A method and system for constructing a shunt network simulation model for a liquid-cooled cold plate

Abstract

The present invention provides a method and system for constructing a shunt network simulation model of a liquid-cooled cold plate. Through step S1, the distribution positions of IT equipment components can be obtained and a two-dimensional IT simulation plane can be established. Through step S2, the cold plate constraint conditions can be obtained and a two-dimensional network simulation plane can be established. The two-dimensional network simulation plane is marked with a heat source level. Through step S3, the two-dimensional network simulation plane is divided into units and a shunt network connection is performed to generate multiple three-dimensional shunt network models. Through step S4, simulation is performed and a discrimination operation is performed. Based on the discrimination operation, step S5 or step S6 is executed. If step S5 is executed, multiple improved three-dimensional shunt network models are generated based on the shunt improvement model, and step S4 is re-executed. If step S6 is executed, the three-dimensional shunt network model is sorted and output based on the discrimination operation. Designers can refer to the three-dimensional shunt network model to design the shunt network of the liquid-cooled cold plate, thereby improving the R&D rate.

Description

Technical Field

[0001] The present invention belongs to the field of liquid cooling simulation technology, and more specifically, relates to a method and system for constructing a shunt network simulation model of a liquid cooling cold plate. Background Art

[0002] Liquid-cooled cold plates remove heat generated by electronic equipment by flowing coolant in a shunt network. The design of the shunt network directly affects the heat dissipation effect. A reasonable shunt network can ensure even distribution of coolant, avoid the generation of local hot spots, and improve heat dissipation efficiency.

[0003] At present, the design of the shunt network of the liquid-cooled cold plate has relative limitations. The traditional design method mainly relies on the experience of engineers and repeated physical experiments. Physical experiments not only require a lot of time and cost to produce samples, but also it is difficult to accurately control and measure all parameters during the experiment. The present invention provides a model construction method that can simulate the shunt network in the liquid-cooled cold plate, thereby providing auxiliary design for designers and improving the research and development design rate of the liquid-cooled cold plate. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides a method and system for constructing a shunt network simulation model for a liquid-cooled cold plate. The purpose and efficacy of the method and system for constructing a shunt network simulation model for a liquid-cooled cold plate of the present invention are achieved by the following specific technical means:

[0005] A method for constructing a flow distribution network simulation model of a liquid-cooled cold plate includes the following steps:

[0006] S1: Obtaining the distribution positions of IT equipment components, and establishing a two-dimensional IT simulation plane based on the distribution positions of the IT equipment components;

[0007] S2: Acquire cold plate constraints, where the cold plate constraints represent restrictions on the installation of a liquid-cooled cold plate on IT equipment; establish a two-dimensional network simulation plane based on the cold plate constraints and the two-dimensional IT simulation plane; and mark the two-dimensional network simulation plane with a heat source level;

[0008] S3: Dividing the two-dimensional network simulation plane into units, obtaining shunt nodes, performing shunt network connections based on the shunt nodes, and generating multiple three-dimensional shunt network models;

[0009] S4: Simulate starting the IT equipment and injecting coolant into multiple three-dimensional flow distribution network models, obtain multiple simulation data matrices, perform a discrimination operation on the multiple simulation data matrices, and execute step S5 or step S6 based on the discrimination operation;

[0010] S5: importing the multiple three-dimensional flow diversion network models into the improved flow diversion model respectively, generating multiple improved three-dimensional flow diversion network models based on the improved flow diversion model, and executing step S4;

[0011] S6: Sort and output the three-dimensional diversion network model based on the discriminant operation.

[0012] As a further solution of the present invention, establishing a two-dimensional network simulation plane based on the cold plate constraint condition and the two-dimensional IT simulation plane, and marking the two-dimensional network simulation plane with a heat source level includes:

[0013] establishing a two-dimensional cold plate simulation plane based on the cold plate constraint condition and the two-dimensional IT simulation plane, and overlapping the two-dimensional cold plate simulation plane with the two-dimensional IT simulation plane to obtain the two-dimensional network simulation plane;

[0014] Obtaining the operating power of IT equipment components based on the IT equipment, and meshing the two-dimensional network simulation plane into a plurality of rectangular liquid cooling grids of equal area;

[0015] Establishing a liquid cooling two-dimensional coordinate system based on the two-dimensional network simulation plane, mapping the plurality of rectangular liquid cooling grids and the distribution positions of the IT equipment components into the liquid cooling two-dimensional coordinate system, using the intersection of the lower right corner edges of the rectangular liquid cooling grids as the X-axis coordinate, and using the intersection of the upper right corner edges of the rectangular liquid cooling grids as the Y-axis coordinate;

[0016] Based on the distribution position of the IT equipment components, each rectangular liquid cooling grid in the liquid cooling two-dimensional coordinate system is traversed and queried, and the rectangular liquid cooling grid is marked with a heat source level according to the operating power of the IT equipment components. The heat source level mark includes a zero-level heat source mark, a first-level heat source mark, a second-level heat source mark, and a third-level heat source mark.

[0017] As a further solution of the present invention, the traversal query of each rectangular liquid cooling grid in the liquid cooling two-dimensional coordinate system based on the distribution position of the IT equipment components and the heat source level marking of the rectangular liquid cooling grid according to the operating power of the IT equipment components include:

[0018] Obtaining a first ratio value of each rectangular liquid cooling grid containing the IT equipment component distribution position, and setting the first ratio value to zero if the rectangular liquid cooling grid does not contain the IT equipment component distribution position;

[0019] Obtaining a second ratio value of each rectangular liquid cooling grid containing the two-dimensional cold plate simulation plane, and setting the second ratio value to zero if the rectangular liquid cooling grid does not contain the two-dimensional cold plate simulation plane;

[0020] Obtaining a third proportional value based on the first proportional value and the operating power of the IT equipment component, and setting the third proportional value to zero if the first proportional value is zero;

[0021] Each of the rectangular liquid cooling grids is marked with a heat source level based on the first ratio value, the second ratio value, and the third ratio value.

[0022] As a further embodiment of the present invention, the method further comprises:

[0023] Each of the rectangular liquid cooling grids includes a first proportional value, a second proportional value, and a third proportional value;

[0024] The zero-level heat source mark indicates an area with no self-heating temperature after the IT equipment is in operation. If any of the first ratio value, the second ratio value, and the third ratio value is zero, the rectangular liquid cooling grid is marked as a zero-level heat source.

[0025] The first-level heat source mark indicates the area where the self-heating temperature of the IT equipment is low after operation, the second-level heat source mark indicates the area where the self-heating temperature of the IT equipment is normal after operation, and the third-level heat source mark indicates the area where the self-heating temperature of the IT equipment is high after operation;

[0026] If the first ratio value, the second ratio value, and the third ratio value are all non-zero, the rectangular liquid cooling grid is marked as a first-level heat source, a second-level heat source, or a third-level heat source based on a set ratio value range.

[0027] As a further solution of the present invention, the two-dimensional network simulation plane is divided into units, shunt nodes are obtained, shunt network connections are made based on the shunt nodes, and multiple three-dimensional shunt network models are generated, including:

[0028] Performing unit division on the two-dimensional network simulation plane based on the Voronoi diagram algorithm to divide the two-dimensional network simulation plane into a plurality of the diversion nodes;

[0029] Numbering the plurality of shunt nodes based on the heat source level mark, the node numbers including a first category number, a second category number, a third category number, and a fourth category number; defining the plurality of shunt nodes based on the node numbers as trunk nodes, shunt nodes, gross nodes, and irrelevant nodes;

[0030] Performing shunt network connections based on the trunk nodes, the shunt nodes, and the gross and detailed points to generate multiple two-dimensional shunt network models;

[0031] A plurality of three-dimensional flow shunt network models are generated based on the plurality of two-dimensional flow shunt network models and the cold plate constraint conditions.

[0032] As a further solution of the present invention, the shunting network connection based on the trunk node, the shunting node and the capillary point is generated to generate multiple two-dimensional shunting network models, including:

[0033] The node numbers of the trunk nodes, the shunt nodes and the hairy and detailed points are grouped into a node number data matrix, wherein the node number data matrix includes the node numbers corresponding to the trunk nodes, the shunt nodes and the hairy and detailed points;

[0034] Randomly sampling and extracting the numbered data matrix based on a Monte Carlo algorithm, and generating a plurality of randomly numbered data matrices, wherein the randomly numbered data matrix represents the connection order of the trunk node, the diversion node, and the hair detail point;

[0035] Based on the multiple randomly numbered data matrices, a shunt network connection is performed to generate multiple two-dimensional shunt network models.

[0036] As a further solution of the present invention, the plurality of shunt nodes are numbered based on the heat source level mark, the node numbers including a first category number, a second category number, a third category number, and a fourth category number. Based on the node numbers, the plurality of shunt nodes are defined as trunk nodes, shunt nodes, gross detail points, and irrelevant nodes, respectively, including:

[0037] The first category number is represented by , wherein A represents the zero-level heat source mark included in the heat source level mark, X represents the Xth shunt node, and the shunt node with the first category number is defined as an irrelevant node, and the irrelevant node represents a node that is irrelevant to the shunt network;

[0038] The second category number is represented by , where B represents the first-level heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node of the second category number is defined as a capillary detail point, which is represented by a node in the shunt network for establishing a capillary cooling channel;

[0039] The third category number is represented by , where C represents the secondary heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node with the third category number is defined as a shunt node, and the shunt node represents a node in the shunt network for establishing a shunt cold channel;

[0040] The fourth category number is represented by , where D represents the third-level heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node with the fourth category number is defined as the trunk node, which is represented by the node used to establish the trunk cold channel in the shunt network.

[0041] As a further embodiment of the present invention, the method of simulating starting IT equipment and injecting coolant into multiple three-dimensional shunt network models, obtaining multiple simulation data matrices, performing a discrimination operation on the multiple simulation data matrices, and executing step S5 or step S6 based on the discrimination operation includes:

[0042] The simulation data matrix includes a plurality of simulation data vectors, and the plurality of simulation data vectors respectively correspond to a plurality of the three-dimensional shunt network models;

[0043] Acquire a design expected data vector, and sequentially compare a plurality of the simulation data vectors in the simulation data matrix with the design expected data vector;

[0044] If the simulation data vector satisfies the design expected data vector, the simulation data vector is derived, and a deviation value between the simulation data vector and the design expected data vector is calculated, and the three-dimensional shunt network model is sorted based on the deviation value, and step S6 is executed;

[0045] If the plurality of simulation data vectors do not meet the design expected data vector, step S5 is executed.

[0046] As a further solution of the present invention, the plurality of three-dimensional shunt network models are respectively imported into the shunt improvement model, a plurality of improved three-dimensional shunt network models are generated based on the shunt improvement model, and step S4 is performed, including:

[0047] Acquiring historical experimental data, wherein the historical experimental data represents data of past designs and experimental simulations of a flow distribution network within a liquid-cooled cold plate;

[0048] An improved model to be trained is constructed based on a convolutional neural network, and the historical experimental data is imported into the improved model to be trained as training data. When the improved model to be trained reaches the maximum number of iterations, the improved diversion model is exported, and multiple three-dimensional diversion network models are imported into the improved diversion model to generate multiple improved three-dimensional diversion network models.

[0049] A system for constructing a flow distribution network simulation model for a liquid-cooled cold plate, comprising:

[0050] An acquisition module, the acquisition module is used to obtain the distribution position of IT equipment components and cold plate constraints;

[0051] An establishment module, the establishment module being used to establish a two-dimensional IT simulation plane, and to establish a two-dimensional network simulation plane based on the two-dimensional IT simulation plane and the cold plate constraint condition;

[0052] a marking module, the marking module being used to mark the heat source level of the two-dimensional network simulation plane;

[0053] A shunt module, the shunt module is used to divide the two-dimensional network simulation plane into units and generate multiple three-dimensional shunt network models;

[0054] A simulation module is used to simulate starting up IT equipment and injecting coolant into the three-dimensional shunt network model, obtaining multiple simulation data matrices for performing discrimination operations;

[0055] An improved module for generating multiple improved three-dimensional diversion network models;

[0056] The output module is used to sort and output the three-dimensional diversion network model.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] First, step S1 can be used to obtain the distribution position of IT equipment components, and a two-dimensional IT simulation plane can be established based on the distribution position of IT equipment components. According to step S1, the component layout of IT equipment can be understood. Then, step S2 can be used to obtain the cold plate constraint condition, thereby restricting the two-dimensional cold plate simulation plane to prevent the two-dimensional cold plate simulation plane from being incompatible with the IT equipment. A two-dimensional network simulation plane is established based on the cold plate constraint condition and the two-dimensional cold plate simulation plane. Then, step S3 is used to divide the two-dimensional network simulation plane into units, obtain the shunt nodes related to the two-dimensional network simulation plane, and perform shunt network connection to generate multiple three-dimensional shunt network models. Then, step S 4 can simulate the three-dimensional shunt network model to obtain a simulation data matrix, and judge the simulation data matrix through a discrimination operation to determine whether it meets the expected design requirements. Finally, through step S6, the three-dimensional shunt network model that meets the expected design requirements can be output, or the three-dimensional shunt network model can be improved through step S5 to obtain an improved three-dimensional shunt network model, and step S4 can be re-executed. Designers can use the output of multiple three-dimensional shunt network models to assist in the design of the liquid-cooled cold plate shunt network, thereby improving the research and development rate. The output of multiple three-dimensional shunt network models is not limited to a single output, and designers can have more reference solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of the steps of a method for constructing a shunt network simulation model of a liquid-cooled cold plate according to the present invention;

[0060] Figure 2This is a principle block diagram of a method for constructing a shunt network simulation model of a liquid-cooled cold plate according to the present invention;

[0061] Figure 3 It is a schematic diagram of a system for building a shunt network simulation model of a liquid-cooled cold plate according to the present invention. DETAILED DESCRIPTION

[0062] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the technical solutions of the present invention, but are not intended to limit the scope of protection of the present invention.

[0063] Example:

[0064] As attached Figure 1 、 Figure 2 、 Figure 3 As shown:

[0065] The present invention provides a method for constructing a shunt network simulation model of a liquid-cooled cold plate, comprising the following steps:

[0066] Step S1: obtaining the distribution positions of IT equipment components, and establishing a two-dimensional IT simulation plane based on the distribution positions of the IT equipment components.

[0067] Specifically, the distribution positions of IT equipment components, such as the physical coordinates of the CPU or GPU, can be obtained by scanning the IT equipment or using CAD design files. Then, a two-dimensional IT simulation plane can be constructed based on the physical coordinates of the heat-generating core components, and the overall polygonal area of ​​the IT equipment can be mapped.

[0068] S2: Obtain cold plate constraints, where the cold plate constraints represent restrictions on the installation of a liquid-cooled cold plate on IT equipment. Establish a two-dimensional network simulation plane based on the cold plate constraints and the two-dimensional IT simulation plane, and mark the two-dimensional network simulation plane with a heat source level.

[0069] In this embodiment, step S2 includes:

[0070] Step S21 : establishing a two-dimensional cold plate simulation plane based on the cold plate constraint condition and the two-dimensional IT simulation plane, and overlapping the two-dimensional cold plate simulation plane with the two-dimensional IT simulation plane to obtain the two-dimensional network simulation plane.

[0071] Understandably, cold plate constraints, such as bolt hole positions and avoidance areas, can be used to establish a two-dimensional IT simulation plane to avoid incompatibility between the designed liquid-cooling shunt network and the physical structure of the actual IT equipment, and to avoid assembly failures due to spatial interference, such as the connection interfaces of actual IT equipment. By pre-defining these mechanical constraint boundaries, the subsequent shunt network design can be forced to optimize the layout within a manufacturing framework, fundamentally eliminating the cost loss caused by the design being divorced from physical reality.

[0072] Step S22 : obtaining the operating power of IT equipment components based on the IT equipment, and performing grid division on the two-dimensional network simulation plane into a plurality of rectangular liquid cooling grids of equal area.

[0073] Step S23: establish a liquid cooling two-dimensional coordinate system based on the two-dimensional network simulation plane, map the multiple rectangular liquid cooling grids and the distribution positions of the IT equipment components into the liquid cooling two-dimensional coordinate system, and use the intersection of the lower right corner edge lines of the rectangular liquid cooling grid as the X-axis coordinate, and use the intersection of the upper right corner edge lines of the rectangular liquid cooling grid as the Y-axis coordinate.

[0074] In this embodiment, after the two-dimensional grid simulation plane is gridded, a liquid-cooled two-dimensional coordinate system is established in the two-dimensional network simulation plane, and when establishing the liquid-cooled two-dimensional coordinate system, the coordinates of each rectangular liquid-cooled grid are defined, so that subsequent traversal queries can be performed in the order of X-axis coordinates and Y-axis coordinates in the liquid-cooled two-dimensional coordinate system, thereby accelerating the speed of the traversal query. In the actual traversal query, the traversal query is first performed based on the X-axis coordinates of the liquid-cooled two-dimensional coordinate system. When the query of the rectangular liquid-cooled grid of a row of X-axis coordinates is completed, a single rectangular liquid-cooled grid is moved upward based on the Y-axis coordinates of the liquid-cooled two-dimensional coordinate system to continue the traversal query based on the X-axis coordinates of the liquid-cooled two-dimensional coordinate system, and this cycle is repeated until the traversal query is completed.

[0075] Step S24, based on the distribution position of the IT equipment components, each of the rectangular liquid cooling grids in the liquid cooling two-dimensional coordinate system is traversed and queried, and the rectangular liquid cooling grids are marked with a heat source level according to the operating power of the IT equipment components. The heat source level mark includes a zero-level heat source mark, a first-level heat source mark, a second-level heat source mark, and a third-level heat source mark.

[0076] Furthermore, step S24 includes:

[0077] Step S241 : obtaining a first ratio value of each rectangular liquid cooling grid containing the IT equipment component distribution position; if the rectangular liquid cooling grid does not contain the IT equipment component distribution position, setting the first ratio value to zero.

[0078] In this embodiment, when each rectangular liquid cooling grid in the liquid cooling two-dimensional coordinate system is traversed and queried based on the distribution positions of IT equipment components, that is, the first proportion value of each liquid cooling grid unit is obtained, since the distribution positions of IT equipment components in the rectangular liquid cooling grid may occupy all area positions or part of the area positions, or the rectangular liquid cooling grid may not contain any IT equipment components, for example, a part of the CPU of the actual IT equipment occupies all area positions in the rectangular liquid cooling grid, then the first proportion value is 1, and if a part of the CPU of the actual IT equipment only occupies 10% of the position in the rectangular liquid cooling grid, then the first proportion value is 0.1. The rectangular liquid cooling grid can also be regarded as the peripheral heat dissipation area of ​​the core component. When the rectangular liquid cooling grid does not contain any IT equipment components, this rectangular liquid cooling grid does not have a heat dissipation function, and the first proportion value is assigned to zero.

[0079] Step S242 , obtaining a second ratio value of each rectangular liquid cooling grid containing the two-dimensional cold plate simulation plane, and setting the second ratio value to zero if the rectangular liquid cooling grid does not contain the two-dimensional cold plate simulation plane.

[0080] In this embodiment, the second ratio value is represented by the ratio of the rectangular liquid cooling grid containing the two-dimensional cold plate simulation plane. When the rectangular liquid cooling grid does not contain the two-dimensional cold plate simulation plane, it means that the rectangular liquid cooling grid is meaningless, and the second ratio value of the rectangular liquid cooling grid is assigned to zero.

[0081] Step S243: Obtain a third ratio value based on the first ratio value and the operating power of the IT equipment component. If the first ratio value is zero, set the third ratio value to zero.

[0082] In this embodiment, since the power of the core heating components of different IT devices is actually different, the operating power of the IT device components needs to be taken into consideration. When taking this into consideration, the relationship between the first ratio value and the operating power of the IT device components also needs to be considered. The specific relationship needs to be specifically set by the designer during the specific implementation process based on the different power of the core heating components of different IT devices.

[0083] Furthermore, each of the rectangular liquid cooling grids includes a first ratio value, a second ratio value, and a third ratio value.

[0084] Step S244 : marking a heat source level of each rectangular liquid cooling grid based on the first ratio value, the second ratio value, and the third ratio value.

[0085] In this embodiment, the zero-level heat source mark indicates an area with no self-heating temperature after the IT equipment is running. If any of the first ratio value, the second ratio value, and the third ratio value is zero, the rectangular liquid cooling grid is marked as a zero-level heat source.

[0086] In this embodiment, the first-level heat source mark indicates an area with a lower self-heating temperature after the IT equipment is running, the second-level heat source mark indicates an area with a normal self-heating temperature after the IT equipment is running, and the third-level heat source mark indicates an area with a higher self-heating temperature after the IT equipment is running.

[0087] In this embodiment, if the first ratio value, the second ratio value and the third ratio value are all non-zero, the rectangular liquid cooling grid is marked as a first-level heat source, a second-level heat source or a third-level heat source based on the set ratio value range.

[0088] It can be understood that using heat source level marking to mark the two-dimensional network simulation plane in advance can facilitate the subsequent definition of nodes after the two-dimensional network simulation plane is divided. The heat source level marking can be completed in advance as a prerequisite work to improve the subsequent simulation speed.

[0089] Specifically, first, the two-dimensional IT simulation plane is established based on the cold plate constraint conditions to ensure that the two-dimensional cold plate simulation plane is compatible with the actual IT equipment, avoid the problem that the cold plate cannot be adapted to the actual IT equipment after the design is completed, and reduce cost losses. Then, the two-dimensional network simulation plane is gridded and analyzed in blocks. In the division process, the area of ​​each rectangular liquid cooling grid is ensured to be equal to prevent the dimensional differences in the subsequent first ratio value, second ratio value, and third ratio value. After the grid division is completed, a liquid cooling two-dimensional coordinate system can be established for the two-dimensional network simulation plane to assign coordinates to each rectangular liquid cooling grid, thereby accelerating the subsequent traversal query speed. Finally, the rectangular liquid cooling grid is traversed and queried based on the liquid cooling two-dimensional coordinate system to obtain the first ratio value, second ratio value, and third ratio value of each rectangular liquid cooling grid, so as to mark the heat source level of the two-dimensional cold plate simulation plane in advance.

[0090] Step S3: divide the two-dimensional network simulation plane into units, obtain shunt nodes, perform shunt network connections based on the shunt nodes, and generate multiple three-dimensional shunt network models.

[0091] In this embodiment, step S3 includes:

[0092] Step S31 : dividing the two-dimensional network simulation plane into units based on a Voronoi diagram algorithm, and dividing the units into a plurality of the diversion nodes.

[0093] It can be understood that the Voronoi diagram algorithm divides a two-dimensional plane into multiple convex polygonal units. Each unit contains a heat source node as a seed point. The distance from any position in the unit to the seed point is less than the distance to other seed points. In this application, the unit division with non-uniform density is automatically generated according to the heat source level mark of the IT equipment components.

[0094] Step S32, numbering the plurality of shunt nodes based on the heat source level mark, wherein the node number includes a first category number, a second category number, a third category number and a fourth category number, and defining the plurality of shunt nodes based on the node number as trunk nodes, shunt nodes, gross detail points and irrelevant nodes respectively.

[0095] Specifically, the first category number is represented as , where A represents the zero-level heat source mark included in the heat source level mark, X represents the Xth shunt node, and the shunt node with the first category number is defined as an irrelevant node, and the irrelevant node is represented as a node that is irrelevant to the shunt network.

[0096] Specifically, the second category number is represented by , where B represents the first-level heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node with the second category number is defined as a capillary detail point, which is represented by a node in the shunt network for establishing a capillary cooling channel.

[0097] Specifically, the third category number is represented by , where C represents the secondary heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node with the third category number is defined as a shunt node, and the shunt node is represented as a node in the shunt network for establishing a shunt cold channel.

[0098] Specifically, the fourth category number is represented by , where D represents the third-level heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node with the fourth category number is defined as the trunk node, which is represented by the node used to establish the trunk cold channel in the shunt network.

[0099] Step S33 , performing shunt network connections based on the trunk nodes, the shunt nodes, and the capillary points to generate a plurality of two-dimensional shunt network models.

[0100] Furthermore, step S31 includes:

[0101] Step S311 , grouping the node numbers of the trunk nodes, the shunt nodes, and the fine and fine points into a node number data matrix, wherein the node number data matrix includes the node numbers corresponding to the trunk nodes, the shunt nodes, and the fine and fine points.

[0102] Step S312 , randomly sampling and extracting the numbered data matrix based on a Monte Carlo algorithm, and generating a plurality of randomly numbered data matrices, wherein the randomly numbered data matrix represents the connection order of the trunk nodes, the branch nodes, and the hair detail points.

[0103] It can be understood that the Monte Carlo algorithm generates several random connection orders based on the heat source nodes divided by the Voronoi diagram by randomly sampling the probability distribution space, and uses statistical simulation to efficiently explore the diversion network topology combinations that cannot be covered by the traditional exhaustive method. In this step, each randomly generated node connection scheme is automatically pre-screened by fluid simulation to generate multiple randomly numbered data matrices.

[0104] Step S313 , performing shunt network connection based on the multiple randomly numbered data matrices to generate multiple two-dimensional shunt network models.

[0105] It is understandable that in a plurality of randomly numbered data matrices, each row represents a connection of a cold channel, for example, the first row is 、 、 , it means that when the cold channel is connected for the first time, the connection method of connecting the No. 1 trunk node to the No. 2 branch node and then to the No. 3 capillary branch node is adopted.

[0106] Step S34 : generating a plurality of three-dimensional flow distribution network models based on the plurality of two-dimensional flow distribution network models and the cold plate constraint conditions.

[0107] It is understandable that when converting a two-dimensional shunt network model into a three-dimensional shunt network model, it is necessary to add cold plate constraints and then stretch the two-dimensional shunt network model into a three-dimensional shunt network model. Designers can set the cold channel diameters of the main cold channel, shunt cold channel, and capillary cold channel in the shunt network according to the cold plate conditions. For example, the main cold channel can be set to 8mm, the shunt cold channel can be set to 4mm, and the capillary cold channel can be set to 1mm.

[0108] Step S4, simulate starting the IT equipment and injecting coolant into multiple three-dimensional shunt network models, obtain multiple simulation data matrices, and perform a discrimination operation on the multiple simulation data matrices, and execute step S5 or step S6 based on the discrimination operation.

[0109] Specifically, the simulation data matrix includes multiple simulation data vectors, each simulation data vector may contain simulation data such as cold channel flow rate, IT equipment component temperature, IT equipment component temperature change rate, etc., and the multiple simulation data vectors correspond to multiple three-dimensional diversion network models respectively.

[0110] In this embodiment, step S4 includes:

[0111] Step S41 , obtaining a design expected data vector, and sequentially comparing a plurality of the simulation data vectors in the simulation data matrix with the design expected data vector.

[0112] Step S41-1: If the simulation data vector satisfies the design expected data vector, the simulation data vector is derived, and the deviation value between the simulation data vector and the design expected data vector is calculated. The three-dimensional shunt network model is sorted based on the deviation value, and step S6 is executed.

[0113] Specifically, the designer can set the design expected data vector when initializing the method of this embodiment. For example, it can be set to an expected temperature of 65 degrees Celsius, with a fluctuation of 2 degrees Celsius. When calculating the deviation value of the simulation data vector and the design expected data vector, the fluctuation of 2 degrees Celsius is not considered, and only the deviation value between the simulation temperature and the expected temperature is calculated. When outputting, the output is in descending order from large to small, which more intuitively arranges the preferred options in the front row, and the designer can set the descending sorting selection method, such as selecting simulation temperature or simulation flow rate.

[0114] Step S41 - 2 : If the plurality of simulation data vectors do not meet the design expected data vector, then execute step S5 .

[0115] It can be understood that when multiple simulation vectors do not meet the design expected data vectors, it indicates that the multiple groups of solutions do not meet expectations, and step S5 needs to be performed to improve the solutions.

[0116] Step S5: importing the multiple three-dimensional flow diversion network models into the improved flow diversion model respectively, generating multiple improved three-dimensional flow diversion network models based on the improved flow diversion model, and executing step S4.

[0117] Specifically, historical experimental data is obtained, and the historical experimental data represents data of past design and experimental simulation of the shunt network in the liquid-cooled cold plate. An improved model to be trained is constructed based on a convolutional neural network, and the historical experimental data is imported into the improved model to be trained as training data. When the improved model to be trained reaches the maximum number of iterations, the improved shunt model is exported, and multiple three-dimensional shunt network models are imported into the improved shunt model to generate multiple improved three-dimensional shunt network models.

[0118] It is understandable that when the convolutional neural network training reaches the maximum iteration, further training of the model will have little effect and is prone to overfitting problems. When building a convolutional neural network, it is necessary to consider the hardware configuration of the input layer, convolution layer and output layer of the convolutional neural network. For example, the input layer can be set to 200 groups of cold plate experimental data, and the convolution layer can be set to a 3*3 convolution kernel and 65 channels.

[0119] It is understandable that after outputting multiple improved three-dimensional flow distribution network models, it is not possible to confirm whether the improved three-dimensional flow distribution network model meets expectations, so the improved three-dimensional flow distribution network model needs to be re-executed in step S4 to be judged.

[0120] In step S6, the three-dimensional shunt network models are sorted and output based on the discrimination operation. Designers can use multiple three-dimensional shunt network models as references to assist in the design of the actual shunt network, thereby improving the R&D speed.

[0121] A system for constructing a flow distribution network simulation model for a liquid-cooled cold plate, comprising:

[0122] The acquisition module 101 is used to acquire the distribution positions of IT equipment components and cold plate constraints.

[0123] The establishing module 102 is used to establish a two-dimensional IT simulation plane, and to establish a two-dimensional network simulation plane based on the two-dimensional IT simulation plane and the cold plate constraint condition.

[0124] The marking module 103 is used to mark the heat source level of the two-dimensional network simulation plane.

[0125] The flow distribution module 104 is used to divide the two-dimensional network simulation plane into units and generate multiple three-dimensional flow distribution network models.

[0126] The simulation module 105 is used to simulate starting up the IT equipment and injecting coolant into the three-dimensional flow distribution network model, obtain multiple simulation data matrices, and perform discrimination operations.

[0127] The improvement module 106 is used to generate multiple improved three-dimensional flow distribution network models.

[0128] The output module 107 is used to sort and output the three-dimensional flow distribution network model.

[0129] The specific usage and function of this embodiment are as follows:

[0130] First, step S1 can be used to obtain the distribution position of IT equipment components, and a two-dimensional IT simulation plane can be established based on the distribution position of IT equipment components. According to step S1, the component layout of IT equipment can be understood. Then, step S2 can be used to obtain the cold plate constraint condition, thereby restricting the two-dimensional cold plate simulation plane to prevent the two-dimensional cold plate simulation plane from being incompatible with the IT equipment. A two-dimensional network simulation plane is established based on the cold plate constraint condition and the two-dimensional cold plate simulation plane. Then, step S3 is used to divide the two-dimensional network simulation plane into units, obtain the shunt nodes related to the two-dimensional network simulation plane, and perform shunt network connection to generate multiple three-dimensional shunt network models. Then, step S 4 can simulate the three-dimensional shunt network model to obtain a simulation data matrix, and judge the simulation data matrix through a discrimination operation to determine whether it meets the expected design requirements. Finally, through step S6, the three-dimensional shunt network model that meets the expected design requirements can be output, or the three-dimensional shunt network model can be improved through step S5 to obtain an improved three-dimensional shunt network model, and step S4 can be re-executed. Designers can use the output of multiple three-dimensional shunt network models to assist in the design of the liquid-cooled cold plate shunt network, thereby improving the research and development rate. The output of multiple three-dimensional shunt network models is not limited to a single output, and designers can have more reference solutions.

[0131] An electronic device, comprising:

[0132] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method proposed in the embodiment of the present invention.

[0133] The following is a detailed introduction to the various components of electronic equipment:

[0134] Among them, the processor is the control center of the electronic device, which can be a single processor or a collective term for multiple processing elements. For example, the processor is one or more central processing units (CPUs), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement an embodiment of the present invention, such as one or more microprocessors (digital signal processors, DSPs) or one or more field programmable gate arrays (FPGAs).

[0135] The processor can execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0136] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0137] The memory may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only (CD-ROM), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these. The memory may be integrated with the processor or exist independently and be coupled to the processor through an interface circuit of the electronic device. This is not specifically limited in the embodiments of the present invention.

[0138] The above embodiments may be implemented in whole or in part through software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in accordance with the embodiments of the present invention are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a limited means (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes a collection of one or more available media. The available medium may be a magnetic medium (such as a floppy disk, hard disk, or tape), an optical medium (such as a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state drive.

[0139] It should be understood that the term "and / or" in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the previous and next associated objects are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the previous and next context for specific understanding.

[0140] It should be understood that in the embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0141] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A method for constructing a shunt network simulation model for a liquid-cooled cold plate, characterized in that: The following steps are included: S1: Obtaining the distribution positions of IT equipment components, and establishing a two-dimensional IT simulation plane based on the distribution positions of the IT equipment components; S2: Acquire cold plate constraints, where the cold plate constraints represent restrictions on the installation of a liquid-cooled cold plate on IT equipment; establish a two-dimensional network simulation plane based on the cold plate constraints and the two-dimensional IT simulation plane; and mark the two-dimensional network simulation plane with a heat source level; S3: Dividing the two-dimensional network simulation plane into units, obtaining shunt nodes, performing shunt network connections based on the shunt nodes, and generating multiple three-dimensional shunt network models; S4: Simulate starting the IT equipment and injecting coolant into multiple three-dimensional flow distribution network models, obtain multiple simulation data matrices, perform a discrimination operation on the multiple simulation data matrices, and execute step S5 or step S6 based on the discrimination operation; S5: importing the multiple three-dimensional flow diversion network models into the improved flow diversion model respectively, generating multiple improved three-dimensional flow diversion network models based on the improved flow diversion model, and executing step S4; Acquiring historical experimental data, wherein the historical experimental data represents data of past designs and experimental simulations of a flow distribution network within a liquid-cooled cold plate; Constructing an improved model to be trained based on a convolutional neural network, importing the historical experimental data as training data into the improved model to be trained, deriving the improved diversion model when the improved model to be trained reaches a maximum number of iterations, and importing multiple three-dimensional diversion network models into the improved diversion model to generate multiple improved three-dimensional diversion network models; S6: Sort and output the three-dimensional diversion network model based on the discriminant operation.

2. The method for constructing a flow distribution network simulation model of a liquid-cooled cold plate according to claim 1, characterized in that: The step of establishing a two-dimensional network simulation plane based on the cold plate constraint condition and the two-dimensional IT simulation plane, and marking the two-dimensional network simulation plane with a heat source level includes: establishing a two-dimensional cold plate simulation plane based on the cold plate constraint condition and the two-dimensional IT simulation plane, and overlapping the two-dimensional cold plate simulation plane with the two-dimensional IT simulation plane to obtain the two-dimensional network simulation plane; Obtaining the operating power of IT equipment components based on the IT equipment, and meshing the two-dimensional network simulation plane into a plurality of rectangular liquid cooling grids of equal area; Establishing a liquid cooling two-dimensional coordinate system based on the two-dimensional network simulation plane, mapping the plurality of rectangular liquid cooling grids and the distribution positions of the IT equipment components into the liquid cooling two-dimensional coordinate system, using the intersection of the lower right corner edges of the rectangular liquid cooling grids as the X-axis coordinate, and using the intersection of the upper right corner edges of the rectangular liquid cooling grids as the Y-axis coordinate; Based on the distribution position of the IT equipment components, each rectangular liquid cooling grid in the liquid cooling two-dimensional coordinate system is traversed and queried, and the rectangular liquid cooling grid is marked with a heat source level according to the operating power of the IT equipment components. The heat source level mark includes a zero-level heat source mark, a first-level heat source mark, a second-level heat source mark, and a third-level heat source mark.

3. The method for constructing a flow distribution network simulation model of a liquid-cooled cold plate according to claim 2, characterized in that: The traversing and querying each of the rectangular liquid cooling grids in the liquid cooling two-dimensional coordinate system based on the distribution position of the IT equipment components and marking the rectangular liquid cooling grids with a heat source level according to the operating power of the IT equipment components includes: Obtaining a first ratio value of each rectangular liquid cooling grid containing the IT equipment component distribution position, and setting the first ratio value to zero if the rectangular liquid cooling grid does not contain the IT equipment component distribution position; Obtaining a second ratio value of each rectangular liquid cooling grid containing the two-dimensional cold plate simulation plane, and setting the second ratio value to zero if the rectangular liquid cooling grid does not contain the two-dimensional cold plate simulation plane; Obtaining a third proportional value based on the first proportional value and the operating power of the IT equipment component, and setting the third proportional value to zero if the first proportional value is zero; Each of the rectangular liquid cooling grids is marked with a heat source level based on the first ratio value, the second ratio value, and the third ratio value.

4. The method for constructing a flow distribution network simulation model of a liquid-cooled cold plate according to claim 3, characterized in that: The method further comprises: Each of the rectangular liquid cooling grids includes a first proportional value, a second proportional value, and a third proportional value; The zero-level heat source mark indicates an area with no self-heating temperature after the IT equipment is in operation. If any of the first ratio value, the second ratio value, and the third ratio value is zero, the rectangular liquid cooling grid is marked as a zero-level heat source. The first-level heat source mark indicates the area where the self-heating temperature of the IT equipment is low after operation, the second-level heat source mark indicates the area where the self-heating temperature of the IT equipment is normal after operation, and the third-level heat source mark indicates the area where the self-heating temperature of the IT equipment is high after operation; If the first ratio value, the second ratio value, and the third ratio value are all non-zero, the rectangular liquid cooling grid is marked as a first-level heat source, a second-level heat source, or a third-level heat source based on a set ratio value range.

5. The method for constructing a flow distribution network simulation model of a liquid-cooled cold plate according to claim 1, characterized in that: The unit division of the two-dimensional network simulation plane, obtaining the shunt nodes, performing shunt network connections based on the shunt nodes, and generating multiple three-dimensional shunt network models include: Performing unit division on the two-dimensional network simulation plane based on the Voronoi diagram algorithm to divide the two-dimensional network simulation plane into a plurality of the diversion nodes; Numbering the plurality of shunt nodes based on the heat source level mark, the node numbers including a first category number, a second category number, a third category number, and a fourth category number; defining the plurality of shunt nodes based on the node numbers as trunk nodes, shunt nodes, gross nodes, and irrelevant nodes; Performing shunt network connections based on the trunk nodes, the shunt nodes, and the gross and detailed points to generate multiple two-dimensional shunt network models; A plurality of three-dimensional flow shunt network models are generated based on the plurality of two-dimensional flow shunt network models and the cold plate constraint conditions.

6. The method for constructing a flow distribution network simulation model of a liquid-cooled cold plate according to claim 5, characterized in that: The generating of a plurality of two-dimensional shunt network models by connecting the trunk nodes, the shunt nodes and the hairline and detail points comprises: The node numbers of the trunk nodes, the shunt nodes and the hairy and detailed points are grouped into a node number data matrix, wherein the node number data matrix includes the node numbers corresponding to the trunk nodes, the shunt nodes and the hairy and detailed points; Randomly sampling and extracting the numbered data matrix based on a Monte Carlo algorithm, and generating a plurality of randomly numbered data matrices, wherein the randomly numbered data matrix represents the connection order of the trunk node, the diversion node, and the hair detail point; Based on the multiple randomly numbered data matrices, a shunt network connection is performed to generate multiple two-dimensional shunt network models.

7. The method for constructing a flow distribution network simulation model of a liquid-cooled cold plate according to claim 5, characterized in that: The plurality of shunt nodes are numbered based on the heat source level mark, the node numbers including a first category number, a second category number, a third category number, and a fourth category number. The plurality of shunt nodes are defined based on the node numbers as trunk nodes, shunt nodes, gross nodes, and irrelevant nodes, respectively, including: The first category number is represented by , wherein A represents the zero-level heat source mark included in the heat source level mark, X represents the Xth shunt node, and the shunt node with the first category number is defined as an irrelevant node, and the irrelevant node represents a node that is irrelevant to the shunt network; The second category number is represented by , where B represents the first-level heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node of the second category number is defined as a capillary detail point, which is represented by a node in the shunt network for establishing a capillary cooling channel; The third category number is represented by , where C represents the secondary heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node with the third category number is defined as a shunt node, and the shunt node represents a node in the shunt network for establishing a shunt cold channel; The fourth category number is represented by , where D represents the third-level heat source mark included in the heat source grade mark, X represents the Xth shunt node, and the shunt node with the fourth category number is defined as the trunk node, which is represented by the node used to establish the trunk cold channel in the shunt network.

8. The method for constructing a flow distribution network simulation model of a liquid-cooled cold plate according to claim 1, characterized in that: The step of simulating starting the IT equipment and injecting coolant into the multiple three-dimensional shunt network models, obtaining multiple simulation data matrices, performing a discrimination operation on the multiple simulation data matrices, and executing step S5 or step S6 based on the discrimination operation includes: The simulation data matrix includes a plurality of simulation data vectors, and the plurality of simulation data vectors respectively correspond to a plurality of the three-dimensional shunt network models; Acquire a design expected data vector, and sequentially compare a plurality of the simulation data vectors in the simulation data matrix with the design expected data vector; If the simulation data vector satisfies the design expected data vector, the simulation data vector is derived, and a deviation value between the simulation data vector and the design expected data vector is calculated, and the three-dimensional shunt network model is sorted based on the deviation value, and step S6 is executed; If the plurality of simulation data vectors do not meet the design expected data vector, step S5 is executed.

9. A system for constructing a shunt network simulation model for a liquid-cooled cold plate, characterized in that: include: An acquisition module, the acquisition module is used to obtain the distribution position of IT equipment components and cold plate constraints; An establishment module, the establishment module being used to establish a two-dimensional IT simulation plane, and to establish a two-dimensional network simulation plane based on the two-dimensional IT simulation plane and the cold plate constraint condition; a marking module, the marking module being used to mark the heat source level of the two-dimensional network simulation plane; A shunt module, the shunt module is used to divide the two-dimensional network simulation plane into units and generate multiple three-dimensional shunt network models; A simulation module is used to simulate starting up IT equipment and injecting coolant into the three-dimensional shunt network model, obtaining multiple simulation data matrices for performing discrimination operations; An improved module for generating multiple improved three-dimensional diversion network models; Acquiring historical experimental data, wherein the historical experimental data represents data of past designs and experimental simulations of a flow distribution network within a liquid-cooled cold plate; Constructing an improved model to be trained based on a convolutional neural network, importing the historical experimental data as training data into the improved model to be trained, deriving a diversion improved model when the improved model to be trained reaches a maximum number of iterations, and importing multiple three-dimensional diversion network models into the diversion improved model to generate multiple improved three-dimensional diversion network models; The output module is used to sort and output the three-dimensional diversion network model.

Images