CN113705077A - Turbine blade anti-problem design method based on machine learning, computer-readable storage medium and electronic device - Google Patents
Turbine blade anti-problem design method based on machine learning, computer-readable storage medium and electronic device Download PDFInfo
- Publication number
- CN113705077A CN113705077A CN202110790078.8A CN202110790078A CN113705077A CN 113705077 A CN113705077 A CN 113705077A CN 202110790078 A CN202110790078 A CN 202110790078A CN 113705077 A CN113705077 A CN 113705077A
- Authority
- CN
- China
- Prior art keywords
- blade
- parameters
- model
- problem design
- geometric
- 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
- 238000013461 design Methods 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000010801 machine learning Methods 0.000 title claims abstract description 39
- 238000003860 storage Methods 0.000 title claims abstract description 7
- 238000012549 training Methods 0.000 claims abstract description 57
- 238000004364 calculation method Methods 0.000 claims abstract description 22
- 238000005457 optimization Methods 0.000 claims abstract description 9
- 230000001419 dependent effect Effects 0.000 claims abstract description 7
- 238000012360 testing method Methods 0.000 claims description 33
- 230000006870 function Effects 0.000 claims description 27
- 238000004422 calculation algorithm Methods 0.000 claims description 17
- 238000003062 neural network model Methods 0.000 claims description 11
- 230000004913 activation Effects 0.000 claims description 9
- 238000011156 evaluation Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 6
- 230000008901 benefit Effects 0.000 abstract description 3
- 238000013528 artificial neural network Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 5
- 238000004088 simulation Methods 0.000 description 5
- 238000009826 distribution Methods 0.000 description 4
- 230000001537 neural effect Effects 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- 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
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Turbine Rotor Nozzle Sealing (AREA)
Abstract
The invention relates to the technical field of turbine blade optimization design, in particular to a turbine blade inverse problem design method based on machine learning, a computer readable storage medium and electronic equipment, wherein the turbine blade inverse problem design method based on machine learning comprises the following steps: establishing a data set from the geometric parameters of the blades to the aerodynamic parameters of the blades by taking the geometric parameters of the first blades as independent variables and the aerodynamic parameters of the first blades as dependent variables; creating a training model of the turbine blade inverse problem design; training and loss evaluating the training model using the data set to determine the inverse problem design model; substituting the second blade aerodynamic parameter into the inverse problem design model to reversely predict a second blade geometric parameter. According to the method, the geometric parameters of the blade can be obtained through calculation according to the aerodynamic parameters of the blade, and meanwhile, the turbine blade anti-problem design method based on machine learning also has the advantage of high calculation precision.
Description
Technical Field
The invention relates to the technical field of optimization design of turbine blades, in particular to a turbine blade anti-problem design method based on machine learning, a computer-readable storage medium and electronic equipment.
Background
The high-efficiency turbine pneumatic design can effectively reduce the quality of the engine, improve the efficiency of the engine and enhance the performance of the engine. The inverse problem is a part of turbine pneumatic design and is also called as a blade geometric optimization and modification design problem, namely under the condition of some pneumatic performance target parameters in a given flow field, two-dimensional parameters such as speed distribution, pressure distribution, load distribution and circulation distribution, and one-dimensional total parameters such as efficiency, power and flow, the blade geometric parameters are obtained through continuous iterative calculation according to the physical relationship between the pneumatic parameters and the blade profile geometric modeling.
In the related technology, after the target aerodynamic parameters of the turbine blade are given, the traditional high-precision numerical simulation method is usually adopted, the blade geometry is continuously changed, and simulation calculation is carried out, so that the calculation precision is high, but the calculation amount is large, and the calculation time is long; although the low-precision simulation can quickly obtain a calculation result, the low-precision simulation cannot be used for accurate flow field analysis; the experimental measurement can reflect the actual situation of the flow field, but the time cost of manpower and material resources is high, and the influence of measurement and manufacturing technical errors can be received.
Machine learning provides a new entry point for turbine blade inverse problem design research. As a sub-field of current artificial intelligence technology, machine learning algorithms are algorithms that rely on patterns and statistical inferences to enable computer systems to accomplish specific tasks without requiring specialized programming, i.e., soft programming. Researchers can instruct computers to train appropriate machine learning models by using known data by using machine learning algorithms, and can predict new data under new situations by using the trained models. The neural network is used as an algorithm model in machine learning and has strong nonlinear mapping capability. The neural network has a multi-layer network structure, each layer is composed of a plurality of neurons, and the neurons are connected by adopting weights. Dividing data into training data and testing data, training and learning the model by using the training data, changing the weight of the neuron to obtain regression prediction capability, and generating the model with generalization capability. The test data is used as data which is not seen by the neural network, and is input into the trained model for testing the prediction accuracy of the model. The method applies a machine learning method to design the inverse problem of the turbine blade, gives the aerodynamic parameters of the blade, can quickly and accurately predict the geometric parameters, reduces the calculated amount and improves the design efficiency, and is a new attempt besides the traditional numerical simulation calculation method, so the following requirements are generated in the field: the design method can accurately predict the geometric parameters of the blade based on the aerodynamic parameters of the blade to be designed by utilizing the regression prediction capability of machine learning.
Disclosure of Invention
The present invention is directed to solving, at least to some extent, one of the technical problems in the related art.
Therefore, the embodiment of the invention provides a turbine blade anti-problem design method based on machine learning, the turbine blade anti-problem design method can calculate and obtain the geometric parameters of the blade according to the aerodynamic parameters of the blade, and meanwhile, the turbine blade anti-problem design method also has the advantage of high calculation precision.
The turbine blade anti-problem design method based on machine learning comprises the following steps: establishing a data set from the geometric parameters of the blades to the aerodynamic parameters of the blades by taking the geometric parameters of the first blades as independent variables and the aerodynamic parameters of the first blades as dependent variables; creating a training model of the turbine blade inverse problem design; training and loss evaluating the training model using the data set to determine the inverse problem design model; substituting the second blade aerodynamic parameter into the inverse problem design model to reversely predict a second blade geometric parameter.
According to the turbine blade inverse problem design method based on machine learning, the geometric parameters of the blades can be obtained through calculation according to the aerodynamic parameters of the blades, and meanwhile, the turbine blade inverse problem design method has the advantage of high calculation accuracy.
In some embodiments, the establishing a data set from the blade geometric parameters to the blade aerodynamic parameters with the first blade geometric parameters as independent variables and the first blade aerodynamic parameters as dependent variables includes: sampling the value range by adopting a sampling algorithm based on the value range of the blade geometric parameters to obtain a blade profile file consisting of the first blade geometric parameters;
calculating a blade shape file to obtain the pneumatic parameters of the first blade; and normalizing the first blade geometric parameters and the first blade aerodynamic parameters corresponding to the first blade geometric parameters to obtain the data set.
In some embodiments, the creating a training model of a turbine blade inverse problem design comprises: creating a 4-layer neural network model consisting of an input layer, a first hidden layer, a second hidden layer and an output layer; setting an input layer and a first hidden layer as RBF layers; and setting the second hidden layer and the output layer as full connection layers.
In some embodiments, said training and loss assessment of said training model using said dataset to determine said inverse problem design model comprises: dividing the data set into training data and test data, wherein the test data comprises the third vane pneumatic parameters and third vane geometric parameters corresponding to the third vane pneumatic parameters; training and loss evaluation are carried out on the 4-layer neural network model by utilizing the training data so as to obtain the anti-problem design model; and verifying the anti-problem design model by using the test data to determine the anti-problem design model.
In some embodiments, the training and loss evaluation of the 4-layer neural network model using the training data to obtain the inverse problem design model includes: carrying out loss evaluation on the training model by using a loss function; performing iterative optimization on the loss function by using an optimization algorithm; and evaluating the calculation result of the loss evaluation to obtain the inverse problem design model.
In some embodiments, said validating said inverse problem design model using said test data to determine said inverse problem design model comprises: substituting the pneumatic parameters of the third blade of the test data into the inverse problem design model to obtain the geometric parameters of the fourth blade; performing inverse normalization processing on the geometric parameters of the fourth blade to obtain the geometric parameters of the fifth blade; establishing a prediction geometric model based on the fifth blade geometric parameter; performing inverse normalization processing on the geometric parameters of the third blade of the test data to obtain sixth geometric parameters; establishing a true geometric model based on the sixth blade geometric parameter; and comparing and evaluating the leaf profiles of the prediction geometric model and the truth geometric model on the same leaf height to determine the inverse problem design model.
In some embodiments, after the step of substituting the third pneumatic parameter of the test data into the inverse problem design model to obtain the sixth blade geometric parameter, the method further includes: calculating the geometric parameters of the sixth blade to obtain fourth pneumatic parameters; based on the third and fourth aerodynamic parameters, an average error is calculated.
In some embodiments, the activation functions of the input layer and the first hidden layer are both radial basis functions; the activation function of the second hidden layer comprises a ReLU function; the activation function of the output layer comprises a sigmoid function.
In some embodiments, the first blade geometry parameters include stagger angle, inlet geometry angle, outlet geometry angle, leading edge pressure face wedge angle, leading edge suction face wedge angle, trailing edge wedge angle, back bend angle, leading edge diameter geometry parameters, and the first aerodynamic parameters include efficiency, power, flow, outlet relative mach number, outlet absolute mach number, outlet relative airflow angle, outlet absolute airflow angle, and reaction degree.
According to an embodiment of the invention, a computer readable storage medium has stored thereon computer instructions which, when executed by a processor, implement the turbine blade inverse problem design method based on machine learning as described in any of the above embodiments.
An electronic device according to an embodiment of the present invention includes: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to: performing the turbine blade inverse problem design method based on machine learning described in any of the above embodiments.
Drawings
FIG. 1 is a flow diagram of a method for machine learning based turbine blade anti-problem design provided in accordance with some embodiments of the present invention;
FIG. 2A is a schematic flow chart diagram illustrating step S1 of a method for designing turbine blade inverse problems based on machine learning according to some embodiments of the present invention;
FIG. 2B is a schematic flow chart diagram illustrating step S3 of a method for designing turbine blade inverse problems based on machine learning according to some embodiments of the present invention;
FIG. 2C is a schematic flow chart diagram illustrating step S33 of a method for designing turbine blade inverse problems based on machine learning according to some embodiments of the present invention;
FIG. 2D is a schematic flow chart of a comparison of third blade aerodynamic parameters and fourth blade aerodynamic parameters in a method for turbine blade anti-problem design based on machine learning provided in accordance with some embodiments of the present invention;
FIG. 3 is a schematic diagram of a 4-layer neural network model provided in accordance with some embodiments of the present invention;
FIG. 4 is a graph comparing loss function curves for a 4-layer neural network model provided in accordance with some embodiments of the present invention during training and testing;
FIG. 5 is a profile comparison plot at a guide vane root section for a predictive geometry model and a true geometry model provided in accordance with some embodiments of the invention;
FIG. 6 is a profile comparison plot at a guide vane top section of a predictive geometry model and a truth geometry model provided in accordance with some embodiments of the present invention;
FIG. 7 is a blade profile comparison plot at a bucket root section for a predicted geometry model and a true geometry model provided in accordance with some embodiments of the invention;
FIG. 8 is a profile comparison plot of a section in the middle of a bucket for a predicted geometry model and a true geometry model provided in accordance with some embodiments of the invention;
FIG. 9 is a profile comparison plot at a bucket tip section of a predicted and true geometric model provided in accordance with some embodiments of the invention.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings. The embodiments described below with reference to the drawings are illustrative and intended to be illustrative of the invention and are not to be construed as limiting the invention.
As shown in FIG. 1, the turbine blade inverse problem design method based on machine learning according to the embodiment of the invention comprises the following steps:
and S1, establishing a data set from the geometric parameters of the blades to the aerodynamic parameters of the blades by taking the geometric parameters of the first blades as independent variables and the aerodynamic parameters of the first blades as dependent variables.
In some embodiments, the data set may be pre-constructed, or the first blade geometric parameter may be used as an input condition, and the target output of the first blade aerodynamic parameter is calculated by using the multi-stage S2 flow surface calculation program, so that the data set of the first blade geometric parameter and the first blade aerodynamic parameter may be obtained. 8 target outputs can be obtained through calculation of a multi-stage S2 flow surface calculation program, and the 8 target outputs, namely the first pneumatic parameters, are efficiency, power, flow, relative outlet Mach number, absolute outlet Mach number, relative outlet airflow angle, absolute outlet airflow angle and reaction degree respectively.
In some embodiments, as shown in fig. 2A, establishing a data set from the blade geometric parameters to the blade aerodynamic parameters with the first blade geometric parameters as independent variables and the first blade aerodynamic parameters as dependent variables includes:
and S11, sampling the value range by adopting a sampling algorithm based on the value range of the geometric parameters of the blade to obtain a blade shape file consisting of the first geometric parameters.
It should be noted that the blade profile file is a specific blade geometric parameter obtained by a sampling algorithm in a blade geometric value range, and 8 specific blade geometric parameters form one blade profile file, for example, as shown in table 1, in this embodiment, 5 sections of a blade are selected, the 5 sections are respectively a guide vane root, a guide vane top, a movable vane root, a movable vane middle part, and a movable vane top, and each section includes the 8 geometric parameters. Specific values of 8 geometric parameters of different sections can be uniformly sampled and taken in the value range of table 1 by using a latin hypercube sampling algorithm, and a plurality of leaf profile files of which 8 specific geometric parameters are one leaf profile file are obtained.
TABLE 1 value ranges of geometric parameters
And S12, calculating the leaf file to obtain the first aerodynamic parameter.
It is noted that the turbine blade input boundary conditions for the multi-stage S2 flow surface calculation procedure are shown in Table 2 during calculation using the multi-stage S2 flow surface calculation procedure.
TABLE 2S2 program input boundary conditions
| Boundary condition | Value taking | Unit of |
| Inlet total pressure | 1012340 | Pa |
| Total inlet temperature | 1330 | K |
| Static pressure at outlet | 300000 | Pa |
| Radius of the outlet | 295.75 | mm |
And S13, normalizing the first geometric parameters and the first aerodynamic parameters corresponding to the first geometric parameters to obtain a data set.
It should be noted that, in this embodiment, the normalization formula may be:
in the above formula, x**Parameters in the normalized data set; x is the number ofmaxThe upper limit of the value range of x; x is the number ofminThe lower limit of the value range of x; x is a first geometric parameter and a first aerodynamic parameter which need to be normalized.
S2, a training model of the turbine blade inverse problem design is created.
It should be noted that, in the present invention, the construction and use platform of the training model are not limited, for example, the training model may be constructed by a matlab platform, or may be constructed by other platforms.
It should be noted that, in the present invention, the algorithm of the training model is not limited, for example, the training model may be established based on a BP neural network algorithm, may also be established based on a BRF neural network algorithm, may also be established based on a BP neural network and a BRF neural network together, and may also be established based on a random forest algorithm.
In some embodiments, the training model may be constructed based on a BP neural network and a BRF neural network, specifically, as shown in fig. 3, the training model is constructed by a 4-layer neural network model, the 4-layer neural network sequentially includes an input layer, a first hidden layer, a second hidden layer, and an output layer from a data set input end to a calculation result output end, the input layer and the first hidden layer are RBF layers, activation functions of the input layer and the first hidden layer are both radial basis functions, the second hidden layer and the output layer are fully connected layers, the activation function of the second hidden layer includes a ReLU function, and the activation function of the output layer includes a sigmoid function.
The training model is trained and loss-assessed using the data set to determine an inverse problem design model S3.
It should be noted that, during the training process, there is no limitation on the type of the loss function and the optimization algorithm of the training model, for example, the loss function may be an average absolute error or a mean square error, and in some embodiments, the loss function is defined by the mean square error; the optimization algorithm may be a Nesterov gradient acceleration method, or may employ an Adam optimizer, which in some embodiments is employed.
In some embodiments, the weight calculation parameter beta of the input layer and the first hidden layer is set to 0.2, the number of neural units of the input layer is 400, the number of neural units of the first hidden layer is 400, the number of neural units of the second hidden layer is 256, the number of neural units of the output layer is 40, the training cycle number (epochs) of the training model is 50000, and the mini-batch size is 256. It will be understood that the above description is exemplary only, and in some embodiments is not limiting.
FIG. 2B is a schematic flow chart illustrating step S3 of a method for designing turbine blade inverse problems based on machine learning according to some embodiments of the present invention; as shown in fig. 2B, the training and loss evaluation of the training model using the data set to determine the inverse problem design model includes:
and S31, dividing the data set into training data and test data, wherein the test data comprise a third blade pneumatic parameter and a third geometric parameter corresponding to the third blade pneumatic parameter.
It should be noted that, the number and the proportion of the training data and the test data are not limited, and in some embodiments, 9781 groups of data are selected as a data set, wherein 8803 groups of data are used as training data for training the training model to have generalization capability, 978 groups of data are used as test data, and test data are used as data that are not seen by the training model for verifying the model performance.
And S32, training and loss evaluation are carried out on the 4-layer neural network model by using the training data to obtain an inverse problem design model.
The inverse problem design model is validated using the test data to determine an inverse problem design model S33.
FIG. 4 is a graph comparing loss function curves for a 4-layer neural network model during training and testing, provided in accordance with some embodiments of the present invention; as shown in fig. 4, the thick solid line and the thin solid line respectively represent the average Mean Square Error (MSE) of the training data and the test data, the final average MSE (MSE) of the training is 0.00038, and the average MSE (MSE) of the test is 0.00035, so that it can be seen that the overall prediction error of the model is very small, the loss of the test data is equal to the loss of the training data, no overfitting phenomenon occurs, and it is shown that the inverse problem design model has good prediction capability.
FIG. 2C is a schematic flow chart illustrating step S33 of a method for designing turbine blade inverse problems based on machine learning according to some embodiments of the present invention; as shown in fig. 2C, the verifying the inverse problem design model using the test data to determine the inverse problem design model includes:
and S331, substituting the pneumatic parameters of the third blade of the test data into the inverse problem design model to obtain the geometric parameters of the fourth blade.
And S332, performing inverse normalization processing on the geometric parameters of the fourth blade to obtain the geometric parameters of the fifth blade.
It should be noted that, in this embodiment, the inverse normalization formula is:
x＝x*×(xmax-xmin)+xmin
in the above formula, x is the fifth geometric parameter after inverse normalization, x*A fourth geometric parameter obtained based on the test data; x is the number ofmaxThe upper limit of the value range of x; x is the number ofminThe lower limit of the value range of x.
And S333, establishing a prediction geometric model based on the fifth blade geometric parameter.
And S334, performing inverse normalization processing on the geometric parameters of the third blade of the test data to obtain sixth geometric parameters.
And S335, establishing a true geometric model based on the sixth blade geometric parameter.
And S336, comparing and evaluating the leaf profiles of the prediction geometric model and the truth geometric model on the same leaf height to determine the inverse problem design model.
The leaf profile comparison conditions of the predicted geometric model and the truth geometric model on 5 designed sections are shown in fig. 5 to 9, the coincidence degree of the 5 sections of the predicted geometric model and the truth geometric model on the front edge and the tail edge, the pressure surface molded line and the suction surface molded line is very high, and it can be considered that the inverse problem design model realizes high-precision prediction on the blade geometry.
FIG. 2D is a schematic flow chart comparing second blade aerodynamic parameters and third blade aerodynamic parameters in a method for machine learning based turbine blade inverse problem design provided in accordance with some embodiments of the present invention; referring to fig. 2D, unlike fig. 2C, the method for designing an anti-problem of a turbine blade based on machine learning according to this embodiment further includes, after the step of substituting the second aerodynamic parameter of the test data into the anti-problem design model to obtain the third blade geometric parameter:
s337, calculating the geometric parameters of the sixth blade to obtain fourth pneumatic parameters;
it should be noted that the sixth blade geometric parameter may be substituted into the multi-stage S2 flow surface calculation program to obtain the fourth aerodynamic parameter.
And S338, calculating an average error based on the third pneumatic parameter and the fourth pneumatic parameter.
Table 3 is a comparison of the 6 sets of third vane aerodynamic parameters and the calculated fourth vane aerodynamic parameters to verify whether the fourth vane aerodynamic parameters and the third vane aerodynamic parameters are close, wherein the top row of the same example is the third aerodynamic parameter.
TABLE 3 calculation results of the third vane aerodynamic parameter and the fourth vane aerodynamic parameter
Calculating the average error of the aerodynamic parameters of the 6 groups of third blades and the calculated aerodynamic parameters of the fourth blade in the table 3, and obtaining the result: the average errors of efficiency, power, flow, relative outlet mach number, absolute outlet mach number, relative outlet flow angle, absolute outlet flow angle, and reaction are 0.15%, 0.57%, 0.82%, 0.67%, 1.81%, 0.32%, 1.37%, and 2.13%, respectively. Based on the above 6 examples of the calculation, the overall average error is 0.98%, it can be considered that the third blade aerodynamic parameter is very close to the fourth blade aerodynamic parameter, and since the fourth blade aerodynamic parameter is calculated based on the third blade geometric parameter and the fourth blade geometric parameter is predicted based on the anti-problem design model, the prediction of the anti-problem design model of the turbine blade anti-problem design method based on machine learning is more accurate.
And S4, substituting the second blade aerodynamic parameter into the inverse problem design model to reversely predict the second blade geometric parameter.
It should be noted that the second blade aerodynamic parameter is a blade aerodynamic parameter matched with the blade to be designed, and the second blade geometric parameter is a geometric parameter similar to the blade to be designed as a target output.
Particular embodiments of the computer-readable storage medium of the present invention have computer instructions stored thereon which, when executed by a processor, implement the machine learning-based turbine blade inverse problem design method.
In a specific embodiment of the electronic device of the present invention, the electronic device includes: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to: executing the turbine blade inverse problem design method based on machine learning.
In the description of the present invention, it is to be understood that the terms "first", "second" and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implying any number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless explicitly specifically defined otherwise.
In the description herein, references to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.
Claims (10)
1. A turbine blade anti-problem design method based on machine learning is characterized by comprising the following steps:
establishing a data set from the geometric parameters of the blades to the aerodynamic parameters of the blades by taking the geometric parameters of the first blades as independent variables and the aerodynamic parameters of the first blades as dependent variables;
creating a training model of the turbine blade inverse problem design;
training and loss evaluating the training model using the data set to determine the inverse problem design model;
substituting the second blade aerodynamic parameter into the inverse problem design model to reversely predict a second blade geometric parameter.
2. The machine learning-based turbine blade inverse problem design method according to claim 1, wherein the establishing of the data set from the blade geometric parameters to the blade aerodynamic parameters with the first blade geometric parameters as independent variables and the first blade aerodynamic parameters as dependent variables comprises:
sampling the value range by adopting a sampling algorithm based on the value range of the blade geometric parameters to obtain a blade profile file consisting of the first blade geometric parameters;
calculating a blade shape file to obtain the pneumatic parameters of the first blade;
and normalizing the first blade geometric parameters and the first blade aerodynamic parameters corresponding to the first blade geometric parameters to obtain the data set.
3. The machine learning based turbine blade inverse problem design method of claim 2, wherein said creating a training model of a turbine blade inverse problem design comprises:
creating a 4-layer neural network model consisting of an input layer, a first hidden layer, a second hidden layer and an output layer;
setting an input layer and a first hidden layer as RBF layers;
and setting the second hidden layer and the output layer as full connection layers.
4. The machine learning based turbine blade inverse problem design method of claim 3, said training and loss assessment of the training model using the data set to determine the inverse problem design model, comprising:
dividing the data set into training data and test data, wherein the test data comprises the third vane pneumatic parameters and third vane geometric parameters corresponding to the third vane pneumatic parameters;
training and loss evaluation are carried out on the 4-layer neural network model by utilizing the training data so as to obtain the anti-problem design model;
and verifying the anti-problem design model by using the test data to determine the anti-problem design model.
5. The machine learning based turbine blade inverse problem design method of claim 4, wherein the training and loss assessment of the 4-layer neural network model using the training data to obtain the inverse problem design model comprises:
carrying out loss evaluation on the training model by using a loss function;
performing iterative optimization on the loss function by using an optimization algorithm;
and evaluating the calculation result of the loss evaluation to obtain the inverse problem design model.
6. The machine learning based turbine blade inverse problem design method of claim 5, wherein said validating the inverse problem design model using the test data to determine the inverse problem design model comprises:
substituting the pneumatic parameters of the third blade of the test data into the inverse problem design model to obtain the geometric parameters of the fourth blade;
performing inverse normalization processing on the geometric parameters of the fourth blade to obtain the geometric parameters of the fifth blade;
establishing a prediction geometric model based on the fifth blade geometric parameter;
performing inverse normalization processing on the geometric parameters of the third blade of the test data to obtain sixth geometric parameters;
establishing a true geometric model based on the sixth blade geometric parameter;
and comparing and evaluating the leaf profiles of the prediction geometric model and the truth geometric model on the same leaf height to determine the inverse problem design model.
7. The machine learning-based turbine blade inverse problem design method of claim 6, wherein after the step of substituting a third aerodynamic parameter of test data into the inverse problem design model to obtain a sixth blade geometric parameter, further comprising:
calculating the geometric parameters of the sixth blade to obtain fourth pneumatic parameters;
based on the third and fourth aerodynamic parameters, an average error is calculated.
8. The machine learning based turbine blade inverse problem design method of claim 7, wherein the activation functions of the input layer and the first hidden layer are both radial basis functions; the activation function of the second hidden layer comprises a ReLU function; the activation function of the output layer comprises a sigmoid function.
9. A computer readable storage medium having computer instructions stored thereon, wherein the computer instructions, when executed by a processor, implement the method of any of claims 1 to 8.
10. An electronic device, comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor is configured to: performing the method of any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110790078.8A CN113705077A (en) | 2021-07-13 | 2021-07-13 | Turbine blade anti-problem design method based on machine learning, computer-readable storage medium and electronic device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110790078.8A CN113705077A (en) | 2021-07-13 | 2021-07-13 | Turbine blade anti-problem design method based on machine learning, computer-readable storage medium and electronic device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113705077A true CN113705077A (en) | 2021-11-26 |
Family
ID=78648518
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110790078.8A Pending CN113705077A (en) | 2021-07-13 | 2021-07-13 | Turbine blade anti-problem design method based on machine learning, computer-readable storage medium and electronic device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113705077A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115983137A (en) * | 2023-01-31 | 2023-04-18 | 西安交通大学 | Turbine flow field prediction method based on similarity principle and deep learning and related device |
| CN116976201A (en) * | 2023-07-10 | 2023-10-31 | 哈尔滨工业大学 | Self-programming parameterized modeling method and modeling system for micro turbine blade of breathing machine, computer readable storage medium and electronic equipment |
| CN117113552A (en) * | 2023-07-19 | 2023-11-24 | 哈尔滨工业大学 | Self-programming one-dimensional design method and design system for micro turbine blade of breathing machine |
| CN117332671A (en) * | 2023-07-06 | 2024-01-02 | 哈尔滨工业大学 | Respirator micro turbine design optimization method and optimization system based on machine learning, computer readable storage medium and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814272A (en) * | 2020-07-07 | 2020-10-23 | 中国科学院工程热物理研究所 | Turbine pneumatic-dynamic response intelligent optimization design method based on machine learning |
| CN111914461A (en) * | 2020-09-08 | 2020-11-10 | 北京航空航天大学 | Intelligent assessment method for one-dimensional cold efficiency of turbine guide vane |
| CN112380794A (en) * | 2020-12-08 | 2021-02-19 | 中北大学 | Multi-disciplinary parallel cooperation optimization design method for aviation turbine engine blade |
| CN112437928A (en) * | 2018-07-03 | 2021-03-02 | 西门子能源全球有限两合公司 | Design and manufacture of fluid machinery blade |
| CN112800663A (en) * | 2021-01-03 | 2021-05-14 | 复旦大学 | Inverse design method of aero-engine compressor rotor blade based on neural network |
-
2021
- 2021-07-13 CN CN202110790078.8A patent/CN113705077A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112437928A (en) * | 2018-07-03 | 2021-03-02 | 西门子能源全球有限两合公司 | Design and manufacture of fluid machinery blade |
| CN111814272A (en) * | 2020-07-07 | 2020-10-23 | 中国科学院工程热物理研究所 | Turbine pneumatic-dynamic response intelligent optimization design method based on machine learning |
| CN111914461A (en) * | 2020-09-08 | 2020-11-10 | 北京航空航天大学 | Intelligent assessment method for one-dimensional cold efficiency of turbine guide vane |
| CN112380794A (en) * | 2020-12-08 | 2021-02-19 | 中北大学 | Multi-disciplinary parallel cooperation optimization design method for aviation turbine engine blade |
| CN112800663A (en) * | 2021-01-03 | 2021-05-14 | 复旦大学 | Inverse design method of aero-engine compressor rotor blade based on neural network |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115983137A (en) * | 2023-01-31 | 2023-04-18 | 西安交通大学 | Turbine flow field prediction method based on similarity principle and deep learning and related device |
| CN115983137B (en) * | 2023-01-31 | 2024-05-31 | 西安交通大学 | Turbine flow field prediction method and related device based on similarity principle and deep learning |
| CN117332671A (en) * | 2023-07-06 | 2024-01-02 | 哈尔滨工业大学 | Respirator micro turbine design optimization method and optimization system based on machine learning, computer readable storage medium and electronic equipment |
| CN116976201A (en) * | 2023-07-10 | 2023-10-31 | 哈尔滨工业大学 | Self-programming parameterized modeling method and modeling system for micro turbine blade of breathing machine, computer readable storage medium and electronic equipment |
| CN117113552A (en) * | 2023-07-19 | 2023-11-24 | 哈尔滨工业大学 | Self-programming one-dimensional design method and design system for micro turbine blade of breathing machine |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113705077A (en) | Turbine blade anti-problem design method based on machine learning, computer-readable storage medium and electronic device | |
| CN110851929B (en) | Two-dimensional leaf-type optimization design method and device based on self-adaptive grid | |
| CN104778327B (en) | Aircraft wing Optimization Design based on artificial neural network | |
| CN114372318B (en) | Design parameter determining method, device, electronic equipment and readable storage medium | |
| Verstraete et al. | Design and optimization of the internal cooling channels of a high pressure turbine blade—Part II: Optimization | |
| Li et al. | Numerical optimization for stator vane settings of multi-stage compressors based on neural networks and genetic algorithms | |
| CN114297779B (en) | Rapid prediction method, system and equipment for aerodynamic coefficient of aircraft | |
| CN114154275B (en) | Low-pressure turbine blade profile pneumatic design method based on optimal load distribution model optimization | |
| CN112182954B (en) | LSTM-based fluid simulation data prediction model | |
| CN115983137A (en) | Turbine flow field prediction method based on similarity principle and deep learning and related device | |
| Biswas et al. | Application of artificial neural network for performance evaluation of vertical axis wind turbine rotor | |
| Cheng et al. | Robust optimization and uncertainty quantification of a micro axial compressor for unmanned aerial vehicles | |
| Zhang et al. | Optimized design of wind turbine airfoil aerodynamic performance and structural strength based on surrogate model | |
| CN112270139A (en) | Pneumatic optimization design method of fuel cell centrifugal compressor based on master model library | |
| CN114021492B (en) | Supercritical airfoil buffeting optimization method based on neural network | |
| CN115481461A (en) | Structural wind pressure statistic prediction method based on deep learning | |
| Yao et al. | Prediction of aerodynamic forces at the tip of the compressor blades based on multi-scale 1DCNN combined with CBAM | |
| Zhang et al. | A parameterized-loading driven inverse design and multi-objective coupling optimization method for turbine blade based on deep learning | |
| CN114861315A (en) | Two-dimensional impeller profile optimization method based on machine learning | |
| CN113868765A (en) | Ship main scale parameter optimization method based on approximate model | |
| CN116595874A (en) | Impeller mechanical performance prediction model parameter optimization method and device and storage medium | |
| Guo et al. | An aero-structure-acoustics evaluation framework of wind turbine blade cross-section based on Gradient Boosting regression tree | |
| Angelini et al. | A multidimensional extension of Balje chart for axial flow turbomachinery using artificial intelligence-based meta-models | |
| CN114046959B (en) | Five-hole pneumatic probe calibration method based on two-stage artificial neural network | |
| CN116110630A (en) | Reactor core thermal power measurement method and device in DCS system |
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 |