-fit analysis module and performance analysis module to formulate a preliminary parts stacking sequence and stack press-fit displacement.

[0011] S2: Component stacking: Use the assembly sequence obtained in S1 to guide component stacking, establish a stacking process data model, collect assembly errors in real time during the stacking process, gradually improve the stacking process data model, and call the stacking analysis module to update the assembly sequence of the components to be stacked in real time until the stacking of the entire stack of components is completed;

[0012] S3: Pressing of the battery stack. Based on the stacking process data model obtained in the S2 assembly sequence, the battery stack pressing analysis module and the performance analysis module are called to update the pressing displacement of the battery stack, and a pressing process data model is established. During the pressing process, the battery stack assembly parameters are collected in real time, and the pressing process data model is gradually updated. The pressing displacement is updated in real time by calling the pressing analysis module and the performance analysis module until the pressing of the battery stack is completed.

[0013] S4: Stack test: Perform performance test on the pressed stack and collect performance data. Compare the actual performance data of the stack with the predicted data of the performance model in the digital space, so as to calibrate the accuracy of the stack assembly analysis model, provide more accurate guidance for the subsequent stack assembly, and finally obtain the assembled fuel cell stack.

[0014] The physical layer, when executing the actual fuel cell stack assembly process, can collect specific parameters of the four main assembly process flows, including component screening, component stacking, stack pressing and stack testing, and transmit them to the digital space through the interactive layer, thereby completing the actual fuel cell stack assembly based on the assembly guidance of the digital space.

[0015] The model layer is composed of three types of models, including data model, multi-physics field behavior analysis model and stack assembly service model, wherein the data model and multi-physics field behavior analysis model together constitute the fuel cell stack assembly digital model. The data model contains all the data processed in the fuel cell assembly process, including component geometry information, material information, stacking process parameters, press-fitting process parameters, assembly errors, etc.; the multi-physics field behavior analysis model includes the mechanism model and its proxy model for analyzing and predicting force behavior, electrical behavior, thermal behavior, water vapor transmission behavior, etc. in the fuel cell stack assembly process, which is solved by calling the parameter information in the data model and dynamically updating the data model according to the result data; the stack assembly service model acts on the functional layer, and uses the results of the analysis model as input to solve the optimized assembly scheme, including the component forming quality evaluation model, the stacking sequence optimization model, the press-fitting displacement optimization model, and the fuel cell performance prediction model. In order to improve the accuracy of the above models, the prediction results of the model are compared with the data collected in the physical space during the system operation, so as to realize the iterative update of the model layer.

[0016] The steps for constructing the data model, multi-physics field behavior analysis model and stack assembly service model in the model layer are as follows:

[0017] 1) Establishing a data model based on the information obtained by the data acquisition device;

[0018] 2) Based on the data model established in step 1), a mechanism model and its proxy model are constructed to form a multi-physical field behavior analysis model to achieve parameter prediction of multi-physical fields in the assembly process;

[0019] 3) Based on the multi-physical layer behavior analysis model constructed in step 2), and combined with the multi-objective optimization algorithm, the service model is constructed to achieve process optimization of the battery stack assembly process.

[0020] Specifically: the data model construction and update method mainly includes the following steps:

[0021] 101) Based on the theoretical geometric model constructed by the modeling software, the theoretical geometric state of the physical object is reproduced in the digital space, thus forming a basic component data model;

[0022] 102) Based on the component data model of step 101), a scanner or the like is used to collect data on the status information of the components in the physical space, the information of the components is obtained and reconstructed through point cloud registration, reverse engineering and other technologies, so as to reflect the real status of the components to be assembled in the digital space, such as the bipolar plate size error, the bipolar plate shape error, etc., and the component data model is updated through the parameter interface in the component entity class;

[0023] 103) Based on the component data model of step 102), a multi-physics field behavior analysis model is used to solve the physical behavior in the assembly process, and the obtained results are combined with the component data model to construct a basic stacking process data model and a press-fitting process data model;

[0024] 104) Based on the stacking process data model and the pressing process data model of step 103), the data collected in real time by sensors are processed in combination with a variety of filtering algorithms to obtain the assembly information of the actual battery stack assembly, such as the straightness error of the four sides of the battery stack and the parallelism error of the two end faces, and the stacking process data model and the pressing process data model are updated using the parameter interface in the battery stack assembly class.

[0025] As a preferred technical solution, the acquisition of the physical space component status information described in step 102) is specifically carried out as follows: for the geometric dimension data collection of the component, a high-precision laser scanner is used to obtain the contour information of the plate part, and based on the formula:

[0026]

[0027] The iterative closest point algorithm (ICP) is selected to perform registration processing on the actual point cloud, where and is the origin cloud P s and the target point cloud P t The corresponding point, R * With t * are respectively the values ​​of the rotation matrix R and the translation vector t that minimize the above formula, thereby obtaining the contour deviation of the bipolar plate.

[0028] As a preferred technical solution, the step 104) is to collect and process the assembly process data, and the specific method is: during assembly, the rotation servo position, speed, linear servo position, speed, assembly pressure and other parameters obtained by the sensor are based on the formula:

[0029]

[0030] Among them, y(n) represents the output signal of the filter, x(n) represents the input signal of the filter, and windowSize represents the window size of the filter; each term in the differential equation represents the weighted average of the input signal over a period of time, and the smoothness of the filter can be controlled by adjusting the window size; the sliding filtering technology is used to smooth the sensor signal and filter the raw data to obtain relatively stable and noise-free data.

[0031] As a preferred technical solution, when measuring the straightness of the side of the battery stack, the two-terminal point connection method is used to find the straightness measurement value with the minimum error.

[0032] As a preferred technical solution, when measuring the flatness error of the battery stack, the LSM (Least Square Method) algorithm is used to fit the center plane of the least squares as the reference plane using the measured data point set on the measured plane.

[0033] As a preferred technical solution, when measuring the parallelism error of the two end plates of the stack, an iterative method is used to evaluate the parallelism of the two end surfaces of the stack.

[0034] The construction of the multi-physics field behavior analysis model mainly includes the following steps:

[0035] S1) establishing a geometric model including design information and processing information of fuel cell stack components;

[0036] S2) Based on the geometric model constructed in step S1), structural simplification, material setting, boundary setting and meshing processing are performed in sequence to obtain the influence of component size errors and other main assembly parameters on the internal pressure distribution of the stack and the overall mechanical state of the stack, and to construct a force behavior model, an electrical behavior model, a thermal behavior model and a water vapor transmission behavior model of the fuel cell stack assembly process;

[0037] S3) Solving the model in step S2), constructing a simulation data set based on the results, and using the data set in combination with a deep learning method to establish a multi-physical field behavior analysis model for a fuel cell stack to improve the efficiency of the mechanism model.

[0038] The force behavior model is divided into a local mechanism model and an overall mechanism model. The local mechanism model analyzes the influence of component size errors and main assembly parameters on the internal pressure distribution of the battery stack, and the overall mechanism model is used to evaluate the influence of end plate deformation and materials on the mechanical state of the battery stack.

[0039] 201) Construction of local mechanism model of force behavior in fuel cell assembly process:

[0040] The model includes component geometry parameters, component material parameters, boundary condition parameters, assembly process parameters, etc., which are used to characterize the bipolar plate processing error. The main assembly parameters include the number of flow channels, the number of stack sections, and the influence of clamping force on the contact pressure between the bipolar plate and the gas diffusion layer in the fuel cell stack in the plane and through the plane. Specifically, it includes the following steps:

[0041] 2011) established a mathematical description of dimensional error based on actual bipolar plates manufactured by the forming process:

[0042]

[0043] Among them, Δh represents the size error, which is the difference between the height values ​​of the edge and the middle;

[0044] 2012) evaluated the in-plane gas diffusion layer contact pressure based on the following formula:

[0045]

[0046] in, and are the maximum contact pressure and the minimum contact pressure at the contact interface between the i-th gas diffusion layer and the electrode plate, respectively;

[0047] 2013) evaluated the contact pressure of the gas diffusion layer in the through-plane direction based on the following formula:

[0048]

[0049] in is the contact pressure difference between the i-th and (i-1)-th battery layers under the ridge at the same position;

[0050] 2014) developed a parameterized physical model to simulate the assembly process of a fuel cell stack containing multiple cells, and based on the above evaluation formula, obtained the relationship between the dimensional error in the plane and through-plane directions of the fuel cell stack, the number of flow channels, the number of stack sections, and the clamping force and contact pressure;

[0051] 2015) built an experimental platform to verify the results obtained from the numerical model.

[0052] 202) Construction of overall mechanism model of force behavior in fuel cell assembly process:

[0053] Similar to the local force behavior model of the assembly process, the overall force behavior mechanism model includes the overall geometric size parameters of the stack, boundary condition parameters, etc. Based on the composite model and material performance equivalent method, the macroscopic assembly parameters, such as the end plate material, the stack compression ratio and the number of stack nodes, are characterized to affect the end plate deformation and the end plate clamping force. Specifically, the following steps are included:

[0054] 2021) The composite model equivalent method is implemented based on the following formula:

[0055]

[0056] where v is the volume of the composite material, σ ij , ε ij and Respectively represent the stress, equivalent stress, strain and equivalent strain of the raw material and equivalent material on the ij plane; represents the equivalent strain on the kl plane, is the stiffness matrix of the equivalent model;

[0057] 2022) The material equivalent properties of the gas diffusion layer are obtained based on the following formula:

[0058]

[0059] Where P is the total force on the bipolar plate, L is the length of the gas diffusion layer and the bipolar plate, and u c is the equivalent compression performance of the gas diffusion layer;

[0060] 2023) Combine the core components of the fuel cell into a composite component, and make equivalents to other microscopic features based on model equivalence and performance equivalence, and establish an equivalent fuel cell stack model in finite element software;

[0061] 2024) Based on the established equivalent physical model and pressure evaluation index, the influence of end plate material, stack compression ratio and stack number on end plate deformation and end plate clamping force is obtained;

[0062] 2025) The correctness of the constructed equivalent finite element model was verified through experiments.

[0063] 203) Construct prediction models for the electrical, thermal and water vapor transmission behaviors of the stack assembly: the electrical behavior model mainly analyzes the contact resistance and body resistance of the fuel cell interface; the thermal behavior model mainly analyzes the thermal conductivity and body thermal conductivity of the internal interface of the stack; the water vapor transmission behavior model mainly analyzes the water vapor transmission resistance inside the stack;

[0064] The model specifically includes the contact properties between the membrane electrode and the gas diffusion layer, the electrode plate and the gas diffusion layer, the material type and the material surface treatment state, and is used to characterize the relationship between the interface contact pressure and the contact resistance, contact thermal resistance and water vapor transmission resistance, and to predict the contact resistance, contact thermal resistance and water vapor transmission resistance in the fuel cell assembly. Specifically, it includes the following steps:

[0065] 2031) Obtain the constitutive relationship between the internal contact pressure and contact resistance, contact thermal resistance and water vapor transmission resistance of the battery stack;

[0066] 2032) Establish an "electric-force" coupling assembly model for simulating electric field and force field, a "heat-force" coupling assembly model for simulating temperature field and force field, and a "water vapor transmission-force" coupling assembly model for simulating flow field and force field. Specifically, the following steps are included:

[0067] 20321) establishing an assembly geometry model inside the battery stack, wherein the battery stack assembly components used are the same as those in step 2031), and defining material properties and applying boundary conditions, etc.;

[0068] 20322) Establish contact relations: the force contact relations are established by the traditional contact pair method, and the electrical, thermal and water vapor transmission contact relations are established based on the constitutive relations between the corresponding multi-physics field behaviors and the contact pressure;

[0069] 20323) applying force field load and multi-physics field load simultaneously based on the model established in the above steps to obtain the contact resistance, contact thermal resistance and water vapor transmission resistance between the membrane electrode and the gas diffusion layer, between the electrode plate and the gas diffusion layer, and by applying a series of assembly pressures, obtain the relationship between the interface contact resistance, contact thermal resistance and water vapor transmission resistance and the contact pressure;

[0070] 2033) By comparing with the obtained constitutive relationship results, the correctness of the "electric-mechanical", "thermal-mechanical" and "water-gas transmission-mechanical" coupled assembly models is verified.

[0071] 204) Constructing a multi-physics behavior analysis agent model for the fuel cell assembly process:

[0072] The input of the model is geometric parameters and process parameters such as displacement load or force load, and the output is parameters that characterize physical behavior characteristics such as force, electricity, heat and water vapor transmission after specific physical behavior. It can be used as a proxy model of the mechanism model to quickly predict the multi-physical field behavior of fuel cells, thereby improving the execution efficiency of the mechanism model. Specifically, it includes the following steps:

[0073] 2041) running the constructed multi-physics field behavior analysis mechanism model, obtaining result displacement information as output information, and geometric parameters and process parameters such as displacement load or force load as input information, thereby constructing a simulation data set;

[0074] 2042) Based on the result of step 2041), input and output parameters are selected according to specific requirements, and a neural network model with specified functions is established on the basis of the simulation data set;

[0075] 2043) Compare the result of step 2042) with the running result of the mechanism model, and then replace the mechanism model with the neural network model to perform multi-physics field behavior analysis.

[0076] The service models mainly include: component evaluation model, component stacking sequence optimization model, battery stack press displacement optimization model and battery stack performance prediction model.

[0077] As a preferred technical solution, the component evaluation model is mainly constructed based on the satisfaction function of the multi-objective optimization algorithm; by running the force prediction model established in the multi-physical field behavior analysis agent model of the fuel cell assembly process in step 204), the contact resistance, porosity and contact pressure distribution between flow channels under the influence of different dimensional errors are obtained and the satisfaction calculation is performed to obtain the relationship between the overall satisfaction and the dimensional error. According to the requirements of different satisfaction, the acceptable dimensional error is determined to screen and optimize the components.

[0078] The satisfaction function is established by the following formula:

[0079]

[0080] where d R , They are contact resistance, gas diffusion layer porosity, and contact pressure distribution standard deviation, respectively.

[0081] As a preferred technical solution, the component stacking sequence optimization model is constructed based on the satisfaction function, and the force behavior prediction model established in the multi-physical field behavior analysis agent model of the fuel cell assembly process is run in step 204 to obtain the distribution of contact resistance, porosity and contact pressure between flow channels under the influence of different assembly errors and perform satisfaction calculations to obtain the relationship between the overall satisfaction and the assembly error, and determine the acceptable assembly error according to the requirements of different satisfaction levels, and adjust and optimize the assembly sequence in real time;

[0082] As a preferred technical solution, the stack press-fit optimization analysis model is constructed based on a satisfaction function, and by running step 204) the force behavior prediction model established in the multi-physical field behavior analysis proxy model of the fuel cell assembly process, the total contact resistance, porosity and contact pressure distribution between flow channels under different press-fit displacements are obtained and satisfaction calculations are performed to obtain the relationship between the overall satisfaction and the press-fit displacement, and according to the requirements of different satisfaction levels, the acceptable press-fit displacement is determined to optimize the press-fit displacement;

[0083] The battery stack performance prediction model takes as input component information, stacking process parameters, etc. in digital or physical space, and outputs specific performance of the battery stack, including battery output VI curve, power density and service life, etc. It is used to predict the performance of the assembled battery stack in advance so that the process flow can be optimized based on performance indicators.

[0084] The model layer is integrated and managed through a unified modeling method, and the product is expressed abstractly based on object-oriented thinking. According to the properties of components, production process technology and multi-physical field coupling characteristics, the fuel cell stack entity class, multi-physical field behavior analysis class and stack assembly service class are constructed. The fuel cell assembly process is regarded as the mutual transmission of messages between the three types of objects, realizing model integration and coordinated operation. Specifically:

[0085] The main attributes of the fuel cell stack entity class include both the attributes of the fuel cell components and the matching characteristics of the assembly relationship between the components. The method is a contact model that describes the component assembly process; the multi-physical field behavior analysis class attributes are the relevant parameters and results of the corresponding multi-physical field analysis model, and the method is a parameterized multi-physical field behavior prediction model; the main methods of the stack assembly service class include forming quality evaluation methods, assembly accuracy calculation methods, and assembly process optimization methods, etc., and its attributes are the main parameters that constitute the evaluation method, such as contact resistance, contact pressure distribution standard deviation, etc.

[0086] The update iteration method of the model layer is specifically:

[0087] Through geometric reconstruction and data reconstruction methods, the basic geometric structure and key numerical information of the multi-physical field behavior analysis model are extracted, and a visual interactive model is constructed. Based on the data interaction method, the fuel cell assembly process parameters collected in the physical space are used to drive step 204) the multi-physical field behavior analysis agent model of the fuel cell assembly process to achieve efficient prediction of the multi-physical field behavior of the actual fuel cell assembly process. Finally, the multi-physical field state information of key components is mapped to the numerical interface of the visual interactive model, thereby realizing the iterative update of the fuel cell assembly state in the digital space.

[0088] The functional layer is mainly composed of four functional modules of the corresponding constructed service model, which realizes the optimization of the whole process of fuel cell stack assembly. Specific assembly method:

[0089] The modules of the functional layer in the digital space are interconnected. The digital space can be ahead of the physical space and perform virtual simulation of the entire process flow. The screening analysis module can evaluate the quality of parts based on the numerical value of the processing deviation of the parts, or it can analyze the influence of the processing deviation of parts on the cumulative error of the battery stack, the uniformity of the whole stack assembly and the performance of the battery stack based on the subsequent modules including the stacking analysis module, the pressing analysis module and the performance analysis module, so as to conduct a deeper evaluation of the processing quality of parts and screen out qualified parts; the stacking analysis module can guide the optimization of the assembly sequence based on the cumulative error of the stacking after the virtual stacking is completed, or it can analyze the influence of the cumulative error of the stacking on the uniformity of the whole stack and the battery performance based on the subsequent modules including the pressing analysis module and the performance analysis module, so as to obtain a more reasonable assembly sequence; the pressing analysis module can optimize the pressing displacement based on the uniformity of the whole stack after the virtual pressing, or it can further analyze the influence of the pressing displacement on the battery performance through the battery performance analysis module, so as to obtain the optimal assembly displacement for pressing guidance of the battery stack.

[0090] The digital assembly system and assembly method for fuel cell stacks provided by the present invention have at least the following beneficial effects compared with the prior art:

[0091] 1) The present invention fully exploits the data information in the fuel cell stack assembly process, obtains various data of the fuel cell assembly process in real time through sensors, and establishes a fuel cell digital model corresponding to the physical space based on this. The fuel cell digital model loads, simulates and iteratively updates various boundary conditions according to the various process parameters obtained, thereby more realistically reflecting the multi-physical field state of the fuel cell physical entity in the entire process assembly process. The present invention can fully exploit the physical characteristics of the fuel cell stack components and related process parameters during the fuel cell stack assembly process, and improve the utilization rate of relevant data in the assembly process.

[0092] 2) The present invention combines the fuel cell assembly digital model and multi-objective optimization method to provide guidance for the optimization control of processes such as component screening, stacking sequence, and press-fit displacement in the actual assembly process, and successively obtains qualified components, the optimal component stacking sequence, and the optimal stack press-fit displacement, thereby achieving the optimization of the fuel cell stack product assembly quality.

[0093] 3) The present invention solves the problems of insufficient utilization of assembly process data, reliance on experience in determining assembly process, inability to dynamically display assembly process, inability to evaluate assembly quality in real time, and inability to adjust assembly process in real time under current battery assembly methods. By combining data analysis technology, the present invention improves fuel cell assembly efficiency, enhances fuel cell performance, and reduces cost losses caused by errors in the assembly process. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] Figure 1 This is an architecture diagram of a digital assembly system for a fuel cell stack in an embodiment;

[0095] Figure 2 It is a schematic diagram of the flow of the method for constructing a local model for predicting the force behavior of a stack assembly in an embodiment;

[0096] Figure 3 It is a schematic diagram of the flow of the method for constructing the overall model for predicting the force behavior of the stack assembly in the embodiment;

[0097] Figure 4 It is a schematic diagram of the process of constructing a prediction model of the electrical behavior of a stack assembly in an embodiment;

[0098] Figure 5 Schematic diagram of constructing a multi-physics field behavior analysis agent model based on a neural network in an embodiment;

[0099] Figure 6 Schematic diagram of the multi-model integration method in the embodiment;

[0100] Figure 7 A schematic diagram of a digital model updating and iterative method in an embodiment;

[0101] Figure 8 Schematic diagram of data interaction between digital space and physical space in an embodiment. DETAILED DESCRIPTION

[0102] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should belong to the scope of protection of the present invention.

[0103] Example 1

[0104] like Figure 1 As shown, the present invention relates to a digital assembly system for fuel cell stacks, including physical space, digital space and information interaction between the two; wherein the physical space, i.e., the physical layer, corresponds to the entire process flow of actual stack assembly, including parts screening, parts stacking, stack pressing and stack testing, as well as devices and materials involved in the process flow; the digital space includes a functional layer, a model layer and a data layer; wherein the functional modules of the functional layer correspond one-to-one to the process flow in the physical layer, and the functional layer analyzes and optimizes the actual stack assembly process by calling a specific model in the model layer; the model layer includes a data model obtained by processing the data stored in the data layer, a multi-physical field behavior analysis model for predicting the stack assembly state based on the data model, and a service model for guiding the stack assembly based on the multi-physical field behavior analysis model, and realizes the integration and management of the model through a unified modeling method, and realizes the visual interaction of the model by using geometric reconstruction and data reconstruction methods; the data layer stores and processes all information of the entire assembly process; the information interaction between the physical space and the digital space is completed through the interaction layer; the five architecture layers work together to realize the assembly guidance of the entire process of the fuel cell stack; based on the system to guide the stack assembly, the specific assembly method includes the following steps:

[0105] Step 1. Parts screening: Use a variety of parts measurement tools to obtain actual processing information such as part geometry and surface morphology and establish a corresponding part data model. Then, based on the part data model, evaluate the part processing quality and screen parts with errors within an acceptable range. Further, call the stacking analysis module, press-fit analysis module and performance analysis module based on the part data model to obtain the impact of part processing deviations on assembly mechanical uniformity and stack performance consistency, check part quality, and screen out parts that meet the requirements. At the same time, formulate a preliminary parts stacking sequence and stack press-fit displacement.

[0106] Step 2, parts stacking: Based on the stacking sequence obtained in step 1, stack the parts, collect assembly errors in real time during the stacking process, and gradually update the stacking process data model; obtain the stacking sequence of the battery stack with an acceptable cumulative assembly error after the virtual stacking is completed through the stacking analysis module; further, based on the stacking process data model, call the press analysis module and the performance analysis module to obtain the impact of the cumulative assembly error on the uniformity of assembly mechanics and the consistency of battery stack performance, and then obtain the real-time optimal assembly sequence, and update the stacking sequence of the remaining parts in real time until the stacking of the parts is completed.

[0107] Step 3, battery stack pressing: Based on the stacking process data model obtained from the stacking sequence in step 2, call the battery stack pressing analysis module and the performance analysis module to set the pressing displacement, and collect the battery stack assembly parameters in real time during the pressing process, and gradually update the pressing process data model until the battery stack pressing is completed.

[0108] Step 4: Stack test: Perform performance test on the pressed stack and collect performance data. Compare the actual performance data of the stack with the predicted data of the performance model in the digital space, so as to calibrate the accuracy of the stack assembly analysis model, provide more accurate guidance for the next stack assembly, and finally obtain a fully assembled fuel cell stack.

[0109] like Figure 1 As shown, the physical layer, when executing the actual stack assembly process, can collect specific parameters of the four main assembly process flows, including component screening, component stacking, stack pressing and stack testing, and transmit them to the digital space through the interaction layer, thereby completing the actual fuel cell stack assembly based on the assembly guidance of the digital space.

[0110] The model layer is composed of three types of models, including data model, multi-physics field behavior analysis model and stack assembly service model, wherein the data model and multi-physics field behavior analysis model together constitute the fuel cell stack assembly digital model. The data model contains all the processed data of the fuel cell assembly process, including component geometry information, material information, stacking process parameters, press-fitting process parameters, assembly errors, etc.; the multi-physics field behavior analysis model includes the mechanism model and its proxy model for analyzing and predicting force behavior, electrical behavior, thermal behavior, water vapor transmission behavior, etc. during the fuel cell stack assembly process, which is solved by calling the parameter information in the data model and dynamically updating the data model according to the result data; the stack assembly service model acts on the functional layer, and uses the results of the analysis model as input to solve the optimized assembly scheme, including the component forming quality evaluation model, the stacking sequence optimization model, the press-fitting displacement optimization model, and the fuel cell performance prediction model. In order to improve the accuracy of the above models, the model prediction results need to be compared with the data collected in the physical space during the system operation, so as to realize the iterative update of the model layer.

[0111] The steps for constructing the data model, multi-physics field behavior analysis model and stack assembly service model in the model layer are as follows:

[0112] 1) Establishing a data model based on the information obtained by the data acquisition device;

[0113] 2) Based on the data model established in step 1), a mechanism model and its proxy model are constructed to form a multi-physical field behavior analysis model to achieve parameter prediction of multi-physical fields in the assembly process;

[0114] 3) Based on the multi-physical field behavior analysis model constructed in step 2), and combined with a multi-objective optimization algorithm, the service model is constructed to achieve process optimization of the stack assembly process.

[0115] Specifically: the data model construction and update method mainly includes the following steps:

[0116] 101) Construct theoretical CAD models of assembly parts based on modeling software including but not limited to Solidworks to reproduce the theoretical geometric state of physical objects in digital space, thereby forming a basic parts data model;

[0117] 102) Acquisition of physical space component status information: Based on the component data model of step 101), a laser 3D scanner with a surface accuracy of 10 is used to acquire the contour information (i.e., geometric dimension data) of components such as plate parts during the component data acquisition process, and reconstruction is achieved through point cloud registration, reverse engineering and other technologies, thereby reflecting the real status of the components to be assembled in the digital space, such as bipolar plate size error, bipolar plate shape error, etc., and updating the component data model through the parameter interface in the component entity class;

[0118] 103) Based on the component data model of step 102), a multi-physics field behavior analysis model is used to solve the physical behavior in the assembly process, and the obtained results are combined with the component data model to construct a basic stacking process data model and a press-fitting process data model;

[0119] 104) Collecting and processing assembly process data: Based on the stacking process data model and the pressing process data model of step 103), the data collected in real time by the pressure sensor, flow sensor, motor driver and FARO portable three-coordinate measuring arm are processed in combination with a variety of filtering algorithms to obtain the assembly information of the actual battery stack assembly, such as the straightness error of the four sides of the battery stack and the parallelism error of the two end faces, etc., and the stacking process data model and the pressing process data model are updated using the parameter interface in the battery stack assembly class.

[0120] The specific method of obtaining the physical space component status information described in step 102) is as follows: for the geometric dimension data collection of the components, a high-precision laser scanner is used to obtain the contour information of the parts such as the plate, and based on the formula:

[0121]

[0122] The iterative closest point algorithm (ICP) is selected to perform registration processing on the actual point cloud, where and is the origin cloud Ps and the target point cloud P t The corresponding point in R * With t * are respectively the values ​​of the rotation matrix R and the translation vector t that minimize the above formula, thereby obtaining the contour deviation of the bipolar plate.

[0123] The specific method of collecting and processing the assembly process data described in step 104) is as follows: during assembly, the rotation servo position, speed, linear servo position, speed, assembly pressure and other parameters obtained by the sensor are used, and based on the formula:

[0124]

[0125] Among them, y(n) represents the output signal of the filter, x(n) represents the input signal of the filter, and windowSize represents the window size of the filter; each term in the differential equation represents the weighted average of the input signal over a period of time, and the smoothness of the filter can be controlled by adjusting the window size; the sliding filtering technology is used to smooth the sensor signal and filter the raw data to obtain relatively stable and noise-free data.

[0126] When measuring the straightness of the side of the battery stack, use the two-terminal point connection method to find the straightness measurement value with the minimum error.

[0127] When measuring the flatness error of the battery stack, the LSM (Least Square Method) algorithm is used to fit the least squares center plane as the reference plane using the measured data point set on the measured plane.

[0128] When measuring the parallelism error of the two end plates of the stack, an iterative method is used to evaluate the parallelism of the two end surfaces of the stack.

[0129] The construction of the multi-physics field behavior analysis model mainly includes the following aspects:

[0130] 201) Construction of local mechanism model of force behavior in fuel cell assembly process:

[0131] The model includes component geometric size parameters, component material parameters, boundary condition parameters, assembly process parameters, etc., which are used to characterize the bipolar plate processing error. The main assembly parameters include the number of flow channels, the number of stack nodes and the influence of clamping force on the contact pressure between the bipolar plate and the gas diffusion layer in the fuel cell stack in the plane and through-plane direction.

[0132] like Figure 2 Specifically, the following steps are included:

[0133] 2011) established the definition of dimensional error based on the actual bipolar plates manufactured by the stamping process, and the mathematical description is:

[0134]

[0135] Among them, Δh represents the size error, which is the difference between the height values ​​of the edge and the middle;

[0136] 2012) evaluated the in-plane gas diffusion layer contact pressure based on the following formula:

[0137]

[0138] in, and are the maximum contact pressure and the minimum contact pressure at the contact interface between the i-th gas diffusion layer and the electrode plate, respectively;

[0139] 2013) evaluated the contact pressure of the gas diffusion layer in the through-plane direction based on the following formula:

[0140]

[0141] in is the contact pressure difference between the i-th and (i-1)-th battery layers under the ridge at the same position;

[0142] 2014) The geometric information of the CAD model of the fuel cell stack components constructed by SolidWorks is mapped to the ABAQUS finite element analysis software, and the material setting, boundary setting, load setting and meshing are performed in sequence, and the parameters are saved to realize the construction of the finite element model. Based on the finite element model, the dimensional error in the gas diffusion plane and the through-plane direction of the fuel cell stack, the number of flow channels, the number of fuel cell stack sections, and the relationship between the clamping force and the contact pressure are obtained under the above evaluation criteria;

[0143] 2015) built an experimental platform to verify the results obtained from the numerical model.

[0144] 202) Construction of overall mechanism model of force behavior in fuel cell assembly process:

[0145] Similar to the local model of force behavior in the assembly process, the overall mechanism model of force behavior includes the overall geometric size parameters of the stack, boundary condition parameters, etc. Based on the composite model and material performance equivalent method, it characterizes the influence of macroscopic assembly parameters such as end plate material, stack compression ratio and stack node number on end plate deformation and end plate clamping force.

[0146] like Figure 3 Specifically, the following steps are included:

[0147] 2021) The composite model equivalent method is implemented based on the following formula:

[0148]

[0149] where v is the volume of the composite material, σ ij , ε ij and Respectively represent the stress, equivalent stress, strain and equivalent strain of the raw material and equivalent material on the ij plane; represents the equivalent strain on the kl plane, is the stiffness matrix of the equivalent model;

[0150] 2022) The equivalent properties of the gas diffusion layer material are obtained based on the following formula:

[0151]

[0152] Where P is the total force on the bipolar plate, L is the length of the gas diffusion layer and the bipolar plate, and u c is the equivalent compression performance of the gas diffusion layer;

[0153] 2023) Combine the bipolar plate and the membrane electrode assembly into a composite component, retain the air manifold, hydrogen manifold and cooling manifold structure, and perform equivalent to other microscopic features based on step 2021) and step 2022). Utilize the symmetry of the fuel cell stack, build a quarter CAD model of the actual fuel cell stack in software including but not limited to SolidWorks, and use finite element analysis software ABAQUS as a modeling platform for exploration, and process the geometric information from the CAD model in turn by material setting, boundary setting, load setting and meshing, and save the parameters to achieve the construction of the finite element model;

[0154] 2024) Based on the established equivalent finite element model and pressure evaluation index, the influence of end plate material, stack compression ratio and stack number on end plate deformation and end plate clamping force is obtained;

[0155] 2025) The correctness of the constructed equivalent finite element model was verified through experiments.

[0156] 203) Construct a prediction model for the electrical, thermal and water vapor transmission behavior of the stack assembly:

[0157] The model includes the contact properties between the membrane electrode and the gas diffusion layer, the electrode plate and the gas diffusion layer, the material type and the surface treatment status of the material. It is used to characterize the relationship between the interface contact pressure and the contact resistance, contact thermal resistance and water vapor transmission resistance, and to predict the contact resistance, contact thermal resistance and water vapor transmission resistance in the fuel cell assembly.

[0158] The establishment of the electrical behavior prediction model is as follows Figure 4 Specifically, the following steps are included:

[0159] 2031) Obtain the “electric coupling” constitutive relationship between contact pressure and contact resistance inside the battery stack;

[0160] 2032) Establish an "electric-force" coupling assembly model that simulates the electric field and force field. Specifically, it includes the following steps:

[0161] 20321) establishing an assembly geometry model inside the battery stack, wherein the battery stack assembly components used are the same as those in step 2031), and defining material properties and applying boundary conditions, etc.;

[0162] 20322) Establish contact relationship: the force contact relationship is established by the traditional contact pair method, and the electrical contact relationship is established based on the constitutive relationship between the corresponding multi-physics field behavior and the contact pressure;

[0163] 20323) applying force field load and multi-physics field load simultaneously based on the model established in the above steps to obtain the contact resistance between the membrane electrode and the gas diffusion layer, and between the electrode plate and the gas diffusion layer, and obtaining the relationship between the interface contact resistance and the contact pressure by applying a series of assembly pressures;

[0164] 2033) By comparing with the obtained constitutive relationship results, the correctness of the "electric-mechanical" coupling assembly model is verified.

[0165] Based on the construction method of the electrical behavior prediction model, a heat and water vapor transmission behavior prediction model is established.

[0166] 204) Constructing a proxy model of force behavior in the fuel cell assembly process:

[0167] The input of this model is geometric parameters and process parameters such as displacement load or force load, and the output is the force-related characteristics after performing specific physical behaviors. It can be used as a proxy model of the mechanism model to quickly predict the contact pressure of the fuel cell, thereby improving the execution efficiency of the mechanism model.

[0168] like Figure 5 Specifically, the following steps are included:

[0169] 2041) running the constructed force behavior analysis mechanism model, using the resulting displacement information, force information, stress information, strain information and contact information as output information, and using geometric parameters and process parameters such as displacement load or force load as input information, thereby constructing a simulation data set;

[0170] 2042) Based on the result of step 2041), input and output parameters are selected according to specific requirements, and an artificial neural network model is established based on the simulation data set;

[0171] 2043) Compare the result of step 2042) with the result of the mechanism model operation, and then use the artificial neural network model to replace the mechanism model for mechanical behavior analysis.

[0172] The service models mainly include: component evaluation model, component stacking optimization model, battery stack pressing analysis model and battery stack performance prediction model, etc.

[0173] The component evaluation model is mainly constructed based on the satisfaction function of the multi-objective optimization algorithm. By running the force prediction model established in step 204), the contact resistance, porosity and contact pressure distribution between flow channels under the influence of different dimensional errors are obtained and the satisfaction calculation is performed to obtain the relationship between the overall satisfaction and the dimensional error. According to the requirements of different satisfaction levels, the acceptable dimensional error is determined to screen and optimize the components.

[0174] The satisfaction function is established by the following formula:

[0175]

[0176] where d R , They are contact resistance, gas diffusion layer porosity, and contact pressure distribution standard deviation, respectively.

[0177] The component stacking order optimization model is constructed based on the satisfaction function. By running the force behavior prediction model established in step 204), the contact resistance, porosity and contact pressure distribution under different assembly errors are obtained and the satisfaction calculation is performed to obtain the relationship between the overall satisfaction and the assembly error. According to the requirements of different satisfaction, the acceptable assembly error is determined to adjust and optimize the real-time assembly order. The battery stack press assembly optimization analysis model is constructed based on the satisfaction function. By running the force prediction model established in step 204), the total contact resistance, porosity and contact pressure distribution under different press assembly displacements are obtained and the satisfaction calculation is performed to obtain the relationship between the overall satisfaction and the press assembly displacement. According to the requirements of different satisfaction, the acceptable press assembly displacement is determined to optimize the press assembly displacement. The battery stack performance prediction model is input as component information, stacking process parameters, etc. in digital or physical space, and output as the specific performance of the battery stack, including battery output VI curve, power density and service life, etc., which is used to predict the performance of the assembled battery stack in advance so that the process can be optimized based on performance indicators.

[0178] like Figure 6As shown in the figure, based on the object-oriented concept, the unified modeling method is used to integrate and manage the model. According to the properties of components, production process technology and multi-physical field coupling characteristics, the fuel cell stack entity class, multi-physical field analysis class and service class are constructed. The fuel cell assembly process is regarded as the mutual transmission of messages between the three types of objects to achieve model integration and coordinated operation. Specifically:

[0179] The fuel cell stack entity class is subdivided into parts class and assembly class, among which the parts class can be further subdivided into end plate class, current collecting plate class, insulating plate class, bipolar plate class, MEA class and sealing class, etc. The main attributes include CAD and physical parameters of fuel cell parts, etc.; the assembly is divided into air-cooled battery, water-cooled battery, and reversible battery class, and the main attributes include both the attributes of the fuel cell components and the main attributes of the matching characteristics of the assembly relationship between the components, and the method is the contact model of the component structure assembly process; the multi-physical field analysis class attributes of the fuel cell are the relevant parameters and results of the corresponding multi-physical field behavior analysis model, and the method is to establish a parameterized multi-physical field behavior prediction model such as force behavior, electrical behavior and thermal behavior; the main methods of the stack assembly service class include forming quality evaluation method, assembly accuracy calculation method and assembly process optimization method, etc., and the attributes are the main parameters of the evaluation method, such as contact resistance, contact pressure distribution standard deviation, etc.; the service class input is the attributes of the analysis class and the entity class, and the output is a specific process flow optimization, which is used to guide the actual fuel cell assembly process.

[0180] The digital model updating and iterative method is as follows: Figure 7 Specifically, the following steps are included:

[0181] The first data mapping of the component information (such as sensor data) in the physical space is carried out based on the geometric model, that is, the actual manufacturing information of the physical object is included in the theoretical CAD model of the assembly component constructed in Solidworks;

[0182] Based on the finite element analysis software ABAQUS, the CAD model is sequentially processed with material setting, boundary setting and meshing, and the parameters and geometric information are mapped to the new INP file to realize the construction of the finite element model;

[0183] Solve the finite element model and perform the third data mapping to store the simulation prediction data of the physical object behavior in the ODB file, which contains both the geometric data of the previous step and the physical behavior data corresponding to the geometric information;

[0184] Based on the constructed OBD file, data is reorganized, and the geometric data and simulation data in the ODB file are extracted to construct a simulation data set. At the same time, the corresponding relationship between the index position and the number in the file geometry information is recorded, and the variables are read into the array in the order of their node numbers. The unit and node numbers are matched with the index positions by reordering, and the geometry and data are reorganized to form the final VTK file as the carrier of the visual digital model.

[0185] Finally, the predicted multi-physical field state information of key components is mapped to the numerical interface in the VTK file in turn using the proxy model, and real-time updates are performed to achieve iterative updates of the model layer of the physical object.

[0186] The information transmission between the digital space and the physical space of the stack assembly is realized through data interaction, such as Figure 8 As shown in the figure, the data interaction of the digital assembly system mainly includes three parts, namely, data interaction within the virtual space, data interaction within the physical space, and data interaction between the virtual and real space. Specifically:

[0187] The internal interaction of virtual space is mainly reflected in the construction of digital twin models, model deployment and application, and data management. Based on the geometric model of fuel cell components and the corresponding multi-physical field analysis model, batch processing tools are used to achieve high-throughput model calculation, obtain the mapping relationship between the input parameters of the fuel cell assembly process and the multi-physical field response of the battery, and then build a model database; at the same time, the production history data of the physical platform for the fuel cell stack assembly in the physical space is also obtained, and an experimental database is built to supplement and correct the model database; with the database as the core, the Tensorflow tool is used to build big data model technologies such as neural network models, and the model is deployed in the cloud based on the open source virtual space management software Docker; at the same time, the modeling tool is used to separate the discrete digital model from the model, and the interface configuration is performed to realize the parameterization of the digital model, and the parameterized digital model deployment is realized locally.

[0188] The interaction within the physical space is mainly reflected in the data transmission between the fuel cell stack assembly physical platform, sensors, data acquisition terminals, user industrial computers, control terminals and controllers. Pressure sensors, flow sensors, barcode scanners and 3D laser scanners are used to receive multi-source data such as assembly pressure, stack air tightness, and stack component material parameters in the assembly platform, and upload them to the data acquisition terminal through the RS232 protocol. 5G technology is then used to communicate data with the user computer, and the user computer implements the execution of the specific control strategy of the motor controller through the RS232 protocol through the control terminal.

[0189] The interaction between virtual and real space is mainly reflected in the combination of battery stack assembly software and user computers. The battery stack assembly software is an interactive terminal of virtual space, which interacts through API interfaces, big data models deployed in the cloud, and parametric digital models deployed locally. User computers, as interactive terminals of physical space, realize the integration of interactive information in physical space. By deploying the software on user computers, it is possible to utilize physical space data, iteratively update virtual space, formulate control strategies, and finally feed back to the physical space to realize data interaction between virtual and real space.

[0190] like Figure 1 As shown, the functional layer is mainly composed of four functional modules of the corresponding constructed service model, which realizes the optimization of the whole process of fuel cell stack assembly. The specific assembly method is:

[0191] The modules of the functional layer in the digital space are interconnected. The digital space can be ahead of the physical space and perform virtual simulation of the entire process flow. The screening analysis module can evaluate the quality of parts based on the numerical value of the processing deviation of the parts, or it can analyze the influence of the processing deviation of parts on the cumulative error of the battery stack, the uniformity of the whole stack assembly and the performance of the battery stack based on the subsequent modules including the stacking analysis module, the pressing analysis module and the performance analysis module, so as to conduct a deeper evaluation of the processing quality of parts and screen out qualified parts; the stacking analysis module can guide the optimization of the assembly sequence based on the cumulative error of the stacking after the virtual stacking is completed, or it can analyze the influence of the cumulative error of the stacking on the uniformity of the whole stack and the battery performance based on the subsequent modules including the pressing analysis module and the performance analysis module, so as to obtain a more reasonable assembly sequence; the pressing analysis module can optimize the pressing displacement based on the uniformity of the whole stack after the virtual pressing, or it can further analyze the influence of the pressing displacement on the battery performance through the battery performance analysis module, so as to obtain the optimal assembly displacement for pressing guidance of the battery stack.

Claims

1. A digital assembly system for fuel cell stacks, characterized in that: The system includes physical space, digital space and information interaction between the two; The physical space, i.e., the physical layer, corresponds to the entire process of actual stack assembly, including component selection, component stacking, stack press assembly and stack testing, as well as the devices and materials involved in the process; The digital space includes a function layer, a model layer and a data layer; wherein the function modules of the function layer correspond one-to-one to the process flow in the physical layer, and the function layer analyzes and optimizes the actual battery stack assembly process by calling the specific model in the model layer; the model layer includes a data model obtained by processing the data stored in the data layer, a multi-physical field behavior analysis model for predicting the battery stack assembly state based on the data model, and a service model for guiding the battery stack assembly based on the multi-physical field behavior analysis model, and realizes the integration and management of the model through a unified modeling method, and realizes the visual interaction of the model by using geometric reconstruction and data reconstruction methods; The data layer stores and processes all information of the entire assembly process; The information interaction between the physical space and the digital space is completed through the interaction layer; each layer works together to realize the assembly guidance of the entire process of the fuel cell stack.

2. A digital assembly system for fuel cell stacks according to claim 1, characterized in that: When executing the actual fuel cell stack assembly process, the physical layer collects specific parameters of the four main assembly process flows, including component screening, component stacking, stack pressing and stack testing, and transmits them to the digital space through the interactive layer, thereby completing the actual fuel cell stack assembly based on the assembly guidance of the digital space.

3. A digital assembly system for fuel cell stacks according to claim 1, characterized in that: The data model and multi-physical field behavior analysis model in the model layer together constitute a digital model of fuel cell stack assembly; the data model contains all the data processed during the fuel cell assembly process, including component geometry information, material information, stacking process parameters, pressing process parameters and assembly errors; the multi-physical field behavior analysis model includes a mechanism model and its proxy model for analyzing and predicting force behavior, electrical behavior, thermal behavior, and water vapor transmission behavior during the fuel cell stack assembly process, which is solved by calling the parameter information in the data model and dynamically updating the data model based on the result data; the service model acts on the functional layer, and uses the results of the multi-physical field behavior analysis model as input to solve an optimized assembly scheme, including a component forming quality assessment model, a stacking sequence optimization model, a pressing displacement optimization model, and a fuel cell performance prediction model; during system operation, the prediction results of the model are compared with the data collected in the physical space, thereby realizing iterative updates of the model layer.

4. A digital assembly system for fuel cell stacks according to claim 1, characterized in that: The construction methods of the data model, multi-physical field behavior analysis model and service model in the model layer are respectively: establishing a data model according to the information obtained by the data acquisition device; based on the data model, constructing a mechanism model and its proxy model to form a multi-physical field behavior analysis model to realize the multi-physical field behavior prediction of the battery stack assembly process; Based on the multi-physical field behavior analysis model and combined with the multi-objective optimization algorithm, the service model is constructed to achieve process optimization of the battery stack assembly process.

5. A digital assembly system for fuel cell stacks according to claim 4, characterized in that , the data model construction and update method comprises the following steps: 501) Based on the theoretical geometric model constructed by the modeling software, the actual geometric state of the physical object and the actual manufacturing information are reproduced in the digital space, thereby constructing a basic component data model; 502) Based on the component data model of step 501), data collection is performed on the component status information in the physical space, the component information is obtained and reconstructed through point cloud registration and reverse engineering, so as to reflect the real status of the assembled components in the digital space, and the component data model is updated through the parameter interface in the component entity class; 503) Based on the component data model of step 502), the physical behavior in the assembly process is solved using a multi-physics field behavior analysis model, and the obtained results are combined with the component data model to construct a basic stacking process data model and a press-fitting process data model; 504) Based on the stacking process data model and the pressing process data model of step 503), the real-time collected data is filtered in combination with a variety of filtering algorithms to obtain the assembly information of the actual battery stack assembly, and the stacking process data model and the pressing process data model are updated using the parameter interface in the battery stack assembly class.

6. A digital assembly system for fuel cell stacks according to claim 4, characterized in that: The multi-physics field behavior analysis model construction method comprises the following steps: 601) Establishing a geometric model including design information and processing information of fuel cell stack components; 602) Based on the geometric model constructed in step 601), a finite element mechanism model of the fuel cell stack assembly process is constructed, including a force behavior model, an electrical behavior model, a thermal behavior model, and a water vapor transmission behavior model; 603) Solve the model in step 602), build a simulation data set based on the results, and use the data set in combination with deep learning methods to establish a fuel cell stack multi-physical field behavior analysis agent model to improve the efficiency of the mechanism model.

7. A digital assembly system for fuel cell stacks according to claim 6, characterized in that: The force behavior model is divided into a local model and an overall model. The local model analyzes the influence of component size errors and main assembly parameters on the pressure distribution inside the battery stack, and the overall model is used to evaluate the influence of end plate deformation and materials on the mechanical state of the battery stack; the electrical behavior model mainly analyzes the interface contact resistance and body resistance of the fuel cell; the thermal behavior model mainly analyzes the interface thermal conductivity and body thermal conductivity inside the battery stack; the water vapor transmission behavior model mainly analyzes the water vapor transmission resistance inside the battery stack; The construction of the finite element mechanism model of the fuel cell stack assembly process mainly includes the following steps: 701) The force behavior model is constructed as follows: based on the geometric model of step 601), structural simplification, material setting, boundary setting and meshing are performed in sequence to obtain the influence of component size errors and other main assembly parameters on the internal pressure distribution of the fuel cell stack and the overall mechanical state of the fuel cell stack; 702) The electrical behavior model is constructed as follows: establishing an assembly geometry model of the interface, defining the material properties of the contact components and setting boundary conditions, and combining the force behavior model established in step 801) on the basis of obtaining the relationship between the interface contact resistance and the contact pressure, thereby obtaining an "electrical-mechanical" coupled assembly model; 703) The thermal behavior model is constructed as follows: establishing an assembly geometry model of the battery stack, defining material properties and setting boundary conditions, and combining the force behavior model established in step 801) on the basis of obtaining the relationship between the contact thermal resistance and contact pressure inside the battery stack, thereby obtaining a "thermal-mechanical" coupled assembly model inside the battery stack; 704) The water vapor transmission behavior model is constructed as follows: establish an assembly geometry model of the fuel cell stack, define material properties and set boundary conditions, and obtain the relationship between the water vapor transmission resistance and contact pressure inside the fuel cell stack in combination with the force behavior model established in step 701), thereby obtaining an assembly model of water vapor transmission inside the fuel cell stack.

8. A digital assembly system for fuel cell stacks according to claim 4 or 7, characterized in that: The service model sets corresponding process analysis targets based on the fuel cell multi-physical field behavior analysis model, including component forming quality, component stacking sequence and stack press displacement, and then constructs it in combination with quality assessment methods and multi-objective optimization algorithms to analyze component processing errors, stack assembly errors, stack press displacement and stack performance, thereby completing process guidance for the entire stack assembly process.

9. A digital assembly system for fuel cell stacks according to claim 3, characterized in that: Each model in the model layer is integrated and managed through a unified modeling method, and the product is expressed abstractly based on the object-oriented idea. According to the properties of components, production process parameters and multi-physical field coupling characteristics, the fuel cell stack entity class, multi-physical field behavior analysis class and stack assembly service class are constructed respectively. The fuel cell assembly process is regarded as the mutual transmission of messages between the three types of objects, thereby realizing model integration and coordinated operation: the attributes of the fuel cell stack entity class include the geometric information of the components of the stack, the material properties and the matching characteristics of the assembly relationship between the components, and the method is a contact model that describes the component assembly process; the attributes of the multi-physical field behavior analysis class are the relevant parameters and results corresponding to the multi-physical field behavior analysis model, and the method is a parameterized multi-physical field behavior prediction model; the main methods of the stack assembly service class include forming quality evaluation method, assembly accuracy calculation method and assembly process optimization method, and its attributes are the main parameters that constitute the evaluation method, including contact resistance and contact pressure distribution standard deviation.

10. A digital assembly method for a fuel cell stack using any of the systems described in 1-9 above, characterized in that: The following steps are involved: 1) Parts screening: Obtain the geometric shape and surface morphology of the parts, establish the corresponding parts data model, evaluate the parts processing quality based on the parts data model, screen out qualified parts and perform quality grading, and call the stacking analysis module, press-fit analysis module and performance analysis module based on the parts data model to formulate a preliminary parts assembly sequence and stack press-fit displacement; 2) Component stacking: Use the assembly sequence obtained in step 1) to guide component stacking, establish a stacking process data model, collect assembly errors in real time during the stacking process, gradually improve the stacking process data model, and update the assembly sequence of the components to be stacked in real time by calling the stacking analysis module until the stacking of the entire stack of components is completed; 3) Stack press assembly: Based on the stacking process data model obtained in the assembly sequence of step 2), the stack press assembly analysis module and the performance analysis module are called to update the stack press assembly displacement, and a press assembly process data model is established. During the press assembly process, the stack assembly parameters are collected in real time, and the press assembly process data model is gradually updated. The press assembly displacement is updated in real time by calling the press assembly analysis module and the performance analysis module until the stack press assembly is completed; 4) Stack test: Perform performance tests on the assembled stack and collect performance data. Compare the actual performance data of the stack with the predicted data of the performance model in the digital space. Calibrate the accuracy of the stack assembly analysis model and the stack performance prediction model to provide more accurate guidance for subsequent stack assembly, and finally obtain a fully assembled fuel cell stack.

Images