Method and device for determining the shape parameters of a fluid path

Abstract

The present disclosure provides a method for determining the shape parameters of a fluid path, comprising: obtaining first simulation data, wherein the first simulation data correlates with the properties of the fluid when flowing in the fluid path with first shape parameters; providing the first simulation data to a machine learning model in response to the simulation conditions being met to obtain second shape parameters of the fluid path; and providing the second shape parameters to a simulation system to obtain second simulation data, wherein the second simulation data correlates with the properties of the fluid when flowing in the fluid path with the second shape parameters.

Description

FIELD OF THE INVENTION

[0001] The present disclosure relates generally to computer-aided design and, more particularly, to a method and apparatus for determining the shape parameters of a fluid path. STATE OF THE ART

[0002] Computer-aided design (CAD) technology is already used in many fields, such as industrial manufacturing, construction, and aerospace, among others. In fluid flow analysis, a fluid path is designed using CAD software (e.g., Creo), followed by simulation analysis using computer-aided engineering (CAE) software (e.g., ANSYS Fluent). This significantly increases efficiency in the design, manufacture, and maintenance of products, resulting in considerable economic benefits.

[0003] However, fluid path design typically requires determining a large number of fluid path shape parameters. During the design and simulation processes using CAD and CAE software, determining the appropriate shape parameters often requires manually modifying one or more parameters to perform large amounts of simulation, which requires significant time and effort. At the same time, due to the complex interrelationships between parameters, designers must have a certain degree of experience to determine the correct direction for parameter optimization, which further complicates the design process.

[0004] For this reason, a highly efficient method and apparatus for determining the shape parameters of a fluid path should be provided in order to quickly and precisely determine the desired shape parameters of fluid paths while reducing manual interventions. DISCLOSURE OF THE INVENTION

[0005] According to the first aspect of the present disclosure, a method for determining the shape parameters of a fluid path is provided, comprising: obtaining first simulation data, wherein the first simulation data correlates with the properties of the fluid when flowing in the fluid path with first shape parameters; providing the first simulation data to a machine learning model in response to the simulation conditions being satisfied to obtain second shape parameters of the fluid path; and providing the second shape parameters to a simulation system to obtain second simulation data, wherein the second simulation data correlates with the properties of the fluid when flowing in the fluid path with the second shape parameters.

[0006] According to another aspect of the present disclosure, there is provided an apparatus for determining the shape parameters of a fluid path, comprising: a memory; and a processor coupled to the memory, the processor configured to perform a method according to the present disclosure.

[0007] According to another aspect of the present disclosure, a computer-readable medium is provided having stored thereon a computer program comprising instructions, the instructions, when executed by a processor, causing the processor to be configured to perform a method according to the present disclosure.

[0008] According to another aspect of the present disclosure, an apparatus for determining the shape parameters of a fluid path is provided, comprising: a module for obtaining first simulation data, wherein the first simulation data correlates with the properties of the fluid when flowing in the fluid path with the first shape parameters; a module for providing the first simulation data to a machine learning model in response to the simulation conditions being met to obtain second shape parameters of the fluid path; and a module for providing the second shape parameters to a simulation system to obtain second simulation data, wherein the second simulation data correlates with the properties of the fluid when flowing in the fluid path with the second shape parameters.

[0009] According to another aspect of the present disclosure, a computer program product is provided, the computer program product comprising a plurality of instructions, the instructions, when executed by a processor, causing the processor to perform the following operations: obtaining first simulation data, wherein the first simulation data correlates with the properties of the fluid when flowing in the fluid path with first shape parameters; providing the first simulation data to a machine learning model in response to the simulation conditions being met to obtain second shape parameters of the fluid path; and providing the second shape parameters to a simulation system to obtain second simulation data, wherein the second simulation data correlates with the properties of the fluid when flowing in the fluid path with the second shape parameters. DESCRIPTION OF THE CHARACTERS

[0010] Various embodiments of the claimed subject matter will now be described by way of example with reference to the figures. In different figures, the same reference numerals are used to refer to the same or similar parts. Fig. 1 shows a schematic representation of a system 100 for determining the shape parameters of a fluid path using a machine learning model according to an embodiment of the present disclosure. Fig. 2 shows a schematic diagram for determining the shape parameters of the fluid path in the distribution zone of a bipolar plate of the cell stack of a fuel cell according to an embodiment of the present disclosure. Fig. 3 shows a schematic diagram for determining the shape parameters of the fluid path in the reaction zone of a bipolar plate of the cell stack of a fuel cell according to an embodiment of the present disclosure. Fig. 4 shows a flowchart of a method 400 for determining the shape parameters of a fluid path using a machine learning model according to an embodiment of the present disclosure. Fig. 5 shows a block diagram of an apparatus 500 for determining the shape parameters of a fluid path according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] In the following description, numerous specific details are set forth in order to provide a thorough understanding of embodiments of the present disclosure. However, those skilled in the art will recognize that the present disclosure may be practiced without one or more of the specific details, or using alternative methods, components, etc. In some embodiments, well-known structures and operations are not shown or described in detail in order to avoid unnecessarily obscuring the present disclosure.

[0012] Fig. Figure 1 shows a schematic representation of a system 100 for determining the shape parameters of a fluid path using a machine learning model according to an embodiment of the present disclosure. The system 100 is composed of a machine learning model unit 110 and a simulation system 120.

[0013] The machine learning model unit 110 serves to provide the shape parameters of a fluid path to the simulation system 120. For example, the machine learning model unit 110 can provide the simulation system 120 with a set of initial shape parameters according to the user's instructions. This set of initial shape parameters describes properties specified by the user relating to the shape of the fluid path, such as the length and width of one or more flow channels, the radius of the critical openings in the flow channels, etc.

[0014] The simulation system 120 is used to modify the fluid path model based on this set of initial shape parameters and to perform simulation operations based on the modified fluid path model to obtain simulation data. The user can specify what type of fluid property data they want to obtain. For example, the user may want to obtain information about pressure and temperature changes of the fluid as it flows through the flow channels. In this example, the simulation system 120 can determine, based on the received initial shape parameters, simulation data corresponding to these initial shape parameters regarding the pressure and temperature distribution of the fluid as it flows through the flow channels.

[0015] The simulation data is then obtained using the machine learning model unit 110, and in response to the fulfillment of the simulation conditions, the simulation data is provided to a machine learning model 150. The simulation conditions are the conditions that must be met to continue the simulations. For example, the machine learning model unit 110 can determine whether the simulation data has reached the threshold conditions (e.g., the design requirements regarding the pressure and temperature distribution in the flow channels), and if it has not yet reached the threshold conditions, subsequent simulation operations must be performed. The machine learning model 150 is integrated with the machine learning model unit 110 and uses machine learning algorithms, such as Bayesian optimization, etc.to perform calculations based on the simulation data and thus determine shape parameters that, after updating, are better than the initial shape parameters. In this context, "better" means that, compared to the initial shape parameters, the updated shape parameters are more likely to achieve the user's desired fluid flow properties. For example, in the example above, the simulation data regarding the pressure and temperature distribution of the fluid in the flow channels that the user obtains using these updated shape parameters may be more consistent with the design requirements.

[0016] The machine learning model unit 110 is then used to provide the updated shape parameters to the simulation system 120, where the simulation system 120 again modifies the fluid path model based on these updated shape parameters and performs simulation processes. By repeating the above-described process several times, the machine learning model unit 110 can continuously update the shape parameters provided to the simulation system 120 and thus obtain the simulation data from the simulation system 120 accordingly. By collecting and learning the simulation data, the machine learning model unit 110 can analyze and calculate the shape parameters provided to the simulation system 120 in the next iteration.

[0017] In one embodiment, the simulation system 120 may further comprise a modeling unit 130 and a simulation unit 140 for performing modeling and simulation operations, respectively. For example, the machine learning model unit 110 may provide the shape parameters to the modeling unit 130 to modify the shape parameters of the fluid path model. The modeling unit 130 may then import the modified model into the simulation unit 140 to perform simulation operations to obtain simulation data. In one embodiment, the modeling unit 130 and the simulation unit 140 may be modular units independent of one another and / or located on different computers. For example, the simulation unit 140 often requires greater computing power, so it may run on a high-performance computer or in a cloud.It should be understood that the individual components of the machine learning model unit 110 and the simulation system 120 may be located on one computer or on multiple computers communicating with each other in any combination without limitation.

[0018] In one embodiment, the machine learning model unit 110 may repeatedly provide simulation data to the machine learning model 150 to obtain updated shape parameters, so that the simulation conditions described above may further include the number of repeated provision of simulation data to the machine learning model 150 being less than a threshold number. For example, when the number of shape parameter updates reaches the threshold number (e.g., 20 times, 25 times), the machine learning model unit 110 may stop executing further iterations.

[0019] In one embodiment, the machine learning model 150 in the machine learning model unit 110 may further utilize neural network algorithms. In this embodiment, large data sets may initially need to be used to train the neural network.

[0020] Compared to the manual analysis of simulation results and parameter modification in conventional design, the solution according to the embodiments of the present disclosure can significantly reduce designer intervention in the simulation process, thus reducing the time and effort required. Furthermore, since machine learning techniques determine the direction of parameter optimization, the parameter optimization process is further simplified, enabling rapid and precise determination of fluid path shape parameters that meet design requirements.

[0021] Fig. 2 shows a schematic diagram for determining the shape parameters of the fluid path in the distribution zone of a bipolar plate of the cell stack of a fuel cell according to an embodiment of the present disclosure.

[0022] The bipolar plate in the cell stack of a fuel cell is one of the core components for carrying out an electrochemical reaction. The bipolar plate not only provides the flow channels for the reaction gases (e.g., hydrogen and air), but also has the task of conducting electrons and supporting the cell stack. Fig. 2, the fluid inlet part of an exemplary bipolar plate 200 is shown, wherein for clarity in Fig. 2, the outlet portion to the left of the dashed line has been omitted. Three fluid inlets are shown on the far right of the bipolar plate 200: the air inlet 230, the coolant inlet 240, and the hydrogen inlet 250, which supply air, coolant (e.g., water), and hydrogen, respectively, to the reaction zone 210 of the bipolar plate. Reaction zone 210 contains several rows of parallel flow channels. Air, coolant, and hydrogen each flow through their respective flow channels and, with the aid of the membrane electrode bonded to the bipolar plate, carry out the electrochemical reaction to generate electrical energy.One of the design requirements for bipolar plates is to ensure the uniform distribution of the reaction gases and cooling liquid in the individual flow channel sets throughout the reaction zone 210 to prevent a reduction in reaction efficiency due to fluid flow differences in the flow channels. To achieve this goal, a distribution zone is further arranged between the fluid inlets of the bipolar plate and the reaction zone 210. In . Fig. 2 shows an example of an arrangement of the distribution zone 220 for the cooling liquid, wherein the cooling liquid 240 first passes through a set of grid cells 260 into the distribution zone 220. The cooling liquid then passes through guide openings 270 of various sizes into the guide channels 280, wherein the guide channels 280 guide the cooling liquid into the cooling liquid flow channels located at various positions in the reaction zone 210 of the bipolar plate. By adjusting the width of the grid cells 260 and the radius of the guide openings 270, a uniform distribution of the cooling liquid flow in the reaction zone 210 can be achieved. It should be noted that the Fig. The grid cells 260 and guide openings 270 shown in Figure 2 are only examples. A uniform distribution of the cooling fluid flow can be achieved by adapting the dimensions of any other shapes of grid cells 260 and guide openings 270.

[0023] In this embodiment, the Fig. The simulation system 120 shown in FIG. 1 contains a set of initial grid cell dimensions and guide opening dimensions. Based on this parameter set, the simulation system 120 modifies the distribution zone model and performs simulation operations based on the modified distribution zone model to obtain simulation data. This simulation data serves to indicate the fluid flow distribution in the reaction zone 210. Subsequently, the simulation data is obtained from the simulation system 120 using the machine learning model unit 110, and the simulation data is provided to the machine learning model 150.The machine learning model 150 performs calculations based on the simulation data to determine the dimensions of the grid cells 260 and the guide openings 270 that may provide a uniform fluid flow distribution in the reaction zone 210, and provides this updated dimensional data to the simulation system 120 to further modify the distribution zone model and perform simulation operations. By repeating the above-described process several times, the machine learning model unit 110 can continuously update the dimensions of the grid cells 260 and the guide openings 270 provided to the simulation system 120, thus obtaining the simulation data of the simulation system 120 accordingly.During this process, the dimensions of the grid cells 260 and the guide openings 270 are continuously optimized until the number of iterations reaches the threshold number or until the standard deviation of the flow rate of all flow channels is less than a threshold standard deviation. The above-described procedure of the present disclosure achieves automatic adjustment of the dimensions of the grid cells 260 and the guide openings 270, which increases design efficiency.

[0024] Fig. 3 shows a schematic diagram for determining the shape parameters of the fluid path in the reaction zone of a bipolar plate of the cell stack of a fuel cell according to an embodiment of the present disclosure.

[0025] In Fig. 3 shows a portion of the cross-sectional area of ​​the reaction zone of an exemplary bipolar plate 300, for example, the bipolar plate shown in Fig. 3 shown cross-sectional area by the cross-sectional area at the dashed line from Fig. 2, comprising an anode plate 330, a membrane electrode 340 and a cathode plate 350. It should be understood that the reaction zone of the cell stack of the fuel cell may comprise a plurality of stacked bipolar plates, wherein in Fig. 3, for the sake of clarity, an arrangement with only one set of anode plate, membrane electrode, and cathode plate is shown. In dashed box 310, the part enclosed by the anode plate 330 and the membrane electrode 340 is an air flow channel; the part enclosed by the cathode plate 350 and the membrane electrode 340 is a hydrogen flow channel. The dashed box 320 shows the part where the anode plate 330 and the cathode plate 350 are bonded to the membrane electrode 340, which is referred to as the web in each pole plate and is the flow channel for the cooling fluid. In an electrochemical reaction, the dimensions of the flow channels of each fluid correlate with the reaction efficiency. To achieve optimal current density, the dimensions of the flow channels must be planned.The shape parameters influencing the current density include, as explained using the anode plate 330 as an example, the width t1 of the oxygen flow channel, its depth h1, the width t2 of the web, and the taper angle α of the connection between the oxygen flow channel and the web. It should be understood that the cathode plate 350 also has the corresponding shape parameters of flow channel width, flow channel depth, web width, and taper angle, etc.

[0026] In this embodiment, the Fig. The simulation system 120 shown in FIG. 1 contains a set of initial parameters: flow channel width, flow channel depth, land width, and chamfer angle. Based on this set of parameters, the simulation system 120 modifies the flow channel model and performs simulation operations based on the modified flow channel model to obtain simulation data. This simulation data serves to indicate the current density of the cell stack of the fuel cell. Subsequently, the simulation data is obtained from the simulation system 120 using the machine learning model unit 110, and the simulation data is provided to the machine learning model 150.The machine learning model 150 performs calculations based on the simulation data to determine the flow channel width, flow channel depth, land width, and draft angle that may result in greater flow density, and provides this updated shape data to the simulation system 120 to further modify the flow channel model and perform simulation operations. By repeating the above-described process several times, the machine learning model unit 110 can continuously update the flow channel width, flow channel depth, land width, and draft angle parameters provided to the simulation system 120, thus obtaining the simulation data from the simulation system 120 accordingly.During this process, the flow channel width, flow channel depth, land width, and draft angle are continuously optimized until the number of iterations reaches the threshold number or until the current density of the fuel cell stack is greater than the current density threshold. The above-described approach of the present disclosure achieves automatic adjustment of the flow channel width, flow channel depth, land width, and draft angle, increasing design efficiency.

[0027] It should be noted that in Fig. 2 and Fig. 3 illustrates two exemplary embodiments for the application of the concept of the present disclosure. The exemplary embodiments of the present disclosure are by no means limited to the embodiments described above, but encompass any other environments in which fluid paths are designed.

[0028] Fig. 4 shows a flowchart of a method 400 for determining the shape parameters of a fluid path using a machine learning model according to an embodiment of the present disclosure. In step S410, the method includes obtaining first simulation data, wherein the first simulation data correlates with the properties of the fluid when flowing in the fluid path with the first shape parameters. In one embodiment, the fluid path is a path for the fluid in the bipolar plate of the cell stack of a fuel cell.

[0029] In step S420, the method includes providing the first simulation data to the machine learning model in response to the simulation conditions being met to obtain second shape parameters of the fluid path. In one embodiment, the machine learning model utilizes Bayesian optimization.

[0030] In step S430, the method comprises providing the second shape parameters to the simulation system to obtain second simulation data, wherein the second simulation data correlate with the properties of the fluid when flowing in the fluid path with the second shape parameters. In one embodiment, the simulation system comprises a modeling unit and a simulation unit, and the method further comprises providing the second shape parameters to the modeling unit, wherein the modeling unit modifies the model of the fluid path based on the second shape parameters, and the simulation unit performs a simulation based on the modified model to obtain the second simulation data. In one embodiment, the modeling unit and the simulation unit run on different computers.In one embodiment, the first shape parameters and the second shape parameters are one or more of the following parameters: the dimensions of the grid cells through which the fluid flows upon entering the distribution zone, the dimensions of the guide openings through which the fluid flows upon entering the guide channel, wherein the first simulation data and the second simulation data correlate with the fluid flow distribution in the reaction zone. In one embodiment, the first shape parameters and the second shape parameters are one or more of the following parameters: flow channel width, flow channel depth, land width, and draft angle, wherein the first simulation data and the second simulation data correlate with the electrical current density of the cell stack of the fuel cell.In one embodiment, method 400 further includes providing the initial parameters to the simulation system to obtain the first simulation data. In one embodiment, the simulation conditions include the following: the number of times the simulation data is repeatedly provided to the machine learning model is less than a threshold number, or the first simulation data does not meet the threshold conditions.

[0031] Fig. Figure 5 shows a block diagram of a device 500 for determining the shape parameters of a fluid path according to an embodiment of the present disclosure. In one embodiment, the device 500 may comprise a computer, such as a desktop computer, a laptop, a small computer, a cloud server, etc.

[0032] The device 500 includes a processor 502 connected to an internal communications bus 508. The processor 502 is operable to execute the instructions in the memory 504 to implement the method for determining the shape parameters of a fluid path described in detail above. Examples of the processor 502 may be a central processing unit (CPU), a microcontroller, and the like. The memory 504 may include any form of memory, such as DRAM. The device 500 may further include an input interface 510 and an output interface 512. The input interface 510 is operable to receive input signals and data from an input device (e.g., a keyboard or mouse coupled to the computer, etc.). The output interface 512 is operable to send output signals and data to an output device (e.g., a display device).Additionally, the device 500 may further include a non-volatile memory device 506 to physically store computer program instructions and data.

[0033] The various methods, steps, operations, units, modules, components, models, and networks described in connection with this disclosure may be implemented as hardware, software executed by a processor, firmware, or any combination thereof. According to one or more aspects of the present disclosure, the computer program product for determining the shape parameters of a fluid path may include processor-executable computer instructions for implementing one or more of the methods described above with reference to Fig. 4. According to another aspect of the present disclosure, computer instructions for determining the shape parameters of a fluid path may be stored on a computer-readable medium, which instructions, when executed by a processor, may cause a processor to perform one or more of the steps described above with reference to Fig. 4 performs the procedures or steps described.

[0034] A computer-readable medium includes both a non-transitory computer storage medium and a communications medium, where a communications medium includes any medium that facilitates the transfer of a computer program from one location to another. Each connection may accordingly be referred to as a computer-readable medium.

[0035] In addition to what is described herein, various modifications may be made to the disclosed embodiments and implementations without departing from the scope of the disclosed embodiments and implementations. Accordingly, the descriptions and examples contained herein should be interpreted in an illustrative rather than a restrictive sense. The scope of the present disclosure should be judged solely by reference to the claims.

Claims

[1] A method for determining the shape parameters of a fluid path, comprising: Obtaining first simulation data, wherein the first simulation data correlates with the properties of the fluid when flowing in the fluid path with first shape parameters; Providing the first simulation data to a machine learning model in response to the satisfaction of the simulation conditions to obtain second shape parameters of the fluid path; and Providing the second shape parameters to a simulation system to obtain second simulation data, wherein the second simulation data correlates with the properties of the fluid when flowing in the fluid path having the second shape parameters. [2] The method of claim 1, wherein the fluid path is a path for the fluid in a bipolar plate of the cell stack of a fuel cell. [3] The method of claim 2, wherein the first shape parameters and the second shape parameters are one or more of the following parameters: the dimensions of the grid cells through which the fluid flows upon entering the distribution zone, the dimensions of the guide openings through which the fluid flows upon entering the guide channel, wherein the first simulation data and the second simulation data correlate with the fluid flow distribution in the reaction zone. [4] The method of claim 2, wherein the first shape parameters and the second shape parameters are one or more of the following parameters: Flow channel width, flow channel depth, web width and chamfer angle, where the first simulation data and the second simulation data correlate with the electrical current density of the cell stack of the fuel cell. [5] The method of claim 1, wherein the machine learning model uses Bayesian optimization. [6] The method of claim 1, wherein the simulation system comprises a modeling unit and a simulation unit, and the method further comprises: Providing the second shape parameters to the modeling unit, wherein the modeling unit modifies the model of the fluid path based on the second shape parameters and the simulation unit performs a simulation based on the modified model to obtain the second simulation data. [7] Method according to claim 6, wherein the modeling unit and the simulation unit run on different computers. [8] The method of claim 1, further comprising providing the initial parameters to the simulation system to obtain the first simulation data. [9] The method of claim 1, wherein the simulation conditions comprise: the number of times the simulation data is repeatedly provided to the machine learning model is less than a threshold number, or the first simulation data does not meet the threshold conditions. [10] Apparatus for determining the shape parameters of a fluid path, comprising: a memory; a processor coupled to the memory, the processor configured to perform a method according to any one of claims 1 to 9. [11] A computer-readable medium having stored thereon a computer program comprising instructions, the instructions, when executed by a processor, causing the processor to be configured to perform a method according to any one of claims 1 to 9. [12] Apparatus for determining the shape parameters of a fluid path, comprising: a module for obtaining first simulation data, wherein the first simulation data correlates with the properties of the fluid when flowing in the fluid path with first shape parameters; a module for providing the first simulation data to a machine learning model in response to the satisfaction of the simulation conditions to obtain second shape parameters of the fluid path; and a module for providing the second shape parameters to a simulation system to obtain second simulation data, wherein the second simulation data correlate with the properties of the fluid when flowing in the fluid path with the second shape parameters. [13] A computer program product, the computer program product comprising a plurality of instructions, the instructions, when executed by a processor, causing the processor to perform the following operations: Obtaining first simulation data, wherein the first simulation data correlates with the properties of the fluid when flowing in the fluid path with first shape parameters; Providing the first simulation data to a machine learning model in response to the satisfaction of the simulation conditions to obtain second shape parameters of the fluid path; and Providing the second shape parameters to a simulation system to obtain second simulation data, wherein the second simulation data correlates with the properties of the fluid when flowing in the fluid path having the second shape parameters.

Images