Construction method, prediction method and equipment for full-flow-channel flow field prediction model of turbomachinery

Abstract

The invention discloses a construction method, a prediction method and equipment of a full-flow-channel flow field prediction model of turbomachinery, and belongs to the technical field of turbomachinery. According to the method, a fluid domain of the turbomachinery is decomposed into different types of physical domains, the different types of physical domains are respectively projected into corresponding rule calculation domains, corresponding agent models are set for training, agent models of the different types of physical domains are obtained, and then a turbomachinery full-flow-channel flow field prediction model is formed; according to the method, the difference between different types of physical domains of the turbomachinery is considered, the corresponding proxy models are trained for the different types of physical domains respectively, and the continuous condition of the physical quantity of each region interface is utilized in the prediction process, so that the prediction of the full-flow-channel flow field is realized by adopting iterative calculation; and the full-flow-channel flow field of the turbomachinery can be efficiently and accurately predicted.

Description

Technical Field

[0001] The present invention belongs to the technical field of impeller machinery, and more specifically, relates to a method for constructing a full-flow channel flow field prediction model of an impeller machinery, a prediction method and a device. Background Art

[0002] As one of the most important energy conversion devices, turbomachinery is widely used in national economy and national defense and military fields such as energy, construction, environment, aviation, and navigation. The rapid and accurate prediction of its flow field information is of great significance for the optimal design of high-performance turbomachinery.

[0003] Existing full-channel flow field prediction methods for impeller machinery often obtain the internal flow field information of the impeller machinery by solving the Navier-Stokes equations. This method has strong applicability and high prediction accuracy and can be applied to complex multi-stage, full-channel impeller machinery. However, it usually requires tedious spatial discretization processing first, and its calculation is time-consuming, making it impossible to efficiently and accurately predict the full-channel flow field of the impeller machinery. Summary of the invention

[0004] In view of the above defects or improvement needs of the prior art, the present invention provides a method for constructing a full-channel flow field prediction model for an impeller machinery, a prediction method and an apparatus, so as to solve the technical problem that the prior art cannot efficiently and accurately predict the full-channel flow field of an impeller machinery.

[0005] In order to achieve the above-mentioned object, in a first aspect, the present invention provides a method for constructing a full-flow channel flow field prediction model of an impeller machinery, comprising:

[0006] Taking the plane where the leading edge point of each blade in each blade row and the plane where the trailing edge point of each blade is located as the dividing plane, the fluid domain of the impeller machinery is divided into a non-blade annular domain and a blade row domain, and further taking the suction surface and pressure surface of each blade as the dividing plane, the blade row domain is divided into multiple blade single flow channel domains, thereby dividing the impeller machinery into different types of physical domains; the blade single flow channel domain includes two types: a stationary guide vane single flow channel domain and a rotating blade single flow channel domain;

[0007] Each type of physical domain is projected into a regular computational domain through coordinate transformation, and a corresponding proxy model is set. The corresponding proxy model is trained using a corresponding training set to obtain proxy models of various types of physical domains, thereby forming a full-channel flow field prediction model for turbomachinery.

[0008] The training set corresponding to the physical domain includes: training samples and corresponding labels;

[0009] The training samples include: velocity field distribution on the left boundary and pressure field distribution on the right boundary of the computational domain corresponding to the physical domain, and coordinate transformation parameters between the physical domain and the corresponding computational domain; the left and right boundaries of the computational domain correspond to the inlet and outlet surfaces of the fluid on the physical domain, respectively;

[0010] The label includes: real flow field information on the computational domain corresponding to the physical domain;

[0011] The flow field information includes: velocity field distribution and pressure field distribution.

[0012] Further preferably, the process of coordinate transformation and projection of the physical domain into the regular computational domain comprises:

[0013] The physical domain is meshed and converted into a regular computational domain through coordinate transformation; each grid point in the computational domain corresponds one-to-one to each grid point in the corresponding physical domain.

[0014] Further preferably, for any grid point coordinate (x, y) in the physical domain, the coordinate of the corresponding grid point in the corresponding computational domain is labeled (ξ, η); wherein ξ = ξ(x, y), η = η(x, y);

[0015] The coordinate transformation parameters between the physical domain and the corresponding computational domain include:

[0016] Further preferably, the calculation domain is a rectangular calculation domain.

[0017] In a second aspect, the present invention provides a method for predicting the flow field of the entire flow passage of a turbomachinery, comprising: predicting the flow field information of the entire flow passage of the turbomachinery based on a prediction model of the flow field of the entire flow passage of the turbomachinery;

[0018] The full-flow channel flow field prediction model of the impeller machinery is constructed by using the construction method provided by the first aspect of the present invention; the flow field information includes: velocity field distribution and pressure field distribution;

[0019] The impeller machine includes: M+1 bladeless annular domains and a blade row domain located between two adjacent bladeless annular domains; M≥1; the mth blade row domain D L,m With the mth bladeless annular region D B,m and the m+1th cascade-free annular region D B,m+1 The interface is denoted as interface Ι m and interface ΙΙ m ; m = 1, 2, ..., M; the number of blade single flow channel domains on the blade row domain is recorded as N, and the nth blade single flow channel domain on the mth blade row domain is recorded as D L,m,n ;

[0020] The prediction of the flow field information of the entire flow channel of the impeller machinery includes:

[0021] Project each cascade-free annular domain and each blade single-channel domain into a regular computational domain through coordinate transformation, and obtain corresponding coordinate transformation parameters;

[0022] For the interface in the computational domain m and interface ΙΙ m The velocity field distribution and pressure field distribution on the surface are predicted to obtain the interface Ι in the calculation domain m and interface ΙΙ m Velocity field distribution And the prediction results of pressure field distribution

[0023] Will and D B,m The corresponding coordinate transformation parameters are input into the non-cascade annular domain proxy model in the full-channel flow field prediction model of the turbomachinery to obtain D B,m The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D B,m Flow field information on S is the velocity field distribution at the left boundary of the calculation domain corresponding to the first pre-collected cascade annular domain;

[0024] Will and D L,m,n The corresponding coordinate transformation parameters are input into the blade single flow channel domain proxy model in the full flow channel flow field prediction model of the impeller machinery to obtain D L,m,n The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D L,m,n Flow field information on for Middle D L,m,n The corresponding prediction results of the velocity field distribution at the left boundary of the computational domain; for Middle D L,m,n The corresponding prediction results of the pressure field distribution at the right boundary of the computational domain;

[0025] Will and D B,m+1 The corresponding coordinate transformation parameters are input into the proxy model of the annular domain without cascades to obtain D B,m+1 The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D B,m+1 Flow field information on P is the pressure field distribution at the right boundary of the calculation domain corresponding to the pre-collected M+1th cascade annular domain.

[0026] Further preferably, the interface Ι in the calculation domain m and interface ΙΙ m The pressure field distribution on the surface is predicted, including:

[0027] A1. Initialize the interface in the computational domain I m and interface II m The pressure field distribution on and and D L,m,n The initial value of the pressure field distribution at the left and right boundaries of the corresponding calculation domain;

[0028] A2. and D B,m The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained, so as to obtain the current interface I in the calculation domain m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain;

[0029] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding velocity field distribution at the right boundary of the computational domain Then we can get the current interface II in the computational domain m The velocity field distribution on

[0030] Will and D B,m+1 The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m+1 The flow field information on the corresponding calculation domain is obtained, thereby obtaining the current interface II in the calculation domain m The pressure field distribution on For the current D L,m,n The corresponding pressure field distribution at the right boundary of the computational domain;

[0031] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain Then we can get the current interface I in the computational domain m The pressure field distribution on

[0032] A3. Judgment and and Are the absolute values ​​of the differences between the two values ​​less than the corresponding preset thresholds? If so, go to A4; otherwise, Go to A2; α1 is the preset adjustment coefficient;

[0033] A4. Set the current interface I in the calculation domain m and interface II m The pressure field distribution on and As the interface in the computational domain m and interface ΙΙ m Prediction results of upper pressure field distribution

[0034] Among them, m=1,2,…,M.

[0035] Further preferably, the interface Ι in the calculation domain m and interface ΙΙ m The velocity field distribution on the surface is predicted, including:

[0036] B1. Initialize the interface in the computational domain I m and interface II m The velocity field distribution on and and D L,m,n The initial value of the velocity field distribution at the left and right boundaries of the corresponding calculation domain;

[0037] B2. and D B,m+1 The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m+1 The flow field information on the corresponding calculation domain is obtained, thereby obtaining the current interface II in the calculation domain m The pressure field distribution on For the current D L,m,n The corresponding pressure field distribution at the right boundary of the calculation domain;

[0038]

[0039] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain Then we get the current interface I m The pressure field distribution on

[0040] Will and D B,m The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained, so as to obtain the current interface I in the calculation domain m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain;

[0041] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding velocity field distribution on the right boundary of the computational domain Then we can get the current interface II in the computational domain m The velocity field distribution on

[0042] B3. Judgment and and Are the absolute values ​​of the differences of are all less than the corresponding preset thresholds? If so, go to B4; otherwise, let Go to B2; α2 is the preset adjustment coefficient;

[0043] B4. Set the current interface I in the calculation domain m and interface II m The velocity field distribution on and As the interface in the computational domain I m and interface II m Prediction results of velocity field distribution on

[0044] Among them, m=M,M-1,…,1.

[0045] Further preferably, the interface Ι in the calculation domain m and interface ΙΙ m The velocity field distribution and pressure field distribution on the surface are predicted, including:

[0046] C1. Initialize the interface in the computational domain Ι m and interface ΙΙ m The velocity field distribution on and And the pressure field distribution and and D L,m,n The initial values ​​of the velocity field distribution at the left and right boundaries of the corresponding computational domain; and D L,m,n The initial values ​​of the pressure field distribution at the left and right boundaries of the corresponding calculation domain;

[0047] C2. and D B,m The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained to obtain the current interface in the calculation domain. m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain;

[0048] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain And the velocity field distribution on the right boundary Then we can get the current interface in the computational domain. m The pressure field distribution on and the current interface ΙΙ m The velocity field distribution on

[0049] Will and D B,m+1 The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m+1 The flow field information on the corresponding calculation domain is obtained to obtain the current interface ΙΙ in the calculation domain m The pressure field distribution on

[0050] C3. Judgment and and and and Are the absolute values ​​of the differences less than the corresponding preset thresholds? If so, go to C4; otherwise, Go to C2; α is the preset adjustment coefficient;

[0051] C4, the current interface Ι m and interface ΙΙ m The velocity field distribution on And pressure field distribution As the interface in the computational domain m and interface ΙΙ m Velocity field distribution And the prediction results of pressure field distribution

[0052] Among them, m=1,2,…,M.

[0053] In a fourth aspect, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor executes the method provided in the first aspect or the second aspect of the present invention when executing the computer program.

[0054] In a fifth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is executed by a processor, the device where the storage medium is located is controlled to execute the method provided in the first aspect or the second aspect of the present invention.

[0055] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects:

[0056] 1. The present invention provides a method for constructing a full-channel flow field prediction model for an impeller machinery, which decomposes the fluid domain of an impeller machinery into different types of physical domains, and projects different types of physical domains into corresponding regular calculation domains respectively, and sets corresponding proxy models for training to obtain proxy models of different types of physical domains, thereby forming a full-channel flow field prediction model for an impeller machinery. The present invention takes into account the differences between different types of physical domains of an impeller machinery, and trains corresponding proxy models for different types of physical domains respectively, so as to efficiently and accurately predict the full-channel flow field of an impeller machinery, and even the full-channel flow field of an impeller machinery with a complex structure can be efficiently and accurately predicted.

[0057] 2. The present invention provides a method for predicting the full-channel flow field of an impeller machinery. The full-channel flow field prediction of the impeller machinery is performed based on the full-channel flow field prediction model of the impeller machinery constructed by the construction method provided by the first aspect of the present invention, and the continuity conditions of the physical quantities of the interfaces in each region are utilized. The full-channel flow field prediction is realized by iterative calculation, and the full-channel flow field of the impeller machinery can be predicted efficiently and accurately.

[0058] 3. Furthermore, the method for predicting the flow field of the entire flow channel of an impeller machine provided by the present invention updates the velocity and pressure information at the interface in steps through a serial prediction method when predicting the velocity field distribution and pressure field distribution on the interface, with small memory usage and strong convergence.

[0059] 4. Furthermore, the present invention provides a method for predicting the flow field of the entire flow channel of a single-stage or multi-stage impeller machinery. When predicting the velocity field distribution and pressure field distribution on the interface, the velocity and pressure information at the interface are simultaneously updated through a parallel prediction method, and the convergence speed is fast. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A flow chart of an agent model for efficiently predicting the flow field of a turbomachinery provided by an embodiment of the present invention;

[0061] Figure 2 The impeller internal flow domain segmentation diagram provided by the embodiment of the present invention;

[0062] Figure 3 A schematic diagram of the physical domain boundary node distribution and internal grid filling process provided by an embodiment of the present invention;

[0063] Figure 4 A schematic diagram of mapping an irregular physical domain to a regular computing domain provided by an embodiment of the present invention;

[0064] Figure 5 A metric parameter matrix for characterizing boundary geometry changes provided by an embodiment of the present invention;

[0065] Figure 6 A cloud diagram of the X-direction velocity information U inside the impeller provided in an embodiment of the present invention;

[0066] Figure 7 A schematic diagram of mapping flow field information of a rotating single flow channel physical domain to a computational domain provided by an embodiment of the present invention;

[0067] Figure 8 A set of input and output parameter matrices of a rotating single flow channel provided in an embodiment of the present invention;

[0068] Fig. 9 A schematic diagram of the rotating single-channel U-net convolutional neural network architecture provided in an embodiment of the present invention;

[0069] Fig.10 A schematic diagram of the prediction effect of a test set sample of a rotating single flow channel provided by an embodiment of the present invention;

[0070] Fig.11 A schematic diagram of the prediction effect of a static single flow channel test set sample provided by an embodiment of the present invention;

[0071] Fig.12 A schematic diagram of the prediction effect of a test set sample in a non-cascade annular region provided by an embodiment of the present invention;

[0072] Fig.13 A flow chart for predicting full-channel impeller flow field information provided by an embodiment of the present invention;

[0073] Fig.14 A flow chart of a serial prediction method for correcting pressure information at each full channel interface provided by an embodiment of the present invention;

[0074] Fig.15 A flow chart of a serial prediction method for correcting velocity information at each full channel interface provided by an embodiment of the present invention;

[0075] Fig.16 A parallel prediction method for correcting flow field information at each full channel interface provided by an embodiment of the present invention;

[0076] Fig.17 A flow field convergence curve diagram provided by an embodiment of the present invention;

[0077] Fig.18 A schematic diagram of the iterative process of flow field information P, U, V at a certain monitoring point provided by an embodiment of the present invention;

[0078] Fig.19 A cloud diagram of the absolute value of the predicted value U of the multi-stage full-flow channel velocity information and the actual value and the difference between the two provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0079] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0080] In order to achieve the above-mentioned object, in a first aspect, the present invention provides a method for constructing a full-flow channel flow field prediction model of an impeller machinery, comprising:

[0081] Taking the plane where the leading edge point of each blade in each blade row and the plane where the trailing edge point of each blade is located as the dividing plane, the fluid domain of the impeller machinery is divided into a non-blade annular domain and a blade row domain, and further taking the suction surface and pressure surface of each blade as the dividing plane, the blade row domain is divided into multiple blade single flow channel domains, thereby dividing the impeller machinery into different types of physical domains; the blade single flow channel domain includes two types: a stationary guide vane single flow channel domain and a rotating blade single flow channel domain;

[0082] Each type of physical domain is projected into a regular computational domain through coordinate transformation, and a corresponding proxy model is set. The corresponding proxy model is trained using a corresponding training set to obtain proxy models of various types of physical domains, thereby forming a full-channel flow field prediction model for turbomachinery.

[0083] The training set corresponding to the physical domain includes: training samples and corresponding labels;

[0084] The training samples include: velocity field distribution on the left boundary and pressure field distribution on the right boundary of the computational domain corresponding to the physical domain, and coordinate transformation parameters between the physical domain and the corresponding computational domain; the left and right boundaries of the computational domain correspond to the inlet and outlet surfaces of the fluid on the physical domain, respectively;

[0085] The label includes: real flow field information on the computational domain corresponding to the physical domain;

[0086] The flow field information includes: velocity field distribution and pressure field distribution.

[0087] It should be noted that the above-mentioned proxy model can be any one of a convolutional neural network, a fully connected layer neural network, a graph neural network, etc.

[0088] In an optional implementation, the process of coordinate transformation and projection of the physical domain into the regular computational domain includes:

[0089] The physical domain is meshed and converted into a regular computational domain through coordinate transformation; each grid point in the computational domain corresponds one-to-one to each grid point in the corresponding physical domain.

[0090] In an optional implementation, for any grid point coordinate (x, y) in the physical domain, the coordinate of the corresponding grid point in the corresponding computational domain is labeled (ξ, η); wherein ξ = ξ(x, y), η = η(x, y);

[0091] The coordinate transformation parameters between the physical domain and the corresponding computational domain include:

[0092] It should be noted that the calculation domain only needs to satisfy the shape rule, and can be a rectangular calculation domain or a square calculation domain. Preferably, in an optional implementation manner, the calculation domain is a rectangular calculation domain.

[0093] In a second aspect, the present invention provides a method for predicting the flow field of the entire flow passage of a turbomachinery, comprising: predicting the flow field information of the entire flow passage of the turbomachinery based on a prediction model of the flow field of the entire flow passage of the turbomachinery;

[0094] The full-channel flow field prediction model of the impeller machinery is constructed by using the construction method provided in the first aspect; the flow field information includes: velocity field distribution and pressure field distribution; the related technical scheme is the same as the construction method provided in the first aspect of the present invention, and will not be described in detail here;

[0095] The impeller machine includes: M+1 bladeless annular domains and a blade row domain located between two adjacent bladeless annular domains; M≥1; the mth blade row domain D L,m With the mth bladeless annular region D B,m and the m+1th cascade-free annular region DB,m+1 The interface is denoted as interface Ι m and interface ΙΙ m ; m = 1, 2, ..., M; the number of blade single flow channel domains on the blade row domain is recorded as N, and the nth blade single flow channel domain on the mth blade row domain is recorded as D L,m,n ;

[0096] The prediction of the flow field information of the entire flow channel of the impeller machinery includes:

[0097] Project each cascade-free annular domain and each blade single-channel domain into a regular computational domain through coordinate transformation, and obtain corresponding coordinate transformation parameters;

[0098] For the interface in the computational domain m and interface ΙΙ m The velocity field distribution and pressure field distribution on the surface are predicted to obtain the interface Ι in the calculation domain m and interface ΙΙ m Velocity field distribution And the prediction results of pressure field distribution

[0099] Will and D B,m The corresponding coordinate transformation parameters are input into the non-cascade annular domain proxy model in the full-channel flow field prediction model of the turbomachinery to obtain D B,m The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D B,m Flow field information on Among them, S is the velocity field distribution at the left boundary of the calculation domain corresponding to the first cascade annular domain pre-collected;

[0100] Will and D L,m,n The corresponding coordinate transformation parameters are input into the blade single flow channel domain proxy model in the full flow channel flow field prediction model of the impeller machinery to obtain D L,m,n The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D L,m,n Flow field information on for Middle D L,m,n The corresponding prediction results of the velocity field distribution at the left boundary of the computational domain; for Middle D L,m,n The corresponding prediction results of the pressure field distribution at the right boundary of the computational domain;

[0101] Will and D B,m+1 The corresponding coordinate transformation parameters are input into the proxy model of the annular domain without cascades to obtain D B,m+1The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D B,m+1 Flow field information on Among them, P is the pressure field distribution at the right boundary of the calculation domain corresponding to the pre-collected M+1th cascade annular domain.

[0102] It should be noted that, in the above process, the value of m gradually increases, from 1 to M in sequence.

[0103] In an optional implementation, a serial prediction method is used to calculate the interface I in the domain. m and interface ΙΙ m The pressure field distribution on the surface is predicted, including:

[0104] A1. Initialize the interface in the computational domain I m and interface II m The pressure field distribution on and and D L,m,n The initial value of the pressure field distribution at the left and right boundaries of the corresponding calculation domain;

[0105] A2. and D B,m The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained, so as to obtain the current interface I in the calculation domain m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain;

[0106] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding velocity field distribution at the right boundary of the computational domain Then we can get the current interface II in the computational domain m The velocity field distribution on

[0107] Will and D B,m+1 The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m+1 The flow field information on the corresponding calculation domain is obtained, thereby obtaining the current interface II in the calculation domain m The pressure field distribution on For the current D L,m,nThe corresponding pressure field distribution at the right boundary of the computational domain;

[0108] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain Then we can get the current interface I in the computational domain m The pressure field distribution on

[0109] A3. Judgment and and Are the absolute values ​​of the differences between the two values ​​less than the corresponding preset thresholds? If so, go to A4; otherwise, Go to A2; α1 is the preset adjustment coefficient;

[0110] A4. Set the current interface I in the calculation domain m and interface II m The pressure field distribution on and As the interface in the computational domain m and interface ΙΙ m Prediction results of upper pressure field distribution

[0111] Among them, m=1,2,…,M.

[0112] It should be noted that, in the above process, the value of m gradually increases, from 1 to M in sequence.

[0113] In an optional implementation, a serial prediction method is used to calculate the interface I in the domain. m and interface ΙΙ m The velocity field distribution on the surface is predicted, including:

[0114] B1. Initialize the interface in the computational domain I m and interface II m The velocity field distribution on and and D L,m,n The initial value of the velocity field distribution at the left and right boundaries of the corresponding calculation domain;

[0115] B2. and D B,m+1 The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m+1The flow field information on the corresponding calculation domain is obtained, thereby obtaining the current interface II in the calculation domain m The pressure field distribution on For the current D L,m,n The corresponding pressure field distribution at the right boundary of the calculation domain;

[0116]

[0117] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain Then we get the current interface I m The pressure field distribution on

[0118] Will and D B,m The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained, so as to obtain the current interface I in the calculation domain m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain;

[0119] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding velocity field distribution on the right boundary of the computational domain Then we can get the current interface II in the computational domain m The velocity field distribution on

[0120] B3. Judgment and and Are the absolute values ​​of the differences of are all less than the corresponding preset thresholds? If so, go to B4; otherwise, let Go to B2; α2 is the preset adjustment coefficient;

[0121] B4. Set the current interface I in the calculation domain m and interface II m The velocity field distribution on and As the interface in the computational domain Im and interface II m Prediction results of velocity field distribution on

[0122] Among them, m=M,M-1,…,1.

[0123] It should be noted that, in the above process, the value of m gradually decreases, from M to 1 in sequence.

[0124] In an optional implementation, a parallel prediction method is used to calculate the interface I in the domain. m and interface ΙΙ m The velocity field distribution and pressure field distribution on the surface are predicted, including:

[0125] C1. Initialize the interface in the computational domain Ι m and interface ΙΙ m The velocity field distribution on and And the pressure field distribution and and D L,m,n The initial values ​​of the velocity field distribution at the left and right boundaries of the corresponding computational domain; and D L,m,n The initial values ​​of the pressure field distribution at the left and right boundaries of the corresponding calculation domain;

[0126] C2. and D B,m The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained to obtain the current interface in the calculation domain. m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain;

[0127] Will and D L,m,n The corresponding coordinate transformation parameters are input into the single-channel proxy model of the blade to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain And the velocity field distribution on the right boundary Then we can get the current interface in the computational domain. m The pressure field distribution on and the current interface ΙΙ m The velocity field distribution on

[0128] Will and D B,m+1 The corresponding coordinate transformation parameters are input into the no-cascade annular domain proxy model to obtain the current D B,m+1 The flow field information on the corresponding calculation domain is obtained to obtain the current interface ΙΙ in the calculation domain m The pressure field distribution on

[0129] C3. Judgment and and and and Are the absolute values ​​of the differences less than the corresponding preset thresholds? If so, go to C4; otherwise, Go to C2; α is the preset adjustment coefficient;

[0130] C4, the current interface Ι m and interface ΙΙ m The velocity field distribution on And pressure field distribution As the interface in the computational domain m and interface ΙΙ m Velocity field distribution And the prediction results of pressure field distribution

[0131] Among them, m=1,2,…,M.

[0132] It should be noted that, in the above process, the value of m gradually decreases, from M to 1 in sequence.

[0133] In an optional implementation, the process of projecting any physical domain in the cascade-free annular domain and the single flow channel domain of each blade into a regular computational domain through coordinate transformation includes:

[0134] The physical domain is meshed and converted into a regular computational domain through coordinate transformation; each grid point in the computational domain corresponds one-to-one to each grid point in the corresponding physical domain.

[0135] In an optional implementation, for any grid point coordinate (x, y) in the physical domain, the coordinate of the corresponding grid point in the corresponding computational domain is labeled (ξ, η); wherein ξ = ξ(x, y), η = η(x, y);

[0136] The coordinate transformation parameters between the physical domain and the corresponding computational domain include:

[0137] It should be noted that the calculation domain only needs to satisfy the shape rule, and may be a rectangular calculation domain, a square calculation domain, etc. Preferably, in an optional implementation manner, the calculation domain is a rectangular calculation domain.

[0138] In order to further illustrate the construction method of the full-flow channel flow field prediction model of the impeller machinery and the corresponding full-flow channel flow field prediction method of the impeller machinery provided by the present invention, the following is a detailed description in conjunction with specific embodiments:

[0139] In this embodiment, the selected three-stage plane blade grid has 6 blade rows, and the number of blades are 10, 9, 10, 9, 10, and 9 respectively. Each stage is in the form of a front stator + a rear rotor, and the velocity inlet and pressure outlet boundary conditions are used in the flow field calculation.

[0140] Figure 1 The flowchart of the full-flow channel flow field prediction model of an impeller machine provided by the present invention is shown, and the specific example steps are as follows:

[0141] S1: Divide the internal flow domain of the impeller: divide the internal flow domain of the impeller machinery into three types of physical domains: the annular domain without cascades, the single flow channel domain of stationary guide vanes, and the single flow channel domain of rotating blades;

[0142] The specific process is: Figure 2 The internal flow domain of the impeller is segmented in the manner shown, where 1 is the annular region without cascades, 2 is the single flow channel region of stationary guide vanes, and 3 is the single flow channel region of rotating blades. In actual operation, multiple physical domains of annular region without cascades, single flow channel of stationary guide vanes, and single flow channel of rotating blades can be segmented and formed. The same type of physical domains of different blade rows can be classified into the same sample library. For example, this example can be divided into 7 physical domains of annular region without cascades, 30 (10+10+10) physical domains of single flow channel of stationary guide vanes, and 27 (9+9+9) physical domains of single flow channel of rotating blades.

[0143] In this embodiment, the inner area of ​​the impeller is divided into multiple bladeless annular domains, stationary guide vane single flow channel domains and rotating blade single flow channel domains through the connecting line of the leading edge points of adjacent blades in each blade row, the connecting line of the trailing edge points, the suction surface and the pressure surface of the blades.

[0144] S2: Mapping irregular physical domains to regular computational domains;

[0145] The obtained bladeless annular, stationary guide vane single flow channel and rotating blade single flow channel domains are often irregular in geometry. Therefore, these irregular physical domains need to be mapped into regular computational domains suitable for agent model training. The regular computational domain used in this embodiment is a rectangular boundary, and its internal grid nodes are evenly distributed.

[0146] The steps for mapping irregular physical domains of non-cascade annular, stationary guide vane single flow channel and rotating blade single flow channel to regular computational domains are the same. Taking a rotating blade single flow channel physical domain as an example, the specific process is:

[0147] ① First set the physical domain boundary grid node distribution, and then use the transfinite interpolation method to generate the internal grid node coordinates of the region according to the boundary grid node coordinates, such as Figure 3 As shown, the number of grid nodes at the boundary of the rotating single-channel physical domain of this embodiment is set to 128. After the internal grid is filled, there are 16384=128×128 grid nodes in total. Of course, in actual operation, the number of grid nodes is not limited to this number, depending on the situation.

[0148] Specifically, due to the large velocity gradient near the blade wall, when setting the grid node distribution on the boundary of each physical domain, the grid distribution can be denser near the blade wall and sparser away from the wall, so as to facilitate the subsequent proxy model to accurately predict the flow field information near the wall. The number of grid nodes on each boundary of the same type of physical domain remains consistent, so as to ensure that only one proxy model needs to be established for the same type of physical domain to predict the flow field.

[0149] ② Map the irregular physical domain (x, y) to the regular computational domain (ξ, η) through coordinate transformation, such as Figure 4 As shown, the regular computational domain grid corresponds one-to-one with the irregular physical domain grid nodes.

[0150] ③ The finite difference method is used to calculate the partial derivatives from (x, y) to (ξ, η) in the computational domain to obtain the metric parameters for measuring the geometric boundary changes of the blades on all grid nodes, and finally 10 128×128 parameter matrices are obtained. like Figure 5 shown.

[0151] Specifically, in the coordinate system of the physical domain (x, y), different geometric figures are represented by different geometric boundaries, but their fluid control equations remain unchanged. In the coordinate system of the computational domain (ξ, η), different geometric figures are represented by the same geometric boundaries, but their fluid control equations are different. The partial derivatives of the physical variables in the control equation with respect to x and y in the physical domain will be converted into partial derivatives with respect to ξ and η in the computational domain. For example, the first-order partial derivative term is expressed as The expressions of the second-order partial derivatives are not listed. For detailed derivation process, please refer to the relevant books on computational fluid dynamics. The above process will generate 10 new parameter matrices It characterizes the changes in the fluid control equations caused by different geometric figures in its computational domain, so these 10 parameter matrices can be used as metric parameters for boundary geometry changes. The finite difference method is used to calculate the partial derivatives from (x, y) to (ξ, η) in the computational domain, and the values ​​of the 10 metric parameters that characterize the changes in the control equations on all grid nodes can be obtained, and then combined to form 10 parameter matrices that measure the geometry changes of the blade boundaries.

[0152] The present invention adopts the coordinate transformation method to map the irregular physical domain to the regular computational domain, so as to use the proxy model (such as the classic CNN network) on the regular computational domain to predict the flow field. The predicted results are then inversely mapped to the irregular physical domain, thereby improving the prediction accuracy of the wall flow field information.

[0153] S3: Obtain flow field information on the computational domain;

[0154] The specific process is:

[0155] ① Spatially discretize the internal flow domain of the impeller machinery and divide the entire flow channel into grids.

[0156] Specifically, the internal flow domain of the impeller machinery is spatially discretized and the entire flow channel is divided into grids. The grids must meet the turbulence model y+ requirements, and then subsequent numerical calculations are performed to output the pressure and velocity information of the entire flow field.

[0157] ② Use commercial computational fluid dynamics software to perform numerical calculations, output the pressure information P and velocity information S of the entire flow field (specifically including: x-direction velocity U and y-direction velocity V information, x-direction is the fluid flow direction, y-direction is perpendicular to x-direction), and obtain flow field information; Figure 6 It is the information cloud diagram of the internal speed U of the impeller.

[0158] ③ Through interpolation processing, the pressure and velocity information of each grid node in each physical domain is obtained and mapped to the corresponding grid nodes in the computational domain. Taking a single flow channel of a rotating blade as an example, the pressure P and velocity U, V of the flow field in the physical domain are mapped to the computational domain, such as Figure 7 shown.

[0159] It should be noted that the above S1-S3 process is to pave the way for S4, and is intended to generate input and output samples for training the proxy model in S4. In this embodiment, the proxy model is a convolutional neural network.

[0160] It should be noted that the flow field information includes velocity information and pressure information, but is not limited to these two physical quantities.

[0161] S4: Obtain training samples and train the proxy model;

[0162] The specific process is:

[0163] ① This embodiment only focuses on the change of impeller shape, so the input parameters do not include operating parameters such as Mach number and speed. Therefore, the 10 parameter measurement matrices of the measurement boundary geometry change obtained by S2 and the speed information U on the left boundary of the calculation domain obtained by S3 are BC 、V BC and the pressure information P on the right boundary BC As input parameters, the calculation domain pressure P and velocity U, V information obtained by S3 are used as output information, that is, like Figure 8 The figure shows the input and output parameter matrix set of a rotating blade single flow channel. The first 13 matrix images are input parameters, and the last 3 matrix images are output parameters. The result is a sample of the rotating blade single flow channel proxy model training. The three-level cascade model of this embodiment can obtain a total of 7 training samples of non-cascade annular area proxy models, 27 rotating blade single flow channel proxy models, and 30 stationary guide vane single flow channel proxy models. To obtain more samples, the blade geometry can be changed to increase training cases and thus obtain more training samples. It should be noted that the acquisition of the training samples of the other two areas, the non-cascade annular area and the stationary guide vane single flow channel, is also processed by sampling a similar method.

[0164] ②Use the U-net convolutional neural network architecture to train the proxy models of the ring without cascade, the single flow channel with stationary guide vanes, and the single flow channel with rotating blades. Taking the single flow channel with rotating blades as an example, the U-net convolutional neural network architecture is as follows: Fig. 9 The trained proxy model is used to predict the normalized test samples outside the training set. The prediction results of the rotating single flow channel, the stationary guide vane single flow channel, and the annular area without cascades are shown as follows: Figure 10-12 shown.

[0165] S5: Predict the new full-channel impeller flow field information, the flow chart is as follows Fig.13 shown.

[0166] The specific process is:

[0167] ① The entire flow channel inside the impeller is divided into multiple annular areas without cascades, single flow channel areas with stationary guide vanes, and single flow channel areas with rotating blades according to the S1 method, and the initial values ​​of the velocity and pressure field distribution at the interface of each sub-area are assigned.

[0168] ② For each annular area without cascades, single flow channel area with stationary guide vanes, and single flow channel area with rotating blades, the corresponding proxy model is used to quickly predict the flow field, so as to obtain the full-channel flow field of the stationary guide vanes, the full-channel flow field of the rotating blades, and the full-channel flow field of the annular area without cascades.

[0169] ③ Determine whether the residuals of the velocity and pressure at the interfaces of the full-channel flow field meet the continuity conditions and meet the convergence criteria. If so, output the flow field and end the calculation. If not, the velocity and pressure of each interface are corrected according to the difference between their initial values ​​and the predicted values ​​of each proxy model. The corrected values ​​are used as new initial values ​​to participate in the next round of prediction process, and iterate until the convergence criteria are met. Inversely map the flow field on the computational domain back to the corresponding physical domain to obtain velocity and pressure information on the physical domain.

[0170] There are two prediction methods for correcting the pressure and velocity at the interface of each full-channel flow field: serial and parallel. The prediction of the full-channel impeller flow field with one blade row is used as an example to illustrate the boundary conditions of which are the inlet velocities U and V and the outlet pressure P. Fig.14 and Fig.15 As shown, the full-channel impeller flow field containing one blade row has three full-channel regions and two corresponding full-channel interfaces, namely ①, ②, ③ and Ⅰ, Ⅱ.

[0171] For string prediction, the iterative process of updating the pressure information of each interface is as follows: Fig.14 As shown, the specific steps are as follows:

[0172] Step 1: Assign initial values ​​to interfaces Ⅰ and Ⅱ in the figure

[0173] Step 2: Based on the left boundary velocity information U, V and the right boundary pressure information on area ① The velocity information on interface Ⅰ can be obtained by using the ring proxy model without blades and Obtained and At the same time, it serves as part of the input information of the proxy model corresponding to region ②;

[0174] Step 3: Based on the left boundary speed information on area ② With right boundary pressure information The velocity information on the interface II can be obtained by using the rotating or stationary single-channel proxy model corresponding to the region for combined prediction. and Obtained and At the same time, it serves as part of the input information of the proxy model corresponding to area ③;

[0175] Step 4: Based on the left boundary speed information on area ③ The pressure information on the right boundary P is predicted by using the blade-free annular proxy model corresponding to this area to obtain the pressure information on the interface II. Obtained At the same time, it serves as part of the input information of the proxy model corresponding to region ②;

[0176] Step 5: Based on the left boundary speed information on area ② With right boundary pressure information The pressure information on the interface I can be obtained by using the rotating or stationary single-channel proxy model corresponding to the region for combined prediction.

[0177] Step 6: Determine the and Whether the absolute value of the difference reaches the convergence standard (i = I, II), if it does not reach the convergence standard, adjust each interface for Among them, α is the set adjustment coefficient, and the adjusted As the new interface initial value, it returns to Step 1 to participate in the next round of flow field prediction until it reaches the continuity convergence standard.

[0178] In the string prediction method, the updating and iteration process of the speed information on each interface is as follows: Fig.15 As shown, the specific steps are as follows:

[0179] Step 1: Assign initial values ​​to interfaces Ⅰ and Ⅱ in the figure j = 1 represents interface I, j = 2 represents interface II;

[0180] Step 2: Based on the left boundary speed information on area ③ The pressure information on the right boundary P can be obtained by using the ring proxy model without blades. Obtained As part of the input information of the proxy model corresponding to region ②;

[0181] Step 3: Based on the left boundary speed information on area ② With right boundary pressure information The pressure information on the interface I can be obtained by using the rotating or stationary single-channel proxy model corresponding to the region for combined prediction. Obtained As part of the input information of the proxy model corresponding to region ①;

[0182] Step 4: Based on the left boundary velocity information U, V and the right boundary pressure information on area ① The velocity information on the interface I can be obtained by using the corresponding bladeless annular proxy model for prediction. Obtained At the same time, it serves as part of the input information of the proxy model corresponding to region ②;

[0183] Step 5: Based on the left boundary speed information on area ② With right boundary pressure information The velocity information on the interface II can be obtained by using the rotating or stationary single-channel proxy model corresponding to the region for combined prediction.

[0184] Step 6: Determine the and The absolute value of the difference and and Whether the absolute value of the difference reaches the convergence standard (j=1, 2), if it does not reach the convergence standard, adjust each interface They are Among them, α is the set adjustment coefficient, and the adjusted As the new interface initial value, it returns to Step 1 to participate in the next round of flow field prediction until it reaches the continuity convergence standard.

[0185] For parallel prediction, such as Fig.16 As shown, the specific steps are as follows:

[0186] Step 1: Assign initial values ​​to interfaces Ⅰ and Ⅱ in the figure

[0187] Step 2: Based on the left boundary flow field information U, V and the right boundary flow field information on area ① The velocity information on interface Ⅰ can be obtained by using the ring proxy model without blades and According to the left boundary flow field information on area ② and right boundary flow field information The pressure information on the interface I can be obtained by using the rotating or stationary single-channel proxy model corresponding to the region for combined prediction. and the velocity information on interface II and According to the left boundary flow field information on area ③ With the right boundary flow field information P, the pressure information on the interface II can be obtained by using the non-blade annular proxy model

[0188] Step 3: Determine the and and and Whether the absolute value of the difference reaches the convergence standard, if it does not reach the convergence standard, Adjust separately Among them, α is the set adjustment coefficient, and the adjusted As the new interface initial value, it returns to Step 1 to participate in the next round of flow field prediction until it reaches the continuity convergence standard.

[0189] In order to highlight the accuracy of this method in predicting the flow field of variable-order impellers, the order of the impeller to be predicted is selected as 4, which is different from the 3-order impeller in the training sample set. In addition, the blade profile of the new impeller is also different from the blade profile of the training sample set. The continuity convergence criterion is set as the difference between the flow field information before and after the update at 60% of the grid nodes on each interface is less than 0.01 before the update. α is set to 0.5. After 126 iterative predictions, the predicted flow field meets the continuity convergence criterion. With the number of iterations as the horizontal axis and the proportion of the number of nodes on the sub-region interface that meet the continuity convergence criterion as the vertical axis, a flow field convergence curve is plotted, as shown in the figure. Fig.17 A node on the sub-interface is randomly selected as the monitoring point, and the iterative process of its flow field information P, U, V is as follows: Fig.18 When all nodes on the sub-region interface meet the continuity convergence standard, the multi-level full-channel flow field information is output. Taking the velocity information U as an example, the absolute value cloud diagram of the predicted value, the actual value and the difference between the two is shown in Fig.19 shown.

[0190] The flow field prediction time of this example is 252s, which is much shorter than the time of solving the conventional NS equation. Fig.19 It can be seen that the difference between the predicted value and the true value of the flow field in most areas is close to 0, and after calculation, the prediction error of the velocity U at 97.43% of the nodes in the impeller flow field is less than 10%, which meets the requirements of engineering applications.

[0191] It should be noted that the present invention combines the proxy models of each sub-region and utilizes the continuity conditions of the physical quantities at the regional interface, and adopts iterative calculation to realize the rapid prediction of the flow in the multi-stage blade channel, which can not only realize the variable blade geometry but also realize the rapid prediction of the flow in the blade channel with variable stages and variable blade numbers.

[0192] At the same time, the present invention does not directly establish a proxy model for prediction of the flow field of the entire channel of a multi-stage blade cascade, but establishes proxy models for flow prediction for different sub-areas, such as a rotating blade cascade single flow channel, a stationary blade cascade single flow channel, and a bladeless channel. Through a single steady flow numerical simulation, flow field data of multiple bladeless channels, rotating blade single flow channels, and stationary blade single flow channel areas can be simultaneously obtained. After processing, they are used as training samples of the neural network models corresponding to the respective sub-areas. Multiple training samples can be obtained in one simulation, thereby avoiding the problem of high cost of constructing and training the proxy model for the multi-stage blade cascade flow.

[0193] In a fourth aspect, the present invention provides an electronic device, comprising: a memory and a processor, wherein the memory stores a computer program, and the processor executes the method provided in the first aspect or the second aspect of the present invention when executing the computer program.

[0194] The related technical solutions are the same as the methods provided in the first and second aspects of the present invention, and will not be described in detail here.

[0195] In a fifth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is executed by a processor, the device where the storage medium is located is controlled to execute the method provided in the first aspect or the second aspect of the present invention.

[0196] The related technical solutions are the same as the methods provided in the first and second aspects of the present invention, and will not be described in detail here.

[0197] It should be noted that the specific embodiments given in this specification are only for illustration and do not constitute the sole limitation of the specific implementation cases of the present invention. For those skilled in the art, based on the embodiments provided by the present invention, the above-mentioned scalable proxy model method for rapid prediction of the flow field of a turbomachinery can be adopted to realize the rapid prediction of the steady flow field of turbomachinery of different types, different working conditions, different rotor and stator arrangements, different stages and different dimensions. The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for constructing a full-flow flow field prediction model for an impeller machine, characterized in that: include: Taking the leading edge points and trailing edge points of each blade in each blade row as the dividing plane, the fluid domain of the impeller machine is divided into a non-blade annular domain and a blade row domain, and further taking the suction surface and pressure surface of each blade as the dividing plane, the blade row domain is divided into multiple blade single flow channel domains, thereby dividing the impeller machine into different types of physical domains; The blade single flow channel domain includes two types: a stationary guide vane single flow channel domain and a rotating blade single flow channel domain; Each type of physical domain is projected into a regular computational domain through coordinate transformation, and a corresponding proxy model is set. The corresponding proxy model is trained using a corresponding training set to obtain proxy models of various types of physical domains, thereby forming a full-channel flow field prediction model for turbomachinery. The training set corresponding to the physical domain includes: training samples and corresponding labels; The training samples include: velocity field distribution on the left boundary and pressure field distribution on the right boundary of the computational domain corresponding to the physical domain, and coordinate transformation parameters between the physical domain and the corresponding computational domain; the left and right boundaries of the computational domain correspond to the inlet and outlet surfaces of the fluid on the physical domain, respectively; The label includes: real flow field information on the computational domain corresponding to the physical domain; The flow field information includes: velocity field distribution and pressure field distribution.

2. The method for constructing a full-flow channel flow field prediction model for a turbomachinery according to claim 1, characterized in that: The process of transforming the physical domain into a regular computational domain includes: The physical domain is meshed and converted into a regular computational domain through coordinate transformation; each grid point in the computational domain corresponds one-to-one to each grid point in the corresponding physical domain.

3. The method for constructing a full-flow channel flow field prediction model for a turbomachinery according to claim 2, characterized in that: For any grid point coordinate (x, y) in the physical domain, the coordinate of the corresponding grid point in the computational domain is labeled (ξ, η); where ξ = ξ(x, y), η = η(x, y); The coordinate transformation parameters between the physical domain and the corresponding computational domain include:

4. The method for constructing a full-flow channel flow field prediction model for a turbomachinery according to any one of claims 1 to 3, characterized in that: The calculation domain is a rectangular calculation domain.

5. A method for predicting the flow field of the entire flow passage of a turbomachinery, characterized in that: include: Based on the full-flow channel flow field prediction model of the impeller machinery, the full-flow channel flow field information of the impeller machinery is predicted; The full-channel flow field prediction model of the impeller machinery is constructed by using the construction method described in any one of claims 1 to 4; The flow field information includes: velocity field distribution and pressure field distribution; The impeller machine comprises: M+1 bladeless annular domains and a blade row domain located between two adjacent bladeless annular domains; M≥1; the mth blade row domain D L,m With the mth bladeless annular region D B,m and the m+1th cascade-free annular region D B,m+1 The interface is denoted as interface Ι m and interface ΙΙ m ; m = 1, 2, ..., M; the number of blade single flow channel domains on the blade row domain is recorded as N, and the nth blade single flow channel domain on the mth blade row domain is recorded as D L,m,n ; The prediction of the flow field information of the entire flow passage of the impeller machinery comprises: Project each cascade-free annular domain and each blade single-channel domain into a regular computational domain through coordinate transformation, and obtain corresponding coordinate transformation parameters; For the interface in the computational domain m and interface ΙΙ m The velocity field distribution and pressure field distribution on the surface are predicted to obtain the interface Ι in the calculation domain m and interface ΙΙ m Velocity field distribution And the prediction results of pressure field distribution Will and D B,m The corresponding coordinate transformation parameters are input into the non-blade annular domain proxy model in the full-flow channel flow field prediction model of the impeller machinery to obtain D B,m The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D B,m Flow field information on S is the velocity field distribution at the left boundary of the calculation domain corresponding to the first pre-collected cascade annular domain; Will and D L,m,n The corresponding coordinate transformation parameters are input into the blade single flow channel domain proxy model in the impeller machinery full flow channel flow field prediction model to obtain D L,m,n The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D L,m,n Flow field information on for Middle D L,m,n The corresponding prediction results of the velocity field distribution at the left boundary of the computational domain; for Middle D L,m,n The corresponding prediction results of the pressure field distribution at the right boundary of the computational domain; Will and D B,m+1 The corresponding coordinate transformation parameters are input into the blade-free annular domain proxy model to obtain D B,m+1 The flow field information on the corresponding computational domain is inversely mapped back to the corresponding physical domain to obtain D B,m+1 Flow field information on P is the pressure field distribution at the right boundary of the calculation domain corresponding to the pre-collected M+1th cascade annular domain.

6. The method for predicting the flow field of the entire flow passage of a turbomachinery according to claim 5, characterized in that: For the interface in the computational domain m and interface ΙΙ m The pressure field distribution on the surface is predicted, including: A1. Initialize the interface in the computational domain I m and interface II m The pressure field distribution on and and D L,m,n The initial value of the pressure field distribution at the left and right boundaries of the corresponding calculation domain; A2. and D B,m The corresponding coordinate transformation parameters are input into the blade-free annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained, so as to obtain the current interface I in the calculation domain m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain; Will and D L,m,n The corresponding coordinate transformation parameters are input into the blade single flow channel domain proxy model to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding velocity field distribution at the right boundary of the computational domain Then we can get the current interface II in the computational domain m The velocity field distribution on Will and D B,m+1 The corresponding coordinate transformation parameters are input into the blade-free annular domain proxy model to obtain the current D B,m+1 The flow field information on the corresponding calculation domain is obtained, thereby obtaining the current interface II in the calculation domain m The pressure field distribution on For the current D L,m,n The corresponding pressure field distribution at the right boundary of the computational domain; Will and D L,m,n The corresponding coordinate transformation parameters are input into the blade single flow channel domain proxy model to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain Then we can get the current interface I in the computational domain m The pressure field distribution on A3. Judgment and and Are the absolute values ​​of the differences between the two values ​​less than the corresponding preset thresholds? If so, go to A4; otherwise, Go to A2; α1 is the preset adjustment coefficient; A4. Set the current interface I in the calculation domain m and interface II m The pressure field distribution on and As the interface in the computational domain m and interface ΙΙ m Prediction results of upper pressure field distribution Among them, m=1,2,…,M.

7. The method for predicting the flow field of the entire flow passage of a turbomachinery according to claim 5 or 6, characterized in that: For the interface in the computational domain m and interface ΙΙ m The velocity field distribution on the surface is predicted, including: B1. Initialize the interface in the computational domain I m and interface II m The velocity field distribution on and and D L,m,n The initial value of the velocity field distribution at the left and right boundaries of the corresponding calculation domain; B2. and D B,m+1 The corresponding coordinate transformation parameters are input into the blade-free annular domain proxy model to obtain the current D B,m+1 The flow field information on the corresponding calculation domain is obtained, thereby obtaining the current interface II in the calculation domain m The pressure field distribution on For the current D L,m,n The corresponding pressure field distribution at the right boundary of the calculation domain; Will and D L,m,n The corresponding coordinate transformation parameters are input into the blade single flow channel domain proxy model to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain Then we get the current interface I m The pressure field distribution on Will and D B,m The corresponding coordinate transformation parameters are input into the blade-free annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained, so as to obtain the current interface I in the calculation domain m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain; Will and D L,m,n The corresponding coordinate transformation parameters are input into the blade single flow channel domain proxy model to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding velocity field distribution on the right boundary of the computational domain Then we can get the current interface II in the computational domain m The velocity field distribution on B3. Judgment and and Are the absolute values ​​of the differences of are all less than the corresponding preset thresholds? If so, go to B4; otherwise, let Go to B2; α2 is the preset adjustment coefficient; B4. Set the current interface I in the calculation domain m and interface II m The velocity field distribution on and As the interface in the computational domain I m and interface II m Prediction results of velocity field distribution on Among them, m=M,M-1,…,1.

8. The method for predicting the flow field of the entire flow passage of a turbomachinery according to claim 5, characterized in that: The interface Ι in the calculation domain m and interface ΙΙ m The velocity field distribution and pressure field distribution on the surface are predicted, including: C1. Initialize the interface in the computational domain Ι m and interface ΙΙ m The velocity field distribution on and And the pressure field distribution and and D L,m,n The initial values ​​of the velocity field distribution at the left and right boundaries of the corresponding computational domain; and D L,m,n The initial values ​​of the pressure field distribution at the left and right boundaries of the corresponding calculation domain; C2. and D B,m The corresponding coordinate transformation parameters are input into the blade-free annular domain proxy model to obtain the current D B,m The flow field information on the corresponding calculation domain is obtained to obtain the current interface in the calculation domain. m The velocity field distribution on For the current D L,m,n The corresponding velocity field distribution at the left boundary of the computational domain; Will and D L,m,n The corresponding coordinate transformation parameters are input into the blade single flow channel domain proxy model to obtain the current D L,m,n The flow field information on the corresponding calculation domain is obtained to obtain the current D L,m,n The corresponding pressure field distribution on the left boundary of the computational domain And the velocity field distribution on the right boundary Then we can get the current interface in the computational domain. m The pressure field distribution on and the current interface ΙΙ m The velocity field distribution on Will and D B,m+1 The corresponding coordinate transformation parameters are input into the blade-free annular domain proxy model to obtain the current D B,m+1 The flow field information on the corresponding calculation domain is obtained to obtain the current interface ΙΙ in the calculation domain m The pressure field distribution on C3. Judgment and and and and Are the absolute values ​​of the differences between the two values ​​less than the corresponding preset thresholds? If so, go to C4; otherwise, let P Ⅰ 0 =P Ⅰ 0 +α(P Ⅰ 1 -P Ⅰ 0 ), Go to C2; α is the preset adjustment coefficient; C4, the current interface Ι m and interface ΙΙ m The velocity field distribution on And pressure field distribution As the interface in the computational domain m and interface ΙΙ m Velocity field distribution And the prediction results of pressure field distribution Among them, m=1,2,…,M.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and the processor executes the method according to any one of claims 1 to 8 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed by a processor, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 8.

Images