CN117972924A - Method for constructing high-precision motor performance database based on dual-model strategy - Google Patents
Method for constructing high-precision motor performance database based on dual-model strategy Download PDFInfo
- Publication number
- CN117972924A CN117972924A CN202410064673.7A CN202410064673A CN117972924A CN 117972924 A CN117972924 A CN 117972924A CN 202410064673 A CN202410064673 A CN 202410064673A CN 117972924 A CN117972924 A CN 117972924A
- Authority
- CN
- China
- Prior art keywords
- model
- error
- motor
- data set
- proxy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000013461 design Methods 0.000 claims abstract description 66
- 230000004044 response Effects 0.000 claims abstract description 25
- 238000012706 support-vector machine Methods 0.000 claims abstract description 7
- 238000004458 analytical method Methods 0.000 claims description 25
- 238000004088 simulation Methods 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 19
- 238000004590 computer program Methods 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 14
- 238000005457 optimization Methods 0.000 claims description 11
- 238000005070 sampling Methods 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 7
- 238000003860 storage Methods 0.000 claims description 6
- 230000009977 dual effect Effects 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 5
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 9
- 239000003795 chemical substances by application Substances 0.000 description 8
- 230000006870 function Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- SLXKOJJOQWFEFD-UHFFFAOYSA-N 6-aminohexanoic acid Chemical compound NCCCCCC(O)=O SLXKOJJOQWFEFD-UHFFFAOYSA-N 0.000 description 2
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 description 2
- 238000013459 approach Methods 0.000 description 2
- 229910052802 copper Inorganic materials 0.000 description 2
- 239000010949 copper Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000013401 experimental design Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 1
- 238000013101 initial test Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010206 sensitivity analysis Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 101150083151 wtip gene Proteins 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2457—Query processing with adaptation to user needs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Genetics & Genomics (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Physiology (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
The invention discloses a method for constructing a high-precision motor performance database based on a dual-model strategy. And then, predicting another group of data sets through the proxy model, comparing the predicted value with the true value, calculating the relative error of the data sets predicted by the proxy model, taking the relative error calculated by predicting each data point by the proxy model as an output response value, taking Euclidean distance between each component of the data points and the data point constructing the proxy model as an input variable, and constructing a support vector machine regression model as an error model for judging whether the response of the proxy model to a certain design is high error response. The method provided by the invention can reduce the time required for constructing the high-precision motor performance database, and has good practicability and economy.
Description
Technical Field
The invention relates to a method for constructing a high-precision motor performance database based on a dual-model strategy, and belongs to the technical field of motor body simulation design.
Background
Most engineering design problems require evaluation of target results using different design parameters through simulation analysis experiments. However, for many practical problems, a single simulation analysis with a high confidence level may take minutes, hours, or even days to complete. And the optimization design, the sensitivity analysis, the hypothesis analysis and the like need to carry out a large amount of simulation analysis and calculation, and a large amount of time is consumed. To improve this, a proxy model may be used to model between design parameters and target results. The agent model is commonly used for optimizing the design in the aspect of motor design, and the calculation result of the agent model is very similar to that of the simulation model, but the calculation amount of solving is small.
Mu Xuefeng et al, in the "research of commonly used proxy models in multidisciplinary design optimization" in one document, mention that proxy model-based optimization often obtains an initial sampling point through experimental design to train an initial proxy model, then searches for a demand design through an optimization algorithm, solves the demand design and adds the result to the sampling point, and repeats this operation until the design is good enough. But this approach may only converge to a locally optimal design and cannot find a globally optimal design.
Gu Jialei in "high-speed permanent magnet motor performance analysis and optimization design for electric turbocharging" discloses a motor multi-objective robustness optimization method based on a local proxy model, wherein partial deterministic optimization data are selected in the aspect of experimental design to construct a proxy model, the optimal design of robustness optimization is obtained through a proxy model on a motor performance prediction result, then simulation analysis and calculation are carried out on the optimal design, the result is added into a proxy model training data set, and the operation is repeated until the termination condition set by an experiment is met. Although the proxy model trained by this method is highly accurate, many redundant sampling designs are still produced.
Sacks, DESIGN AND ANALYSIS of computer experiments, proposes to interpret the mean square error of the Kriging model by statistical methods to calculate the model predictions. However, in practical application, the unique error estimation mode of the Kriging proxy model cannot be popularized to the application of other proxy models, and the estimated value of the Kriging model on the error sometimes has a far difference from the actual error in sensitive design, so that the method is difficult to apply to the actual error prediction.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: the method for constructing the high-precision motor performance database based on the dual-model strategy is provided, the accuracy of the proxy model is improved under the condition that the high-error design can be judged, and the problem that a large amount of time is consumed when the proxy model reaches the required accuracy is solved.
The invention adopts the following technical scheme for solving the technical problems:
The method for constructing the high-precision motor performance database based on the dual-model strategy comprises the following steps:
Acquiring an initial motor performance data set, randomly dividing the initial motor performance data set into an S data set and an X data set, and establishing a proxy model for fitting a mapping relation between motor parameters and motor performance through motor input variables and output response values in the S data set; predicting motor performance corresponding to motor parameters in the X data set through the proxy model, comparing a predicted value with a true value, calculating relative errors of the X data set predicted by the proxy model, and taking the relative errors calculated by each data point in the X data set predicted by the proxy model as an output response value of the proxy model;
constructing an input variable of an error model, and training a support vector machine regression model as an error model for judging whether the response of the motor design to be judged is high error response or not by using the support vector machine regression model as a judgment proxy model;
Screening out a proxy model high-error motor design by using an error model, calculating the high-error motor design by using finite element simulation analysis, adding a calculation result into an S data set, updating a proxy model, predicting motor performance corresponding to motor parameters in an X data set according to the updated proxy model, calculating relative errors of the motor parameters in the X data set predicted by the updated proxy model as output response of the updated error model, updating input variables of the error model, and retraining the error model; and continuously updating the proxy model and the error model until the precision of the proxy model reaches the required precision or meets the set updating times of the proxy model.
As a preferable scheme of the invention, the initial motor performance data set is obtained by carrying out finite element model simulation analysis and calculation on the motor through Latin hypercube sampling.
As a preferable scheme of the invention, the motor design to be judged is obtained through random search, latin hypercube sampling or multi-objective optimization algorithm NSGA-II each generation of decision solution.
As a preferable scheme of the invention, the establishment and update process of the proxy model is as follows:
And performing super-parameter tuning on the proxy model by adopting a grid search method, guiding the proxy model by using an error model to obtain a high-error motor design, calculating the motor performance of the high-error motor design by finite element simulation analysis, adding a calculation result into a data set for training the proxy model, and continuously updating the proxy model until the proxy model meets the precision requirement.
As a preferable scheme of the invention, the error model judges the design standard of the high-error motor as follows:
if the relative error value of a motor design is larger than the average value of the relative errors of the X data sets predicted by the proxy model and the standard deviation of the relative errors of the X data sets predicted by the p multiplied by the proxy model, the motor design is judged to be a high-error motor design, and p is a pre-designed value.
As a preferred scheme of the invention, the method for constructing the input variable of the error model comprises the following specific steps:
Normalizing the S data set and the X data set, for the jth component X j of the data points X in the X data set, selecting a data point S * corresponding to the jth component with the smallest difference value with X j in the jth components of all the data points in the S data set, calculating Euclidean distance between the two data points as the jth component of a new data point X ', and calculating the distance between the data point X and the S data set as the last component of the new data point X'; carrying out the same treatment on each data point in the X data set, and carrying out the second normalization on the new data point set obtained after the treatment is completed to be used as an input variable of an error model;
the j-th component of the new data point x' is specifically as follows:
Where d (X, S *) represents the Euclidean distance between data points X and S *, n is the dimension of each data point of the X dataset, d (X, S) represents the minimum distance between data point X and the S dataset, i.e., d (X, S) =inf { d (X, S) |s ε S }, S 1,...,sn represents the component of data point S in the S dataset, X 'j represents the jth component of the new data point X';
The normalization formula for all datasets is:
Where y j is the j-th component of the pre-normalized data point and y 'j is the j-th component of the post-normalized data point' For the minimum value of the j-th component of all data points in the dataset,/>The maximum value of the j-th component for all data points in the dataset.
As a preferable scheme of the invention, the updating method of the error model is as follows:
Predicting motor performance corresponding to motor parameters in the X data set by using the updated proxy model, comparing the predicted value with a true value, calculating the relative error of the updated proxy model for predicting the X data set, and taking the relative error calculated by each data point in the updated proxy model for predicting the X data set as an output response value of the updated proxy model; the input variable construction mode of the error model is unchanged, and the error model is retrained to obtain an updated error model.
A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, the processor implementing the steps of the method of constructing a high precision motor performance database based on a dual model strategy when the computer program is executed.
A computer readable storage medium storing a computer program which when executed by a processor performs the steps of the method of constructing a high precision motor performance database based on a dual model strategy.
Compared with the prior art, the technical scheme provided by the invention has the following technical effects:
1. According to the method, the motor parameter design with higher prediction error of the proxy model can be effectively distinguished, the process is very simple and easy to understand, after the error model distinguishes the motor design with higher prediction error of the proxy model, the finite element model can be continuously called on the high error design of the proxy model to carry out simulation analysis calculation, the error of the motor design distinguished by the proxy model is reduced, and the unnecessary and extremely time-consuming finite element model analysis calculation is reduced when a higher-precision database is built.
2. The invention distinguishes the high-error motor design of the proxy model through the error model, continuously calls the finite element model to carry out simulation analysis calculation, adds the result into the data set of the training proxy model, updates the proxy model until the precision or the update times set according to the application requirement are met, and finally obtains the proxy model training data set from the high-error motor design of the proxy model distinguished from the initial test design and the error model.
Drawings
FIG. 1 is a flow chart of a method of constructing a high-precision motor performance database based on a dual model strategy of the present invention;
FIG. 2 is a block diagram of an embodiment stator 48 slot rotor 8 pole double V permanent magnet flat wire motor;
FIG. 3 is a graph showing torque waveforms for a slot 8 pole permanent magnet flat wire drive motor according to example 48;
fig. 4 is a graph showing the core loss of the slot 8 pole permanent magnet flat wire drive motor of example 48;
FIG. 5 is a graph showing the eddy current loss of a slot 8 pole permanent magnet flat wire drive motor according to example 48;
FIG. 6 is a graph showing copper loss of a slot 8 pole permanent magnet flat wire drive motor according to example 48;
FIG. 7 is a schematic diagram of the prediction of a high error design of an error model versus a proxy model;
FIG. 8 is a line graph of relative error of proxy model predictions versus update times.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. The embodiments described below by referring to the drawings are exemplary only for explaining the present invention and are not to be construed as limiting the present invention.
The motor parameter design is unavoidable before the motor is put into production, and in the simulation analysis of the motor model, the finite element analysis model has a higher confidence level, but the calculation is huge in time consumption. In a scenario requiring extensive finite element model analysis computation, fitting a mapping relationship between finite element model inputs and outputs using a proxy model is a more common approach. However, if the agent model meets the requirement of higher precision, enough finite element model simulation analysis calculation sample data are acquired to train the agent model. According to the technical scheme, the number of times of finite element model simulation analysis and calculation is reduced as much as possible, the motor parameter design with higher error when the proxy model predicts the motor performance is distinguished, the finite element model simulation analysis and calculation is only carried out on the motor design with higher error when the proxy model is updated, and finally the obtained proxy model with higher precision does not need to consume too much time.
Taking the process of establishing a performance database of a stator 48 slot rotor 8 pole double V permanent magnet flat wire motor as an example, the implementation process of the invention is described in detail, and fig. 2 is a structural diagram of the motor.
The whole implementation process of the invention is shown in figure 1, firstly, an initial motor performance data set is collected through a test design and is randomly divided into two groups, namely an S data set and an X data set, and a proxy model K 0 is established through the S data set and is used for predicting motor performance; then predicting another group of data sets through the agent model K 0, comparing the predicted value with the true value, calculating the relative error e of the data sets predicted by the agent model, and training a support vector machine regression model to serve as an error model C 0 for judging whether the response of the agent model K 0 to a certain design is a high error response or not; and finally, continuously screening out a high-error motor design of the proxy model by using an error model C 0, calculating the high-error motor design by using finite element simulation analysis, adding the result into a proxy model training set S, updating the proxy model, and updating the error model along with the update of the proxy model until the precision of the proxy model reaches the required precision or exceeds the set update times of the proxy model.
In the dual model strategy proposed by the present invention, first, the motor parameter input variables and motor performance output responses of the proxy model database to be constructed are determined, and in this embodiment, the selected motor parameter input variables and initial parameters are shown in table 1.
Table 1 stator 48 slot rotor 8 pole double V permanent magnet flat wire motor input variable and initial parameter
| Sequence number | Input variable description | Variable reference name | Initial value |
| 1 | Stator inner diameter/mm | Rsi | 73.5 |
| 2 | Stator slot width/mm | Wslot_open | 1 |
| 3 | Stator slot height/mm | Wtip | 1 |
| 4 | Stator slot width-flat wire length/mm | Wslot_Wcoil | 0.9 |
| 5 | Stator slot depth-flat wire width 8/mm | Hslot_Lcoil | 0.5 |
| 6 | Distance/mm between outer diameter of stator and slot body | Rso_slot | 20 |
| 7 | L1 permanent magnet in-slot air gap length/mm | d_FB1(d_pm_1-FB_W1) | 2 |
| 8 | L1 permanent magnet width/mm | Wpm_1 | 4.5 |
| 9 | L1 permanent magnet length/mm | Lpm_1 | 20 |
| 10 | L1 permanent magnet bridge width/mm | FB_W1 | 2 |
| 11 | L 2 permanent magnet bridge width/mm | FB_W2 | 2 |
| 12 | L 2 permanent magnet in-slot air gap length/mm | d_FB2(d_pm_2-FB_W2) | 1.77 |
| 13 | L 2 permanent magnet width/mm | Wpm_2 | 3.5 |
| 14 | L 2 permanent magnet length/mm | Lpm_2 | 10 |
| 15 | Flat wire width/mm | Wcoil | 4.3 |
| 16 | Width of flat wire: length of flat wire | W_L_ratio(Wcoil/Lcoil) | 2.5 |
| 17 | Stator slot depth/mm | dcoil | 0.5 |
| 18 | Motor stack thickness/mm | lstk | 150 |
And (3) changing the value of an input variable through Latin hypercube sampling (Latin hypercube sampling, LHS), and performing finite element model simulation analysis and calculation to obtain an initial data set of motor performance. In this embodiment, the obtained motor performance is a torque average value (as shown in fig. 3) of a motor operation condition stabilization period and a total loss average value of the motor, wherein the total loss includes a motor core loss, an eddy current loss and a copper loss (as shown in fig. 4-6 respectively), and 100 groups of data are collected.
Then constructing a proxy model and an error model through the initial data set, namely randomly dividing the initial motor performance data set into two groups, namely an S data set and an X data set, and carrying out the maximum normalization on the data sets, wherein the normalization formula is as follows:
Where y j is the j-th component of the pre-normalized data point and y 'j is the j-th component of the post-normalized data point' For the minimum value of the j-th component of all data points in the dataset,/>The maximum value of the j-th component for all data points in the dataset.
Establishing a proxy model K 0 through motor input variable values and output response values in the S data set, wherein the proxy model K 0 is used for fitting a mapping relation between motor parameters and motor performance, and a Kriging model is selected to construct the proxy model in the embodiment; and then predicting the motor performance corresponding to the motor input variable in the other group of data sets through the proxy model K 0, comparing the predicted value with the true value, and calculating the relative error e of the data sets predicted by the proxy model, wherein the relative error calculation formula is as follows:
Taking the relative error e calculated by predicting each data point by the proxy model K 0 as an output response value, taking the Euclidean distance between the normalized data point X epsilon X and the data point S epsilon S constructing the proxy model as an input variable X ', carrying out second normalization on the relative error e serving as the output response value and the data point X' serving as the input variable, training a support vector machine regression model to serve as an error model C 0 for judging whether the response of the motor design to be judged by the proxy model K 0 is high error response or not, and calculating the j-th component X 'j of the input variable X' of the error model by the following formula:
Where n is the dimension of data point x, d (x, S *) represents the euclidean distance between data points x and S *, S * is the data point in the dataset S where the absolute value of the difference between the jth component and the jth component is the smallest, and d (x, S) represents the smallest distance between data point x and set S, i.e., d (x, S) =inf { d (x, S) |s e S }. And performing super-parameter tuning on the error model C 0 by adopting a grid search method to obtain an error model C 0 with more accurate judgment on the high-error motor design of the proxy model.
The motor design of the agent model to be judged can be obtained through random search, latin hypercube sampling or each generation of decision solution of a multi-objective optimization algorithm NSGA-II.
As shown in fig. 7, the first 50 data points to the left of the dotted line are data points in the X data set, and the 30 data points to the right of the dotted line are randomly set to test the error model prediction effect, wherein the data points satisfying the condition (error is equal to or greater than the average value of the relative error of the proxy model prediction motor design data set+p×the standard deviation of the relative error of the proxy model prediction motor design data set) are set as proxy model high error motor design data points, and p is 0 in this embodiment.
And finally, continuously screening out a high-error motor design of the proxy model by using the error model, calculating the high-error motor design by using finite element simulation analysis, adding the result into a proxy model training set S, updating the proxy model, calculating the relative error of the motor parameter in the updated proxy model prediction training set X as the output response of the updated error model according to the motor performance corresponding to the motor parameter in the updated proxy model prediction training set X, recalculating the input variable of the updated error model according to the output response of the updated error model, retraining the error model, and keeping the data set X used for training the error model unchanged all the time in the process. And continuously updating the proxy model and the error model until the precision of the proxy model reaches the required precision or meets the set updating times of the proxy model. As shown in fig. 8, the number of updates is set to 3 in the present embodiment, and the average relative error of the proxy model is basically unchanged each time, but the maximum relative error is continuously reduced.
Based on the same inventive concept, an embodiment of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the aforementioned method for constructing a high-precision motor performance database based on a dual-model strategy when executing the computer program.
Based on the same inventive concept, embodiments of the present application provide a computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the aforementioned method of constructing a high-precision motor performance database based on a dual-model strategy.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above embodiments are merely illustrative of the present invention, and not limiting the scope of protection thereof, and those skilled in the art may also change part of the present invention, for example, the motor type and motor performance to be optimized may be selected according to the actual situation, the proxy model type used to fit the mapping relationship between the motor parameter design and the motor performance may be selected according to the actual situation, the proxy model may be defined according to the actual situation to predict the high-error motor design, the appropriate error model type may be selected according to the actual situation, and any equivalent substitution according to the spirit of the present invention falls within the scope of protection of the present invention.
Claims (9)
1. The method for constructing the high-precision motor performance database based on the dual-model strategy is characterized by comprising the following steps of:
Acquiring an initial motor performance data set, randomly dividing the initial motor performance data set into an S data set and an X data set, and establishing a proxy model for fitting a mapping relation between motor parameters and motor performance through motor input variables and output response values in the S data set; predicting motor performance corresponding to motor parameters in the X data set through the proxy model, comparing a predicted value with a true value, calculating relative errors of the X data set predicted by the proxy model, and taking the relative errors calculated by each data point in the X data set predicted by the proxy model as an output response value of the proxy model;
constructing an input variable of an error model, and training a support vector machine regression model as an error model for judging whether the response of the motor design to be judged is high error response or not by using the support vector machine regression model as a judgment proxy model;
Screening out a proxy model high-error motor design by using an error model, calculating the high-error motor design by using finite element simulation analysis, adding a calculation result into an S data set, updating a proxy model, predicting motor performance corresponding to motor parameters in an X data set according to the updated proxy model, calculating relative errors of the motor parameters in the X data set predicted by the updated proxy model as output response of the updated error model, updating input variables of the error model, and retraining the error model; and continuously updating the proxy model and the error model until the precision of the proxy model reaches the required precision or meets the set updating times of the proxy model.
2. The method for constructing a high-precision motor performance database based on a dual-model strategy according to claim 1, wherein the initial motor performance data set is obtained by performing finite element model simulation analysis and calculation on a motor through Latin hypercube sampling.
3. The method for constructing a high-precision motor performance database based on a dual-model strategy according to claim 1, wherein the motor design to be judged is obtained through random search, latin hypercube sampling or multi-objective optimization algorithm NSGA-II per-generation decision solution.
4. The method for constructing a high-precision motor performance database based on a dual-model strategy according to claim 1, wherein the establishment and update process of the proxy model is as follows:
And performing super-parameter tuning on the proxy model by adopting a grid search method, guiding the proxy model by using an error model to obtain a high-error motor design, calculating the motor performance of the high-error motor design by finite element simulation analysis, adding a calculation result into a data set for training the proxy model, and continuously updating the proxy model until the proxy model meets the precision requirement.
5. The method for constructing a high-precision motor performance database based on a dual-model strategy according to claim 1, wherein the error model judges the standard of the high-error motor design as follows:
if the relative error value of a motor design is larger than the average value of the relative errors of the X data sets predicted by the proxy model and the standard deviation of the relative errors of the X data sets predicted by the p multiplied by the proxy model, the motor design is judged to be a high-error motor design, and p is a pre-designed value.
6. The method for constructing a high-precision motor performance database based on a dual-model strategy according to claim 1, wherein the method for constructing the input variable of the error model is as follows:
Normalizing the S data set and the X data set, for the jth component X j of the data points X in the X data set, selecting a data point S * corresponding to the jth component with the smallest difference value with X j in the jth components of all the data points in the S data set, calculating Euclidean distance between the two data points as the jth component of a new data point X ', and calculating the distance between the data point X and the S data set as the last component of the new data point X'; carrying out the same treatment on each data point in the X data set, and carrying out the second normalization on the new data point set obtained after the treatment is completed to be used as an input variable of an error model;
the j-th component of the new data point x' is specifically as follows:
Where d (X, S *) represents the Euclidean distance between data points X and S *, n is the dimension of each data point of the X dataset, d (X, S) represents the minimum distance between data point X and the S dataset, i.e., d (X, S) =inf { d (X, S) |s ε S }, S 1,...,sn represents the component of data point S in the S dataset, X 'j represents the jth component of the new data point X';
The normalization formula for all datasets is:
Where y j is the jth component of the pre-normalized data point, y' j is the jth component of the post-normalized data point, For the minimum value of the j-th component of all data points in the dataset,/>The maximum value of the j-th component for all data points in the dataset.
7. The method for constructing a high-precision motor performance database based on a dual-model strategy according to claim 1, wherein the error model updating method is as follows:
Predicting motor performance corresponding to motor parameters in the X data set by using the updated proxy model, comparing the predicted value with a true value, calculating the relative error of the updated proxy model for predicting the X data set, and taking the relative error calculated by each data point in the updated proxy model for predicting the X data set as an output response value of the updated proxy model; the input variable construction mode of the error model is unchanged, and the error model is retrained to obtain an updated error model.
8. A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor, when executing the computer program, implements the steps of the method of constructing a high-precision motor performance database based on a dual-model strategy as claimed in any one of claims 1 to 7.
9. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method of constructing a high precision motor performance database based on a dual model strategy as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410064673.7A CN117972924A (en) | 2024-01-16 | 2024-01-16 | Method for constructing high-precision motor performance database based on dual-model strategy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410064673.7A CN117972924A (en) | 2024-01-16 | 2024-01-16 | Method for constructing high-precision motor performance database based on dual-model strategy |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117972924A true CN117972924A (en) | 2024-05-03 |
Family
ID=90860876
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410064673.7A Pending CN117972924A (en) | 2024-01-16 | 2024-01-16 | Method for constructing high-precision motor performance database based on dual-model strategy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117972924A (en) |
-
2024
- 2024-01-16 CN CN202410064673.7A patent/CN117972924A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110163429B (en) | Short-term load prediction method based on similarity day optimization screening | |
| CN112018758A (en) | Modeling method of high-proportion new energy-containing alternating current-direct current hybrid system based on digital twinning | |
| CN110276442A (en) | A kind of searching method and device of neural network framework | |
| CN114428803B (en) | Air compression station operation optimization method, system, storage medium and terminal | |
| CN114253157B (en) | Motor multi-parameter optimization method and system based on second-order sensitivity analysis | |
| CN110609975B (en) | Parallel computing global optimization method based on Kriging agent model | |
| CN112597610B (en) | Optimization method, device and equipment for lightweight design of mechanical arm structure | |
| CN111079788A (en) | K-means clustering method based on density Canopy | |
| JP2020113044A (en) | Data expansion program, data expansion method, and data expansion device | |
| CN115481549A (en) | Cylindrical linear motor multi-objective optimization method, equipment and storage medium | |
| CN114072809A (en) | Small and fast video processing network via neural architectural search | |
| CN109074348A (en) | For being iterated the equipment and alternative manner of cluster to input data set | |
| CN114841106A (en) | Integrated circuit optimization method and system based on rule-guided genetic algorithm | |
| CN114564787A (en) | Bayesian optimization method, device and storage medium for target-related airfoil design | |
| CN113592064A (en) | Ring polishing machine process parameter prediction method, system, application, terminal and medium | |
| CN117093847A (en) | Statistical delay characteristic calculation method and system based on Gaussian process regression | |
| CN117972924A (en) | Method for constructing high-precision motor performance database based on dual-model strategy | |
| CN116861772A (en) | Parameter optimization method and system for multi-parameter finish machining | |
| Wang et al. | An estimation of distribution algorithm for the flexible job-shop scheduling problem | |
| CN111241749A (en) | Permanent magnet synchronous motor chaos prediction method based on reserve pool calculation | |
| Xu et al. | Motor fault diagnosis method based on deep learning | |
| CN115423149A (en) | Incremental iterative clustering method for energy internet load prediction and noise level estimation | |
| CN114358251A (en) | Neural network architecture searching method based on graph value network | |
| CN115688873A (en) | Graph data processing method, device and computer program product | |
| CN112465105A (en) | Computer-readable recording medium recording learning program and learning method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |