Large and small blade machine learning optimization design method based on NURBS curved surface modeling
Through machine learning methods based on NURBS surface modeling, turbine blade design is optimized, and the problems of large calculation volume and low accuracy of traditional design methods are solved, fast and accurate blade design is achieved, and aerodynamic performance and calculation efficiency are improved.
Patent Information
- Application Number
- CN202410919652.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-10
- Publication Date
- 2024-11-08
AI Technical Summary
The traditional turbine blade design method has high calculation accuracy but large calculation amount, and it is difficult to quickly obtain accurate flow field analysis. The experimental measurement cost is high and affected by errors, so it is impossible to effectively match the flow performance of large and small blades.
Using machine learning optimization design method based on NURBS surface modeling, by obtaining theoretical data of turbine blades, determining the leaf type value points and control vertices of large and small blades, establishing a three-dimensional flow channel model, performing data dimensionality reduction and neural network construction, and seeking the best Optimize design parameters, reduce calculation amount and improve design efficiency.
It realizes rapid and precise optimization of design parameters, improves the aerodynamic performance of the blade, compresses computing resources, and simplifies the design process.
Smart Images

Figure CN118917174A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of turbine blade optimization design, and in particular to a machine learning optimization design method for large and small blades based on NURBS surface modeling. Background Art
[0002] High-efficiency turbine aerodynamic design can effectively reduce engine mass, improve engine efficiency, and enhance engine performance. Conventional turbine modeling methods have been relatively mature, and a variety of different modeling tools have been developed. Regardless of the method, the basic idea is to parameterize the blade geometry and combine it with smooth curves to complete the modeling. The design of the large and small blade combination cascade that integrates the transition section support plate and the low-pressure turbine guide vane is different from the conventional blade design. The flow matching of large and small blades must be considered in the design, and the structural design requirements for the geometric size and stacking form of large blades must be met. Due to the existence of two different blade shapes, large and small, it is impossible to use a unified modeling parameter representation like conventional blades.
[0003] In related technologies, after the target aerodynamic parameters of the turbine blade are given, the traditional high-precision numerical simulation method is usually used to continuously change the blade geometry and perform simulation calculations. The calculation accuracy is high, but the calculation amount is large and the calculation time is long. Although low-precision simulation can quickly obtain calculation results, it cannot be used for accurate flow field analysis. Experimental measurement can reflect the actual situation of the flow field, but the manpower, material resources and time costs are high, and it will be affected by measurement and manufacturing technology errors.
[0004] In addition, the traditional blade design method requires the derivation of an iterative relationship between each value point and the iterative calculation in sequence. The process formula is cumbersome and the amount of calculation is large. It is even more difficult to implement programming and modification on a computer.
[0005] As the only mathematical method for defining and designing the geometric shapes of industrial products, it is particularly important to reasonably and effectively represent various NURBS curves and surfaces. NURBS is the abbreviation of Non-Uniform Rational B-Splines. This surface modeling method can accurately represent and design standard analytic geometric shapes (i.e. elementary curves and surfaces and complex free-form curves and surfaces, providing a common standard mathematical expression for these curves and surfaces. The control vertices on the curves and surfaces and the corresponding weights are applied separately or in combination to provide flexibility for the design and modification of various curve and surface shapes, among which the introduction of weights replaces the shape parameters in the previous geometric continuous spline curves and surfaces.
[0006] Principal component analysis (PCA) is a common data analysis method, which is often used for dimensionality reduction of high-dimensional data and can be used to extract the main characteristic components of the data. PCA is usually used to reduce the dimensionality of large data sets by reducing the number of indicators in the data set and retaining most of the information of the indicators in the original data set. The advantage of PCA is that it reduces the dimensionality of the data, facilitates the extraction of the main features of the data, makes the data easier to use, reduces computational overhead, removes noise, etc. PCA can not only compress the data, but also make the data features after dimensionality reduction independent of each other. The principal component analysis method is used to reduce the dimensionality of the flow field simulation results, so that the k modal flow fields after dimensionality reduction are independent of each other.
[0007] Machine learning provides a new entry point for the optimization design of turbine blades. As a subfield of current artificial intelligence technology, machine learning algorithms are algorithms that rely on patterns and statistical inferences. They enable computer systems to complete specific tasks without special programming, i.e., soft programming. Researchers use machine learning algorithms to guide computers to train appropriate machine learning models using known data, and use the trained models to make predictions for new data in new situations. As an algorithm model in machine learning, neural networks have strong nonlinear mapping capabilities. It has a multi-layer network structure, each layer consisting of multiple neurons, and neurons are connected by weights. The data is divided into training data and test data, and the model is trained and learned with the training data, the weights of the neurons are changed, so that it can obtain regression prediction capabilities and generate a model with generalization capabilities. The test data is data that the neural network has not seen before, and is input into the trained model to test its prediction accuracy. The use of machine learning methods in the optimization design of large and small blades, given the aerodynamic parameters of the blades, can quickly and accurately predict the geometric parameters, reduce the amount of calculation, and improve the design efficiency. This is a new attempt outside the traditional numerical simulation calculation method. Summary of the invention
[0008] The present application aims to solve one of the technical problems in the related art at least to some extent.
[0009] To this end, the purpose of this application is to propose a machine learning optimization design method for large and small blades based on NURBS surface modeling, aiming to obtain the optimized geometric parameters of small blades according to the three-dimensional parameters of large turbine blades, compress the calculation amount and save computing resources.
[0010] To achieve the above purpose, the present application embodiment proposes a machine learning optimization design method for large and small blades based on NURBS surface modeling, including:
[0011] Obtain theoretical data of the original turbine blade size of the aircraft to be designed, determine the blade value points and control vertices of the large blade using the NURBS surface modeling method, and obtain the blade value points and control vertices of the small blade by scaling;
[0012] Constructing a three-dimensional flow channel model of the original turbine blade size, and meshing the three-dimensional flow channel model of the original turbine blade size;
[0013] Extract the blade design parameters and target parameter data of the original turbine blade size of the aircraft to be designed, establish a data set from the blade NURBS surface control vertices to the blade aerodynamic parameters, and perform data dimension reduction on the data set;
[0014] Create a model for turbine blade optimization design, build a neural network, and seek the optimal design parameter data.
[0015] Among them, the theoretical data of the large and small blades of the aircraft engine to be designed are obtained, and the blade value points and control vertices of the large blade are determined by the NURBS surface modeling method, including:
[0016] Combined with the theoretical data of the specified blades in the aircraft engine, the blades are constructed using the bicubic NURBS surface modeling method;
[0017] The inverse calculation method of bicubic NURBS surface is used to obtain the required blade value points and control vertices on the large blade surface, and the obtained control vertices are used to generate the surface in Python.
[0018] Among them, the blade shape value points and control vertices of the small blade are obtained by scaling, including:
[0019] In the traditional turbine transition section blade modeling method, the small blade is proportionally reduced according to the percentage of the axial chord length of the large blade to obtain its blade theoretical data;
[0020] The large blade shape points and control vertices are determined using the NURBS surface modeling method.
[0021] Among them, the three-dimensional flow channel model of the original turbine blade size is meshed, and the three-dimensional flow channel model of the original turbine blade size is meshed using the O4H grid topology.
[0022] Among them, the blade design parameters and target parameter data of the turbine blades of the aircraft to be designed are extracted, and a data set from the control vertices of the blade NURBS surface to the blade aerodynamic parameters is established, including:
[0023] The training data obtained by using the three-dimensional flow grid calculation is used to extract the turbine blade design parameters and target parameter data;
[0024] Taking blade geometric parameters as independent variables and blade aerodynamic parameters as dependent variables, a data set from blade geometric parameters to blade aerodynamic parameters is established;
[0025] The sample space of the modified turbine blade three-dimensional blade profile is generated by taking the control vertex ±20% normal offset as the upper and lower bounds of the sample space;
[0026] A preset number of three-dimensional blade profiles are selected from the sample space of the three-dimensional blade profiles of modified turbine blades as a database, and the three-dimensional blade profiles of the modified turbine blades contained in the database are numerically simulated using computational fluid dynamics methods to obtain a three-dimensional numerically simulated flow field of the turbine, and the obtained numerically simulated flow fields are combined into a numerically simulated flow field database.
[0027] Among them, the data dimension reduction of the numerical simulation flow field database includes:
[0028] The principal component analysis method is used to reduce the dimension of the numerical simulation flow field database, so that the modal flow fields after dimension reduction are independent of each other.
[0029] The linear combination of modal flow fields is used to reconstruct the numerical simulation flow field of the large and small turbine blades to obtain the reconstructed flow field of the large and small turbine blades.
[0030] Among them, a model for turbine blade optimization design is created and a neural network is built, including:
[0031] The database of modified large and small blade profiles is divided into training set and test set in a ratio of 2:1;
[0032] The three-dimensional parameters of the turbine blade size are used as the input parameters of the neural network, and the weight of the modal flow field is used as the output.
[0033] Among them, the optimal design parameter data is sought, including:
[0034] The neural network model is used as an algorithm evaluation method to optimize and obtain the three-dimensional blade data of turbine blades of different sizes.
[0035] Computational fluid dynamics methods are used to verify the optimization results.
[0036] Different from the prior art, the present invention provides a machine learning optimization design method for large and small blades based on NURBS surface modeling. By obtaining theoretical data of aircraft engine turbine blades, the large blade shape value points and control vertices are determined by the NURBS surface modeling method; the blade shape determination value and control vertex of the small blade are obtained by the scaling method; the model grid of the three-dimensional flow channel of large and small blades is divided; the data of the blade shape design parameters and target parameters of the turbine large and small blades are extracted, and a data set from the blade NURBS surface control vertex to the blade aerodynamic parameters is established; data dimension reduction is performed; a model for the design of turbine blade optimization problems is created, a neural network is built, and the optimal design parameter data is sought. The present invention can improve the aerodynamic performance of the blade and obtain the three-dimensional design parameters of the blade. At the same time, the machine learning optimization design method for large and small blades based on NURBS surface modeling also has the advantages of compressing the amount of calculation and saving computing resources.
[0037] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] Figure 1 It is a flow chart of a method for machine learning optimization design of large and small blades based on NURBS surface modeling provided by the present invention;
[0039] Figure 2 It is a schematic diagram of blade type value points on 9 cross sections of a machine learning optimization design method for large and small blades based on NURBS surface modeling provided by the present invention;
[0040] Figure 3 It is a schematic diagram of control vertices on a turbine blade surface in a machine learning optimization design method for large and small blades based on NURBS surface modeling provided by the present invention;
[0041] Figure 4 The invention provides a NURBS bicubic surface pressure surface view of a turbine blade constructed by a machine learning optimization design method for large and small blades based on NURBS surface modeling;
[0042] Figure 5 The present invention provides a NURBS bicubic leading edge view of a turbine blade constructed by a machine learning optimization design method for large and small blades based on NURBS surface modeling;
[0043] Figure 6 It is a flow chart of data dimensionality reduction in a machine learning optimization design method for large and small blades based on NURBS surface modeling provided by the present invention. DETAILED DESCRIPTION
[0044] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.
[0045] The following describes, with reference to the accompanying drawings, a machine learning optimization design method for large and small blades based on NURBS surface modeling in an embodiment of the present application.
[0046] like Figure 1 As shown, the embodiment of the present application provides a machine learning optimization design method for large and small blades based on NURBS surface modeling, including:
[0047] Step S101: Obtain theoretical data of the original turbine blade size of the aircraft to be designed, determine the blade value points and control vertices of the large blade using the NURBS surface modeling method, and obtain the blade value points and control vertices of the small blade by scaling.
[0048] In the embodiment of the present invention, the theoretical data of a blade on an aircraft engine is known, a section of the blade profile is cut, and 9 key sections along the blade height direction are uniformly selected as an example. Assuming that these key sections (i.e., cross sections) are in the u direction, and the blade height direction is in the v direction, 45 points are selected on each cross section as the shape value points for constructing the bicubic NURBS surface, such as Figure 2 The figure shows the shape value points on nine cross sections, where the shape value points at the edge transition of the blade cross section need to be denser in order to better meet the accuracy requirements. Due to the large amount of data for constructing the curve on the blade cross section, only the generation process of the cubic NURBS curve on one of the cross sections is given here.
[0049] Given 45 type value points p i The coordinate values of (i=0,1,…,44) are as follows. Since the bending is more serious in places where the value points are dense on the cubic NURBS curve, the parameterization method of the data points adopts the centripetal parameterization method. The node vector of the cubic NURBS curve can be obtained by calculation. It is known that the section curve is a closed curve, so the control vertex d i =(i=0,1,…,46) is obtained from the matrix expression of the inverse calculation of the control vertices of the following cubic NURBS closed curve. For a continuous NURBS cubic closed curve, since its first and last data points overlap, n-1 equations are sufficient to find n+1 control vertices, so the matrix expression of the inverse calculation of the control vertices of the NURBS cubic closed curve is:
[0050]
[0051] in:
[0052]
[0053]
[0054]
[0055] e i =(Δ i+1 +Δ i+2 ) i-1
[0056] Δ i =u i+1 -u i (i=1,2,…,n-2)
[0057] The control vertex d obtained by inverse calculation i =(i=0,1,…,46) and the node vector U, combined with the weight factor ω corresponding to the control vertex i (i=0,1,…,46), for convenience, the weight factors here are all taken as 1. The required blade section curve can be constructed by applying the following matrix expression of cubic NURBS curve. The generated curve is as follows Figure 3 As shown, the given type value points are all on the generated curve.
[0058]
[0059] in,
[0060] T 3 =(1 tt 2 t 3 )0≤t≤1
[0061] D i =(ω i d i ω i+1 d i+1 ω i+2 d i+2 ω i+3 d i+3 ) T
[0062] W i =(ω i ω i+1 ω i+2 ω i+3 ) T
[0063]
[0064] Δ i =u i+1-u i (i=0,1,…,n+4)
[0065]
[0066]
[0067]
[0068] The same method as above is used to construct the NURBS cubic curves of the other 8 blade sections. The control vertices on these 9 section curves are used as the “type value points” in the blade height direction (i.e., v direction) to construct these 47 v-direction NURBS cubic curves. Due to the large amount of data, the generation method of only one of the curves is given here.
[0069] Since the V-curve is relatively gentle, the accumulated chord length parameterization method can be used to parameterize the type value points to obtain the node vector.
[0070] The v-curve is an open curve, and the control vertices are calculated by using the matrix expression of the control vertices of the cubic NURBS open curve inverse calculation. j (i=0,1,…,10) weight factors are all taken as 1, combined with the node vector and the following formula, we have the control vertex d of the curve j =(i=0,1,…,10).
[0071]
[0072] a 1 =c n-1 =1,b 1 =c 1 =a n-1 =b n-1 =0
[0073]
[0074] Among them, a i , b i 、c i 、e i Same as the matrix expression in the inverse calculation of the control vertices of a NURBS cubic closed curve, Δ i =u i+1 -u i , p i (i=0,1,…,m) is the given type value point of the curve, p′ 0 and p′ m is the tangent vector of the first and last points of the shape value. Solve the above linear equations, d i =(i=0,1,…,n) are all the required control vertices.
[0075] Control vertex combined node vector V and weight factor ω j (i=0,1,…,10), the required v-curve can be constructed by applying the matrix expression of the cubic NURBS curve.
[0076] Using the same method as above, another 46 v-direction NURBS cubic curves can be constructed. All control vertices on these curves are the control vertices on the blade surface. The blade surface can be constructed by using the following matrix expression of bicubic NURBS surface and combining u and v-direction node vectors and their corresponding weight factors.
[0077]
[0078] in,
[0079] i=0,1,…,m-3; j=0,1,…,n-3
[0080] U=[1 α α 2 α 3 ],V=[1 β β 2 β 3 ]
[0081]
[0082]
[0083]
[0084]
[0085] Furthermore, the blade shape determination value and control vertex of the small blade are obtained by the scaling method. For the blade shapes of large and small blades, the blade shape theoretical data of the small blade is obtained by proportionally reducing the proportion of the axial chord length in percentage, and the parameters of its control vertex are obtained through the NURBS control vertex calculation in the same step 1. Figure 4 and Figure 5 The blade surface is composed of bicubic NURBS surfaces.
[0086] Step S102: constructing a three-dimensional flow channel model of the original turbine blades, and meshing the three-dimensional flow channel model of the original turbine blades.
[0087] For the meshing of the model of the three-dimensional flow channel of the large and small blades, for the blade profile in step S101, an O4H type mesh topology is used for meshing, and the total number of mesh nodes is m.
[0088] Step S103: extracting the blade profile design parameters and target parameter data of the original turbine size of the aircraft to be designed, establishing a data set from the blade NURBS surface control vertices to the blade aerodynamic parameters, and performing data dimension reduction on the data set.
[0089] The blade geometry parameters are used as independent variables and the blade aerodynamic parameters are used as dependent variables to establish a data set from blade geometry parameters to blade aerodynamic parameters. The sample space of the modified turbine blade three-dimensional blade profile is generated by controlling the vertex ±20% normal offset as the upper and lower bounds of the sample space. n three-dimensional blade profiles are selected from the sample as a database, n≥100; the computational fluid dynamics method is used to perform numerical simulation on the n three-dimensional blade profiles included in the database to obtain the turbine three-dimensional numerical simulation flow field. The obtained numerical simulation flow field constitutes the database X={A 1 , A 2 , …, A n} T , where A represents a numerical simulation example of a blade profile, subscript n represents the example number, and A = {a 1 , a 2 , …, a m} T , where a represents the numerical simulation data at a grid point, and the subscript m represents the total number of grid points.
[0090] Data dimensionality reduction includes: neutralizing the above database and constructing covariance. Standardizing the range of continuous initial variables (converting unstructured to structured), calculating the covariance matrix to identify correlations, calculating the eigenvectors and eigenvalues of the covariance matrix to identify principal components, creating eigenvectors to determine which principal components to retain, and recasting the data along the principal component axis. The data dimensionality reduction process is as follows: Figure 6 shown.
[0091] The above database is neutralized and the covariance is constructed. The database is neutralized using the following formula:
[0092]
[0093] Take the eigenvector w corresponding to the largest first k eigenvalues of the covariance matrix 1 ,w 2 ,…,w k , diagonalize the covariance matrix, k≤(1 / 10)n; select a set of eigenvectors w that can reduce the covariance to zero and maximize the variance 1 ,w 2 ,…,w kThe target projection matrix is formed. The principal component analysis method is used for dimensionality reduction, so that the k modal flow fields after dimensionality reduction are independent of each other; the linear combination of the k modal flow fields is used to reconstruct the numerical simulation flow field of the large and small turbine blades to obtain the reconstructed flow field of the large and small turbine blades.
[0094] Step S104: Create a model for turbine blade optimization design, build a neural network, and seek optimal design parameter data.
[0095] The database of n modified blades is divided into training set and test set in a ratio of 2:1. The three-dimensional parameters of the turbine blades are used as the input parameters of the neural network, and the weights of the k modal flow fields are used as the output. The samples of the training set are trained, and the weights of the k modal flow fields obtained by principal component analysis are tested using the test set.
[0096] The three-dimensional reconstructed flow field of the turbine is compared with the numerical simulation flow field of the three-dimensional turbine blade, and the performance parameters such as static pressure rise coefficient and total pressure loss coefficient are compared. The following formula is used as the determination coefficient between the predicted data and the actual data:
[0097]
[0098] In the formula, K O To reconstruct the flow field results, K C is the numerical simulation flow field result, is the average reconstructed flow field result, is the average numerical simulation flow field result;
[0099] The three-dimensional blade profile reconstructed flow field of large and small turbine blades is evaluated, and the value range of the determination coefficient R is 0-1; if the value of the coefficient R tends to 0, the judgment accuracy is low, and the next step is repeated until the value of the determination coefficient R tends to 1, and a neural network that meets the accuracy requirements is obtained.
[0100] The neural network model is used as an evaluation method for the algorithm to perform optimization and obtain the three-dimensional blade data of turbine blades of different sizes. The optimization results are then verified using computational fluid dynamics methods.
[0101] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.
[0102] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.
[0103] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.
[0104] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0105] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0106] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.
[0107] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.
[0108] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0109] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0110] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0111] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A machine learning optimization design method for large and small blades based on NURBS surface modeling, characterized in that: include: Obtain theoretical data of the original turbine blade size of the aircraft to be designed, determine the blade value points and control vertices of the large blade using the NURBS surface modeling method, and obtain the blade value points and control vertices of the small blade by scaling; Constructing a three-dimensional flow channel model of the original turbine blade size, and meshing the three-dimensional flow channel model of the original turbine blade size; Extract the blade profile design parameters and target parameter data of the original turbine size blade of the aircraft to be designed, establish a data set from the blade NURBS surface control vertex to the blade aerodynamic parameters, and perform data dimension reduction on the data set; Create a model for turbine blade optimization design, build a neural network, and seek the optimal design parameter data.
2. The machine learning optimization design method for large and small blades based on NURBS surface modeling according to claim 1 is characterized in that: The method of obtaining theoretical data of large and small blades of the aircraft engine to be designed and determining the blade value points and control vertices of the large blade by using the NURBS surface modeling method includes: Combined with the theoretical data of the specified blades in the aircraft engine, the blades are constructed using the bicubic NURBS surface modeling method; The inverse calculation method of bicubic NURBS surface is used to obtain the required blade value points and control vertices on the large blade surface, and the obtained control vertices are used to generate the surface in Python.
3. The machine learning optimization design method for large and small blades based on NURBS surface modeling according to claim 2 is characterized in that: By scaling, the leaf shape points and control vertices of the small blade are obtained, including: In the traditional turbine transition section blade modeling method, the small blade is proportionally reduced according to the percentage of the axial chord length of the large blade to obtain its blade theoretical data; The large blade shape points and control vertices are determined using the NURBS surface modeling method.
4. The machine learning optimization design method for large and small blades based on NURBS surface modeling according to claim 1 is characterized in that: The three-dimensional flow channel model of the original turbine blades is meshed by using an O4H mesh topology to mesh the three-dimensional flow channel model of the original turbine blades.
5. The machine learning optimization design method for large and small blades based on NURBS surface modeling according to claim 1 is characterized in that: The method of extracting the blade profile design parameters and target parameter data of the turbine blade of the aircraft to be designed, and establishing a data set from the control vertices of the blade NURBS surface to the blade aerodynamic parameters, includes: Extracting data of blade profile design parameters and target parameters of turbine blades of different sizes using the training data obtained by the three-dimensional flow grid calculation; Taking blade geometric parameters as independent variables and blade aerodynamic parameters as dependent variables, a data set from blade geometric parameters to blade aerodynamic parameters is established; Using the control vertex ±20% normal offset as the upper and lower bounds of the sample space, a sample space of the three-dimensional blade profile of the modified turbine blade is generated; A preset number of three-dimensional blade profiles are selected from the sample space of the modified turbine blade three-dimensional blade profiles as a database, and computational fluid dynamics methods are used to perform numerical simulation on the modified turbine blade three-dimensional blade profiles contained in the database to obtain a three-dimensional numerically simulated flow field of the turbine, and the obtained numerically simulated flow fields are assembled into a numerically simulated flow field database.
6. The machine learning optimization design method for large and small blades based on NURBS surface modeling according to claim 5 is characterized in that: Performing data dimension reduction on the numerical simulation flow field database includes: The principal component analysis method is used to reduce the dimension of the numerical simulation flow field database so that the modal flow fields after the dimension reduction are independent of each other; The linear combination of the modal flow fields is used to reconstruct the numerical simulation flow fields of the large and small turbine blades to obtain the reconstructed flow fields of the large and small turbine blades.
7. The machine learning optimization design method for large and small blades based on NURBS surface modeling according to claim 1 is characterized in that: The method of creating a model for the optimization problem of turbine blades and building a neural network includes: The database constructed by the modified large and small blade profiles is divided into a training set and a test set in a ratio of 2:1; The three-dimensional parameters of the turbine blades are used as input parameters of the neural network, and the weight of the modal flow field is used as output.
8. The machine learning optimization design method for large and small blades based on NURBS surface modeling according to claim 1 is characterized in that: Seek the optimal design parameter data, including: The neural network model is used as an algorithm evaluation method to perform optimization and obtain three-dimensional blade profile data of large and small turbine blades; Computational fluid dynamics methods are used to verify the optimization results.
Citation Information
Cited By
Multi-angle radar image-based three-dimensional imaging method for multiple moving targets
CN120669241A
