WO2024113835A1

WO2024113835A1 - Method and apparatus for generating fan model

Info

Publication number: WO2024113835A1
Application number: PCT/CN2023/104029
Authority: WO
Inventors: 李龙婷
Original assignee: 苏州元脑智能科技有限公司
Priority date: 2022-11-30
Filing date: 2023-06-29
Publication date: 2024-06-06
Also published as: CN115544815A; CN115544815B

Abstract

Provided in some embodiments of the present application are a method and apparatus for generating a fan model. The method comprises: acquiring a plurality of fan models, which are generated on the basis of preset optimization parameters; performing simulation calculation on the plurality of fan models, so as to generate a first PQ curve and a first noise value of each fan model; acquiring an optimization function for performing overall optimization on the first PQ curve, and determining respective first PQ optimization values of a plurality of first PQ curves on the basis of the optimization function, wherein the optimization function is used for determining, by means of performing overall translation on the first PQ curve, a translation mode that optimizes the performance of the first PQ curve; and updating the fan model according to the first PQ optimization value and the first noise value. By means of some embodiments of the present application, optimization and adjustment are performed on an overall PQ curve, and an optimization effect of improving the entire PQ performance can be achieved.

Description

A method and device for generating a fan model

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority to a Chinese patent application filed with the China Patent Office on November 30, 2022, with application number 202211520551.1 and application name “A method and device for generating a fan model”, all contents of which are incorporated by reference in this application.

Technical Field

The present application relates to the technical field of fan models, and in particular to a method and device for generating a fan model.

Background technique

With the high performance of electronic products and the increasing attention paid to energy saving and noise reduction by society, fans, as one of the main components of air cooling and heat dissipation systems, have put forward increasingly stringent requirements on their cooling and noise performance. After years of development, the design system of fan units and their application in electronic products have become very mature, and their performance improvement is gradually approaching the limit.

The current design process of electronic fan blades/fan frames is to take the maximum air volume point (also known as the noise test point), the highest wind pressure point or the highest efficiency point on the fan PQ curve as the optimization target after the fan model is initially generated. That is, a single operating point on the PQ curve is selected as the optimization target for iterative optimization.

However, since the aerodynamic performance of the fan is the result of a dynamic balance between flow and pressure difference, that is, the operating point moves dynamically on the PQ curve, focusing only on the performance of a certain point will often ignore the performance in the remaining flow areas. Therefore, the above-mentioned optimization method for a single operating point is actually difficult to obtain better optimization results.

Summary of the invention

In view of the above problems, a method and device for generating a fan model is proposed to overcome the above problems or at least partially solve the above problems, including:

A method for generating a fan model, the method comprising:

Acquire multiple fan models generated based on preset optimization parameters;

Performing simulation calculations on multiple fan models respectively to generate a first PQ curve and a first noise value for each fan model;

Obtaining an optimization function for overall optimization of the first PQ curve, and determining first PQ optimization values of each of the plurality of first PQ curves based on the optimization function; wherein the optimization function is used to determine a translation mode that optimizes the performance of the first PQ curve by overall translation of the first PQ curve;

The fan model is updated according to the first PQ optimization value and the first noise value.

In some embodiments, a fan model generation device is also included, the device comprising:

A fan model acquisition module, used to acquire multiple fan models generated based on preset optimization parameters;

A simulation calculation module, used to perform simulation calculations on multiple fan models respectively, and generate a first PQ curve and a first noise value for each fan model;

an overall optimization module, used to obtain an optimization function for overall optimization of the first PQ curve, and determine first PQ optimization values of each of the plurality of first PQ curves based on the optimization function; wherein the optimization function is used to determine a translation mode that optimizes the performance of the first PQ curve by overall translation of the first PQ curve;

The fan update module is used to update the fan model according to the first PQ optimization value and the first noise value.

In some embodiments, a server is also included, including a processor, a memory, and a computer program product stored in the memory and capable of A computer program running on a processor implements the above-mentioned method for generating a fan model when the computer program is executed by the processor.

In some embodiments, a non-volatile computer-readable storage medium is also included, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned method for generating a fan model is implemented.

Some embodiments of the present application have the following advantages:

In some embodiments, multiple fan models generated based on preset optimization parameters are obtained; simulation calculations are performed on the multiple fan models respectively to generate a first PQ curve and a first noise value for each fan model; an optimization function for overall optimization of the first PQ curve is obtained, and based on the optimization function, the first PQ optimization values of the multiple first PQ curves are determined respectively; wherein the optimization function is used to determine a translation method that optimizes the performance of the first PQ curve by overall translation of the first PQ curve; according to the first PQ optimization value and the first noise value, the fan model is updated, thereby achieving optimization adjustment for the overall PQ curve, thereby achieving an optimization effect in which the overall PQ performance can be improved.

Furthermore, the prediction model is trained with a small amount of data obtained through simulation calculations, so that a large amount of optimized data can be quickly obtained in the prediction model, so that the prediction model can replace a large amount of numerical simulations for optimization work, and can directly obtain the target PQ value in a shorter time, with fewer iterative steps and higher accuracy, avoiding the current multiple iterative steps of taking a single working point on the PQ curve as the optimization target, thereby effectively improving the optimization iteration time of the fan blades and the fan optimization design accuracy.

BRIEF DESCRIPTION OF THE DRAWINGS

In order to more clearly illustrate the technical solution of the present application, the drawings required for use in the description of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

FIG. 1a is a flowchart of a method for generating a fan model provided in some embodiments of the present application;

FIG. 1b is a PQ curve provided in some embodiments of the present application;

FIG2a is a flowchart of another method for generating a fan model provided in some embodiments of the present application;

FIG2b is a PQ curve sampling diagram provided by the present application in some embodiments;

FIG3 is a flowchart of another method for generating a fan model provided in some embodiments of the present application;

FIG4a is a schematic diagram of a fan modeling process provided in some embodiments of the present application;

FIG4b is an optimization process of a proxy model provided by the present application in some embodiments;

FIG4c is a Pareto solution set provided by the present application in some embodiments;

FIG5 is a schematic diagram of the structure of a device for generating a fan model provided in some embodiments of the present application;

FIG6 is a schematic diagram of the structure of an electronic device provided in some embodiments of the present application;

FIG. 7 is a schematic diagram of the structure of a non-volatile computer-readable storage medium provided in some embodiments of the present application.

Detailed ways

In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

The current design process of electronic fan blades/fan frames is mainly based on the system heat dissipation requirements for PQ (air volume and pressure curve), blade type selection and modification, to ensure that the selected blade type has a certain range of attack angle adaptability, and then select the appropriate radial stacking method to generate the blade type into a three-dimensional fan blade to complete the preliminary design of the fan blade/fan frame. Under the premise of meeting the PQ performance, in order to make the final design scheme have the highest efficiency and the lowest noise, it is also necessary to use a set of DOE (Design of Experiments, test design method) optimization method, based on some important design parameters of fan blades/fan frames, optimize the design and determine the final solution.

In the process of fan blade/frame optimization design, the generally widely used practice is to use the maximum air volume point (also known as the noise test point), the highest wind pressure point or the highest efficiency point on the fan PQ curve as the optimization target, that is, to select a single operating point on the PQ curve as the optimization target for iterative optimization, and then perform simulation calculations of the PQ curve and noise. Finally, it is necessary to combine noise and efficiency among many calculated PQ curves to select the most suitable one or several PQ curves as the final solution for proofing testing.

However, in order to meet the cooling requirements of servers with different configurations, usually after optimization, the entire PQ needs to be moved up while meeting the efficiency and noise requirements. Since the aerodynamic performance of the fan is the result of the dynamic balance between flow and pressure difference, that is, the working point moves dynamically on the PQ curve, only focusing on the performance of a certain point will often ignore the performance in the remaining flow areas. Therefore, the above optimization method is actually difficult to obtain an optimization result of the overall upward movement of PQ. When performing multi-objective optimization, the noise performance of the entire PQ and the maximum air volume should be taken into account at the same time.

Based on this, the present application proposes in some embodiments that in the process of multi-objective optimization of the fan, a new optimization function can be used to perform overall optimization on the PQ curve, and the PQ optimization value is determined according to the optimization function, thereby achieving an update of the fan model.

1a, a flowchart of a method for generating a fan model provided in some embodiments of the present application is shown, which may include the following steps:

Step 101, obtaining multiple fan models generated based on preset optimization parameters;

The optimization parameters include any one or more of the following:

Cross-sectional installation angle, bend angle corresponding to radial stacking line, and sweep angle corresponding to radial stacking line.

During the optimization process of the fan model, a plurality of fan models may be generated based on preset optimization parameters, wherein the optimization parameters of the fan may be any one or more of the installation angle, the bend angle and the sweep angle.

It should be noted that in some embodiments of the present application, other parameters of the fan blades may be selected as optimization parameter variables according to requirements.

In practical applications, the optimization parameters can be determined according to the optimization objectives, different optimization methods can be set for different optimization parameters, and an objective function related to the optimization parameters can be set in the optimization method.

When it is determined that a certain optimization parameter needs to be optimized, the maximum or minimum of the objective function is achieved by changing the optimization parameter, thereby determining the parameter value that meets the optimization goal.

In some embodiments, multiple fan models generated based on preset optimization parameters are obtained, including: obtaining an original fan model to be processed and determining optimization parameters for the original fan model; generating a sampling set according to the optimization parameters, the sampling set including multiple sampling results for the optimization parameters; and updating the original fan model based on the sampling set to generate multiple fan models.

In practical applications, a preliminary original fan model can be determined. The original fan model can be a 3D (Three Dimensional) fan model. After the original fan model is determined, any parameters can be used as optimization variables to carry out optimization work, so that the optimization parameters of the original fan model can be determined according to the optimization target of this optimization. Then, the parameter range of the optimization parameters is set according to the optimization parameters, and the sampling data is ensured, so that the optimization parameters can be sampled within the parameter range of the optimization parameters to generate a sampling set, and the original fan model can be updated for each sampling result in the sampling set to generate multiple fan models.

In some embodiments, a sample point matrix may be determined based on an optimal Latin hypercube sampling method, and a fan module may be generated based on the sample point matrix. Each sample corresponds to a set of fan parameters, so that a corresponding daily fan model may be generated.

In some embodiments, the original fan model is generated by the following steps: in response to a blade drawing operation for the original fan model, generating original blades of the original fan model; in response to a drawing operation for the original blades, generating the original fan model based on the original blades.

In practical applications, fan blades can be drawn and fitted based on professional rotating machinery modeling software, and the blade profiles can be stacked into three-dimensional fan blades based on a reasonable radial stacking method.

After the original fan blades are generated, drawing is performed in the modeling software to install the hub for the fitted fan blades and complete the array and chamfering of the fan blades, thereby generating a complete fan 3D model.

In some embodiments, after generating the original fan blade of the original fan model, it also includes: extracting the blade shape parameters and stacking parameters of the original fan blade; and generating a first macro command file for blade fitting according to the blade shape parameters and stacking parameters.

In some embodiments, after generating an original fan model based on the original fan blade in response to a drawing operation on the original fan blade, the method further includes: generating a second macro command file for fan model fitting, wherein the second macro command file can be directly called externally.

In some embodiments, the original fan model is updated based on the sampling set to generate multiple fan models, including:

Based on the sampling set, the first macro command file is called to update the original fan model to generate multiple target fan blades; based on the multiple target fan blades, the second macro command file is called to generate multiple fan models.

In actual applications, when generating the original fan model, the blade fitting and model fitting processes can be generated into a callable macro command file for subsequent calls. During the optimization process of the original fan model, the first macro command file can be called based on the sample set to perform blade fitting and generate fan blades, and then the second macro command file can be called to generate a fan model based on the fan blades.

Step 102, performing simulation calculations on multiple fan models respectively to generate a first PQ curve and a first noise value for each fan model;

After obtaining multiple fan models, each fan model can be placed in a model environment for simulation calculation to simulate the actual operation of the fan model, thereby generating a first PQ curve and a first noise value for each fan model. The PQ curve is an air volume and pressure curve, as shown in FIG1b , where the horizontal axis represents the air volume Q and the vertical axis represents the air pressure P.

Step 103, obtaining an optimization function for overall optimization of the first PQ curve, and determining first PQ optimization values of each of the plurality of first PQ curves based on the optimization function; wherein the optimization function is used to determine a translation mode that optimizes the performance of the first PQ curve by overall translation of the first PQ curve;

In order to achieve the overall upward shift of the PQ curve, an optimization function can be set. The optimization parameter determines the best translation method by evaluating the performance of the PQ curve after translation. In the optimization function, the PQ optimization value is the performance representation of each PQ curve after translation.

For the first PQ curve of each fan model, a corresponding first PQ optimization value may be calculated by an optimization function, and the first PQ optimization value is used to characterize the best optimization method of the first PQ curve.

Step 104: Update the fan model according to the first PQ optimization value and the first noise value.

After obtaining the first PQ optimization value, the first PQ optimization value and the first noise value of each fan model are associated. The PQ optimization value is used to characterize the optimization of the PQ curve, and the noise value is used to characterize the optimization effect of the noise. During the overall optimization, the optimization effects of the PQ curve and the noise need to be considered simultaneously. Therefore, the fan model can be updated in combination with the first PQ optimization value and the first noise value.

In some embodiments, updating the fan model according to the first PQ optimization value and the first noise value includes: obtaining server configuration information of the server where the fan model is located; according to the server configuration information, the first PQ optimization value and the first noise value value, determine the target PQ optimization value and target noise value.

In actual applications, the fan model can be a fan installed on the server to dissipate heat for the server, so it needs to cooperate with the relevant configuration of the server. Therefore, during the optimization process, the real-time configuration information of the server can be obtained, combined with the configuration requirements of the server, the first PQ optimization value and the first noise value, to determine the target PQ optimization value and the target noise value.

2a, a flowchart of another method for generating a fan model provided in some embodiments of the present application is shown, which may include the following steps:

Step 201, obtaining multiple fan models generated based on preset optimization parameters;

Step 202, performing simulation calculations on multiple fan models respectively to generate a first PQ curve and a first noise value for each fan model;

Step 203, performing translation for each first PQ curve to generate a plurality of candidate PQ curves corresponding to the first PQ curve;

In the process of tuning parameters for each fan model, the first PQ curve is translated upward to obtain a plurality of candidate PQ curves corresponding to the first PQ curve.

Step 204, sampling the first PQ curve and the candidate PQ curve corresponding thereto according to a preset sampling rule, determining first sampling data of a first sampling point in the first PQ curve and second sampling data of a second sampling point in the candidate PQ curve; the first sampling point corresponds to the second sampling point one by one;

After generating the candidate PQ curve, the first PQ curve and each candidate PQ curve may be sampled, as shown in FIG2b , sampling rules such as the sampling quantity and sampling density may be preset, and sampling may be performed according to the sampling rules, a plurality of first sampling points may be sampled on the first PQ curve, and first sampling data at the first sampling points may be determined, a plurality of second sampling points may be sampled on the candidate PQ curve, and second sampling data at the second sampling points may be determined, wherein the first sampling points correspond one to one to the second sampling points.

Step 205, determining weight data of the first sampling point;

The sum of the weights of the first sampling points is 1, and the weight data of the first sampling points is determined according to the simulation calculation accuracy of the first sampling points.

In actual applications, on the PQ curve, as the air volume decreases and the wind pressure increases, there will be stall zones of varying degrees. This is mainly because during the design process of this type of fan, in order to take into account the high wind pressure and large air volume requirements of the system, the fan design point is generally set in the middle area of PQ. Therefore, as the wind pressure increases, the fan blade profile will have a positive angle of attack, causing large-scale separation on the back of the blade, thereby inducing the stall phenomenon.

During the fan PQ performance test, there are obvious eddies in the stall zone, so the corresponding speed is unstable, which will lead to a big difference between the simulation results and the test results in the stall zone. Generally, the simulation results of the two points Pmax and Qmax are very accurate. Therefore, in order to maximize the optimization accuracy, when the entire PQ is taken as the optimization target, the areas with large weights can be placed in the area with higher simulation accuracy.

For example, for the point Pmax and the points in the range Q1 to Qmax, the simulation results are relatively accurate. A higher weight will be given, while for points in the range of 0 to Q1, a lower weight will be given due to poor simulation accuracy. At the same time, the weights of all the first sampling points on the first PQ curve are added to 1.

Step 206: Determine a first PQ optimization value of the first PQ curve based on the weight data, the first sampling data, and the second sampling data.

In some embodiments, step 206 may include the following sub-steps: determining target deviation data for the first sampling data and the second sampling data; determining a candidate PQ optimization value corresponding to each candidate PQ curve based on the target deviation data and the weight data; and determining a first PQ optimization value for the first PQ curve from among multiple candidate optimization values.

In practical applications, for each candidate PQ curve, the second sampling data can be subtracted from the first sampling data to obtain target deviation data, and then the target deviation data of each first sampling point is multiplied by the corresponding weight and added to obtain the candidate PQ optimization value.

After calculating the candidate PQ optimization values corresponding to the multiple candidate PQ curves according to the above process, the first PQ optimization value corresponding to the best optimization method can be determined therefrom.

In some embodiments, a maximum value among a plurality of candidate optimization values may be determined as the first PQ optimization value of the first PQ curve.

For example, the optimization function of the PQ curve can be defined as follows:

As shown in Figure 2b below, n represents n points obtained by interpolation on a PQ curve, i represents the i-th point, pq_i represents the obtained new PQ curve (candidate PQ curve), pq_oi represents the original PQ curve (first PQ curve), PQtarget represents the optimized value, ki represents the weight value of the i-th point on the PQ, and i can be set according to requirements.

The principle of taking different weight values in formula (1) is as follows: for the point Pmax and the points in the interval Q1 to Qmax, higher weights will be assigned because the simulation results are more accurate, while for the points in the interval 0 to Q1, lower weights will be assigned because the simulation accuracy is poor; during the optimization process, the goal is to make pq_i above pq_oi, so the judgment will be automatically set in the program, and for the working condition where pq_i-pq_oi is less than 0, it will be eliminated when the target is selected.

Among them, Q1 represents the approximate dividing point between the stall and non-stall zones. As shown in Figure 1, a hump appears at the PQ point on the left, which means that P no longer increases as Q decreases, but P remains at a constant value and cannot go up, so it can be judged that a stall has occurred at point Q1. For the area where the air volume is between 0 and Q1, the fan has a relatively large positive attack angle and a relatively serious flow separation on the back of the blade. It is usually called the stall zone. The flow in this area is very unstable and the vortex separation is serious. The gap between the CFD simulation results and the test results is relatively large, that is, the simulation accuracy is low.

Step 207: Update the fan model according to the first PQ optimization value and the first noise value.

In some embodiments, the candidate PQ curves are obtained by translating the PQ curve, and then the optimization value of each candidate PQ curve is calculated by sampling and weighting to evaluate the optimization effect of the PQ curve, thereby achieving accurate optimization of the overall PQ curve.

3 , a flowchart of another method for generating a fan model provided by the present application in some embodiments is shown, which may include the following steps:

Step 301, obtaining multiple fan models generated based on preset optimization parameters;

Step 302, performing simulation calculations on multiple fan models respectively to generate a first PQ curve and a first noise value for each fan model;

Step 303, obtaining an optimization function for overall optimization of the first PQ curve, and determining first PQ optimization values of each of the plurality of first PQ curves based on the optimization function; wherein the optimization function is used to determine a translation mode that optimizes the performance of the first PQ curve by overall translation of the first PQ curve;

Step 304, obtaining a preset prediction model; the prediction model is used to output a predicted PQ optimization value and a noise value for any input optimization parameter;

In practical applications, multi-objective optimization can be achieved through repeated iterations, which takes a lot of time. In order to shorten the optimization time and improve the optimization efficiency, a prediction model can be set up. The preset model can output the corresponding PQ optimization value and noise value under different optimization methods for the input optimization parameters, thereby replacing a large number of numerical simulations for optimization work.

Step 305: training the prediction model based on the first PQ optimization value and the first noise value to update the model parameters of the prediction model;

The first PQ optimization value and the first noise value are related data obtained based on simulation calculations. A data group consisting of preset optimization parameters, the first PQ curve, the first PQ optimization value and the first noise value is used as training data to train the prediction model.

During the training process, the PQ optimization value and noise value predicted by the prediction model can be obtained by taking the optimization parameters as input values, and then the predicted value is compared with the actual first PQ optimization value and the first noise value, and the model parameters of the prediction model are adjusted until the prediction error is within the prediction range, and the model training is terminated.

Step 306, obtaining target optimization parameters for optimizing the fan model;

After the model training is completed, the target optimization parameters can be obtained to use the target optimization parameters to perform large-scale sampling predictions in the fan model.

Step 307, input the target optimization parameter into the updated prediction model, and output the second PQ optimization value and the second noise value corresponding to the target optimization parameter;

After the target optimization parameters are input into the preset model, a second PQ optimization value and a second noise value can be generated accordingly. The second PQ optimization value and the second noise value represent the optimization effects of the PQ curve and the noise respectively.

Step 308, determining a target PQ optimization value and a target noise value based on the second PQ optimization value and the second noise value;

After obtaining the second PQ optimization value and the second noise value, the second PQ optimization value and the second noise value can be in one-to-one correspondence, and then the two are fitted with a curve to determine a target PQ optimization value and a target noise value that meet the optimization target from the fitting curve.

Step 309: Update the fan model according to the target PQ optimization value and the target noise value.

In some embodiments, the method further includes: generating an optimization curve with the second PQ optimization value as the horizontal axis coordinate and the second noise value as the vertical axis coordinate; sampling the optimization curve to determine third sampling data; and verifying the prediction model according to the third sampling data;

In practical applications, in order to ensure the accuracy of the prediction simulation, the prediction model can also be verified. In some embodiments, the second PQ optimization value can be used as the horizontal axis coordinate, and the second noise value can be used as the vertical axis coordinate to fit and generate an optimization curve. Each value on the optimization curve corresponds to a set of optimization results, so that a number of data are sampled for the optimization curve to obtain third sampling data. For the third sampling data, it can be determined whether there is a large difference between it and the data obtained in the actual simulation calculation process. When there is a large difference, it is determined that the prediction model is unreliable. When the difference is within the preset error threshold range, the prediction model is determined to be a reliable model, thereby realizing verification of the prediction model.

In some embodiments, the third sampling data includes a predicted PQ optimization value and a corresponding predicted noise value, and verifying the prediction model according to the third sampling data includes: determining a predicted PQ curve and a predicted optimization parameter of the third sampling data; The predicted fan model is generated based on the predicted optimization parameters; and the predicted model is verified based on the predicted fan model.

In actual application, the third adopted data may correspond to the prediction optimization parameters of the input value prediction model and the prediction PQ curve generated in the prediction process. Then, the fan modeling and simulation calculation may be performed based on the prediction optimization parameters. The fan modeling and simulation calculation refer to steps 301 to 304. Then, the data output by the prediction model and the data output by the simulation calculation may be compared to verify the prediction model and determine whether the prediction model is reliable.

In some embodiments, the prediction model is verified according to the prediction fan model, including: performing simulation calculations on the prediction fan model to generate a simulated PQ curve and a simulated noise value; determining whether the simulated PQ curve and the simulated noise value match the predicted PQ curve and the predicted noise value; determining that the verification is successful when it is determined that the simulated PQ curve and the simulated noise value match the predicted PQ curve and the predicted noise value; determining that the verification has failed when it is determined that the simulated PQ curve and the simulated noise value do not match the predicted PQ curve and the predicted noise value.

In practical applications, after generating the prediction fan model, simulation calculations can be performed to generate a simulation PQ curve and a simulation noise value, and the matching of the PQ curve and the noise value can be judged to determine whether the verification of the prediction model is successful.

In some embodiments, when the verification fails, the prediction model is retrained based on the optimization curve. When the verification succeeds, the target PQ optimization value and the target noise value are determined based on the second PQ optimization value and the second noise value.

In practical applications, if the verification fails, it means that the current prediction model is unreliable, and the model training can continue. If the verification succeeds, it means that the current prediction model is reliable, and a second PQ optimization value and a second noise value can be generated based on the prediction model to determine the target PQ optimization value and the target noise value.

In some embodiments, a prediction model is trained by performing simulation calculations on a small amount of data and obtaining the first PQ data and the first noise value. After the training is completed, a large amount of data can be optimized based on the optimization parameters, thereby achieving direct acquisition of the target PQ value in a shorter time, with fewer iterative steps, and with higher accuracy, avoiding the current multiple iterative steps of using a single working point on the PQ curve as the optimization target, thereby effectively improving the optimization iteration time of the fan blades and the fan optimization design accuracy.

Some embodiments of the present application are exemplarily described below in conjunction with FIG. 4a to FIG. 4c:

In practical applications, the fan modeling process can include the following steps:

S11, based on professional rotating machinery modeling software, performs blade fitting, and based on a reasonable radial stacking method, stacks the blade profile into a three-dimensional fan blade, extracts the corresponding blade profile and stacking parameters, and the blade fitting macro command script file, or directly extracts the blade profile parameters based on the preliminary design plan.

S12, the fitted fan blades are directly entered into the general modeling software for drawing, so as to install the hub for the fitted fan blades, complete the array and chamfering of the fan blades, and finally generate a complete fan 3D model. For this part of the operation, a macro command script that can be directly called externally can be established.

S12, determine the final selected optimization parameter variables according to the optimization target, and determine the sample point matrix based on the optimal Latin hypercube sampling method, where the optimization parameters can be the installation angle of each section, the bending angle and the sweep angle corresponding to the radial stacking line. In this process, the matrix generation script can be edited, as shown in Figure 4a, which is the automatic modeling process of the fan geometry.

S13, write code based on a general programming language. The main purpose and function of this part is to integrate the macro commands of each part of the software together, and at the same time integrate the sample point matrix mentioned in the above steps, call the macro command mentioned in step (1) based on the sample point matrix, fit the new fan blade, and then execute the above step (2) to generate the geometric model corresponding to the sample point using the optimal super Latin square method. At the end of this step, the automatic modeling process ends.

As shown in FIG4b , the optimization process of the proxy model may include the following steps:

S21, define the optimization variables with range, use Latin square stereo sampling random combination, and simulate the sample points calculate.

For all geometric models generated from step S11 to step S13 (one geometric model is generated for each sampling point), mesh division is performed and simulation calculation is carried out.

Based on the simulation, we can obtain the PQ curves calculated from all sample points and the noise values under Qmax. For example, 20 sample points were designed for Latin square stereo sampling before. At this time, 20 PQ intervals and 20 groups of noise values will be calculated. According to the optimization function, the PQ optimization values of these 20 PQs can be obtained.

S22, based on the sample points obtained by the optimal Latin hypercube method, the proxy model training is performed, the aerodynamic noise under the fan Qmax and the equivalent value of the whole PQ performance target PQtarget are used as the response output, the fan performance is batch simulated and analyzed, the sample analysis results are fitted, and the radial basis neural network/model is constructed.

The selected proxy model may be a radial basis neural network. In fact, different proxy models can be compared. The proxy model with the highest fitting accuracy can be selected as the final prediction model. The proxy models include support vector machines, artificial neural networks, etc.

S23, a proxy model is obtained based on training to replace a large amount of numerical simulation for optimization work.

The optimization algorithm may be a multi-island legacy algorithm, etc. Based on multi-objective optimization, a Pareto function curve will be generated, which represents the optimal solution set for trade-offs between noise and PQ.

According to the requirements for noise and PQ, several points are selected on the curve for CFD and CAA numerical simulation verification. If the results show that the prediction results and the simulation results are highly consistent, it means that the proxy model prediction obtained by training is accurate and can be output. If the prediction is less consistent with the simulation results, the Pareto solution set is put into the training sample, and the proxy model training and optimization steps are repeated until an accurate proxy model is obtained, and then the optimization result is output.

Among them, the Pareto solution set is shown in Figure 4c. It can be seen that the vertical axis represents the noise value and the horizontal axis represents the change in the PQ equivalent value, that is, the larger the change, the smaller the improvement in PQ performance. It can be clearly seen from the figure that from left to right, the noise gradually decreases and the PQ optimization degree gradually decreases, that is, the optimization effects of PQ and noise are contradictory. When selecting the optimization target solution, you can select it according to your needs. If you want an optimization model with better PQ performance, select sample points in the left area. If you want to select an optimization model with lower noise, select sample points in the right area.

It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the order of the actions described, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

5 , a schematic diagram of the structure of a fan model generation device provided in some embodiments of the present application is shown, which may include the following modules:

A fan model acquisition module 501 is used to acquire multiple fan models generated based on preset optimization parameters;

A simulation calculation module 502 is used to perform simulation calculations on multiple fan models respectively to generate a first PQ curve and a first noise value for each fan model;

The overall optimization module 503 is used to obtain an optimization function for overall optimization of the first PQ curve, and determine the first PQ optimization values of the plurality of first PQ curves respectively based on the optimization function; wherein the optimization function is used to determine a translation mode that optimizes the performance of the first PQ curve by overall translation of the first PQ curve;

The fan updating module 504 is configured to update the fan model according to the first PQ optimization value and the first noise value.

In some embodiments, the overall optimization module 503 includes:

A candidate PQ curve generating submodule, used for translating each first PQ curve to generate a plurality of candidate PQ curves corresponding to the first PQ curve;

A curve sampling submodule, for sampling the first PQ curve and the candidate PQ curve corresponding thereto according to a preset sampling rule, and determining first sampling data of a first sampling point in the first PQ curve and second sampling data of a second sampling point of the candidate PQ curve; the first sampling point corresponds to the second sampling point one by one;

A weight data determination submodule, used to determine the weight data of the first sampling point;

The first PQ optimization value determination submodule is used to determine a first PQ optimization value of the first PQ curve based on the weight data, the first sampling data and the second sampling data.

In some embodiments, the first PQ optimization value determination submodule may include:

A target deviation data determining unit, used to determine target deviation data of the first sampling data and the second sampling data;

A candidate PQ optimization value determination unit, used to determine a candidate PQ optimization value corresponding to each candidate PQ curve according to the target deviation data and the weight data;

The first PQ optimization value determining unit is used to determine a first PQ optimization value of a first PQ curve from among a plurality of candidate optimization values.

In some embodiments, when determining the first PQ optimization value of the first PQ curve from a plurality of candidate optimization values, the first PQ optimization value determination unit is configured to: determine a maximum value from the plurality of candidate optimization values as the first PQ optimization value of the first PQ curve.

In some embodiments, the fan update module 504 includes:

The prediction model acquisition submodule is used to obtain a prediction model based on a preset; the prediction model is used to output a predicted PQ optimization value and a noise value for any input optimization parameter;

A prediction model training submodule, used for training the prediction model based on the first PQ optimization value and the first noise value to update the model parameters of the prediction model;

A target optimization parameter determination submodule is used to obtain target optimization parameters for optimizing the fan model;

A data prediction submodule, used for inputting the target optimization parameter into the updated prediction model, and outputting a second PQ optimization value and a second noise value corresponding to the target optimization parameter;

a target PQ optimization value determination submodule, configured to determine a target PQ optimization value and a target noise value based on a second PQ optimization value and a second noise value;

The fan model update submodule is used to update the fan model according to the target PQ optimization value and the target noise value.

In some embodiments,

An optimization curve generating submodule, used to generate an optimization curve with the second PQ optimization value as the horizontal axis coordinate and the second noise value as the vertical axis coordinate;

A third sampling data generating submodule is used to sample the optimization curve and determine the third sampling data;

A model verification submodule, used for verifying the prediction model according to the third sampling data;

The model retraining module is used to retrain the prediction model based on the optimization curve when verification fails.

In some embodiments, the fan update module 504 further includes:

The execution submodule is used to determine the target PQ optimization value and the target noise value based on the second PQ optimization value and the second noise value when the verification succeeds.

In some embodiments, the third sampled data includes a predicted PQ optimization value and a corresponding predicted noise value, and the model verification submodule includes:

A prediction optimization parameter determination unit, used to determine a prediction PQ curve and prediction optimization parameters of the third sampling data;

A predictive fan model generating unit, used to generate a predictive fan model based on the predictive optimization parameters;

The verification unit is used to verify the prediction model according to the prediction fan model.

In some embodiments, the verification unit includes:

A simulation calculation subunit is used to perform simulation calculation on the predicted fan model to generate a simulated PQ curve and a simulated noise value;

A matching judgment subunit is used to judge whether the simulated PQ curve and the simulated noise value match the predicted PQ curve and the predicted noise value;

A matching success subunit is used to determine that the verification is successful when it is determined that the simulated PQ curve and the simulated noise value match the predicted PQ curve and the predicted noise value;

The matching failure subunit is used to determine that the verification fails when it is determined that the simulated PQ curve and the simulated noise value do not match the predicted PQ curve and the predicted noise value.

In some embodiments, the fan update module 504 includes:

A server configuration information acquisition submodule, used to acquire server configuration information of the server where the fan model is located;

The optimization value screening submodule is used to determine a target PQ optimization value and a target noise value according to the server configuration information, the first PQ optimization value and the first noise value.

In some embodiments, the fan model acquisition module 501 may include:

The original fan model acquisition submodule is used to obtain the original fan model to be processed.

An optimization parameter determination submodule, used to determine the optimization parameters for the original fan model;

A sampling set generation submodule, used to generate a sampling set according to the optimization parameters, wherein the sampling set includes a plurality of sampling results for the optimization parameters;

The fan model generation submodule is used to update the original fan model based on the sampling set to generate multiple fan models.

In some embodiments, the original fan model acquisition submodule may include:

An original fan blade generating unit, configured to generate original fan blades of the original fan model in response to a fan blade drawing operation on the original fan model;

The original fan model generating unit is used to generate an original fan model based on the original fan blade in response to a drawing operation on the original fan blade.

In some embodiments, the original fan model acquisition submodule may include:

A blade parameter extraction unit, used to extract blade profile parameters and stacking parameters of the original blade;

The first macro command file generating unit is used to generate a first macro command file for blade fitting according to blade profile parameters and stacking parameters.

In some embodiments, the original fan model acquisition submodule may further include:

The second macro command file generating unit is used to generate a second macro command file for fan model fitting.

In some embodiments, the fan model generation submodule may include:

A target fan blade generating unit, configured to call the first macro command file based on the sampling set to update the original fan model and generate a plurality of target fan blades;

The fan model generating unit is used to generate multiple fan models by calling the second macro command file based on multiple target fan blades.

In some embodiments, the optimization parameters include any one or more of the following:

In some embodiments, the sum of the weights of the first sampling points is 1, and the weight data of the first sampling points is determined according to the simulation calculation accuracy of the first sampling points.

6 , a server provided in some embodiments of the present application is shown, which may include a processor 61, a memory 62, and a computer program stored in the memory and capable of running on the processor, and the computer program implements the above fan model generation method when executed by the processor.

7 , a non-volatile computer-readable storage medium provided in some embodiments of the present application is shown. The non-volatile computer-readable storage medium 70 stores a computer program 710 , and when the computer program 710 is executed by a processor, the above method for generating a fan model is implemented.

As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices, or computer program products. Therefore, the embodiments of the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the embodiments of the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

The present application embodiment is described with reference to the flowchart and/or block diagram of the method, terminal device (system) and computer program product according to the embodiment of the present application. It should be understood that each process and/or box in the flowchart and/or block diagram, and the combination of the process and/or box in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device produce a device for realizing the function specified in one process or multiple processes in the flowchart and/or one box or multiple boxes in the block diagram.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce computer-implemented processing, so that the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device that includes a series of elements includes not only those Elements include not only other elements that are not explicitly listed, but also elements that are inherent to such process, method, article or terminal device. In the absence of more restrictions, elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device that includes the elements.

The above is a detailed introduction to the method and device for generating a fan model. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technical personnel in this field, according to the idea of the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

A method for generating a fan model, characterized in that the method comprises:

Acquire multiple fan models generated based on preset optimization parameters;

Performing simulation calculations on the multiple fan models respectively to generate a first PQ curve and a first noise value for each fan model;

Acquire an optimization function for overall optimization of the first PQ curve, and determine first PQ optimization values of each of the plurality of first PQ curves based on the optimization function; wherein the optimization function is used to determine a translation mode that optimizes the performance of the first PQ curve by overall translation of the first PQ curve;

The fan model is updated according to the first PQ optimization value and the first noise value.
The method according to claim 1, characterized in that the obtaining an optimization function for overall optimization of the first PQ curve, and determining the first PQ optimization values of each of the plurality of first PQ curves based on the optimization function, respectively, comprises:

Perform translation on each first PQ curve to generate multiple candidate PQ curves corresponding to the first PQ curve;

According to a preset sampling rule, a first PQ curve and a candidate PQ curve corresponding thereto are sampled to determine first sampling data of a first sampling point in the first PQ curve and second sampling data of a second sampling point of the candidate PQ curve; the first sampling point corresponds to the second sampling point one by one;

Determining weight data of the first sampling point;

A first PQ optimization value of the first PQ curve is determined based on the weight data, the first sampling data, and the second sampling data.
The method according to claim 2, characterized in that the determining the first PQ optimization value of the first PQ curve based on the weight data, the first sampling data and the second sampling data comprises:

Determine target deviation data between the first sampled data and the second sampled data;

Determine a candidate PQ optimization value corresponding to each candidate PQ curve according to the target deviation data and the weight data;

A first PQ optimal value of the first PQ curve is determined from a plurality of candidate optimal values.
The method according to claim 3, characterized in that determining the first PQ optimization value of the first PQ curve from a plurality of candidate optimization values comprises:

A maximum value among the plurality of candidate optimization values is determined as a first PQ optimization value of the first PQ curve.
The method according to claim 1, characterized in that updating the fan model according to the first PQ optimization value and the first noise value comprises:

Obtain a prediction model based on a preset; the prediction model is used to output a prediction for any input optimization parameter PQ optimization value and noise value;

Training the prediction model based on the first PQ optimization value and the first noise value to update model parameters of the prediction model;

Obtaining target optimization parameters for optimizing the fan model;

Inputting the target optimization parameter into the updated prediction model, and outputting a second PQ optimization value and a second noise value corresponding to the target optimization parameter;

Determining a target PQ optimization value and a target noise value based on the second PQ optimization value and the second noise value;

The fan model is updated according to the target PQ optimization value and the target noise value.
The method according to claim 5, further comprising:

Generate an optimization curve with the second PQ optimization value as the horizontal axis coordinate and the second noise value as the vertical axis coordinate;

Sampling the optimization curve to determine third sampling data;

Verifying the prediction model according to the third sampling data;

When the verification fails, the prediction model is retrained based on the optimization curve.
The method according to claim 6, further comprising:

When the verification succeeds, determining a target PQ optimization value and a target noise value based on the second PQ optimization value and the second noise value is performed.
The method according to claim 6, characterized in that the third sampling data includes a predicted PQ optimization value and a corresponding predicted noise value, and the verifying the prediction model according to the third sampling data comprises:

Determining a predicted PQ curve and predicted optimization parameters of the third sampling data;

generating a predictive fan model based on the predictive optimization parameters;

The predictive model is validated based on the predictive fan model.
The method according to claim 8, characterized in that the verifying the prediction model according to the prediction fan model comprises:

Performing simulation calculation on the predicted fan model to generate a simulated PQ curve and a simulated noise value;

Determining whether the simulated PQ curve and the simulated noise value match the predicted PQ curve and the predicted noise value;

When it is determined that the simulated PQ curve and the simulated noise value match the predicted PQ curve and the predicted noise value, determining that the verification is successful;

When it is determined that the simulated PQ curve and the simulated noise value do not match the predicted PQ curve and the predicted noise value, it is determined that the verification fails.
The method according to claim 1, characterized in that the first PQ optimization value and the first A noise value is used to update the fan model, including:

Obtain server configuration information of the server where the fan model is located;

A target PQ optimization value and a target noise value are determined according to the server configuration information, the first PQ optimization value and the first noise value.
The method according to claim 1, characterized in that the obtaining of multiple fan models generated based on preset optimization parameters comprises:

Get the original fan model to be processed;

Determining optimization parameters for the original fan model;

Generating a sampling set according to the optimization parameters, wherein the sampling set includes a plurality of sampling results for the optimization parameters;

The original fan model is updated based on the sampling set to generate multiple fan models.
The method according to claim 11, characterized in that the original fan model is generated by the following steps:

In response to a blade drawing operation for an original fan model, generating original blades of the original fan model;

In response to a rendering operation on the original fan blade, an original fan model is generated based on the original fan blade.
The method according to claim 12, characterized in that after generating the original fan blade of the original fan model, it also includes:

Extracting blade profile parameters and stacking parameters of the original fan blade;

A first macro command file for blade fitting is generated according to the blade profile parameters and the stacked parameters.
The method according to claim 13, characterized in that, after generating an original fan model based on the original fan blade in response to a drawing operation on the original fan blade, the method further comprises:

Generate the second macro command file for fan model fitting.
The method according to claim 14, characterized in that the updating of the original fan model based on the sampling set to generate multiple fan models comprises:

Calling the first macro command file based on the sampling set to update the original fan model and generate a plurality of target fan blades;

A second macro command file is called based on the multiple target fan blades to generate multiple fan models.
The method according to claim 1, characterized in that the optimization parameters include any one or more of the following:

Cross-sectional installation angle, bend angle corresponding to radial stacking line, and sweep angle corresponding to radial stacking line.
The method according to claim 2 is characterized in that the sum of the weights of the first sampling points is 1, and the weight data of the first sampling points is determined according to the simulation calculation accuracy of the first sampling points.
A device for generating a fan model, characterized in that the device comprises:

A fan model acquisition module, used to acquire multiple fan models generated based on preset optimization parameters;

A simulation calculation module, used to perform simulation calculations on the multiple fan models respectively to generate a first PQ curve and a first noise value for each fan model;

an overall optimization module, configured to obtain an optimization function for overall optimization of the first PQ curve, and to determine first PQ optimization values of each of the plurality of first PQ curves based on the optimization function; wherein the optimization function is configured to determine a translation mode that optimizes the performance of the first PQ curve by overall translation of the first PQ curve;

A fan updating module is used to update the fan model according to the first PQ optimization value and the first noise value.
A server, characterized in that it includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein when the computer program is executed by the processor, the method for generating a fan model as described in any one of claims 1 to 17 is implemented.
A non-volatile computer-readable storage medium, characterized in that a computer program is stored on the non-volatile computer-readable storage medium, and when the computer program is executed by a processor, the method for generating a fan model as described in any one of claims 1 to 17 is implemented.