A bidirectional fluid-structure interaction calculation method based on masked deep neural network

Abstract

The present invention discloses a bidirectional fluid-solid coupling calculation method based on a masked deep neural network, and the main process of its implementation includes: step one, establishing a database; step two, constructing a training set and a test set according to a time series; step three, building and training a masked deep neural network; step four, combining the masked deep neural network with a computational structural mechanics solver to perform fluid-solid coupling calculations. The present invention establishes a flow field reduction model based on a masked deep neural network to replace the computational fluid dynamics solver part in the bidirectional fluid-solid coupling solution system, realizes rapid prediction of the flow field solution part during the solution process, and couples with the computational structural mechanics solver to reduce the computational cost of the bidirectional fluid-solid coupling solution system, realizes rapid calculation of bidirectional fluid-solid coupling problems while ensuring calculation accuracy, solves the problems of high computational cost and long computational cycle in engineering applications, and helps to optimize the design of fluid-solid coupling engineering cases.

Description

Technical Field

[0001] The present invention relates to the field of fluid-solid coupling calculation, and in particular to a bidirectional fluid-solid coupling calculation method based on a masked deep neural network. Background Art

[0002] Fluid-solid coupling problems are widely present in many engineering fields such as civil engineering, water conservancy, shipbuilding, aerospace, etc., such as wind-induced vibration of high-rise buildings and suspension bridges, vortex-induced vibration of offshore oil risers and submarine pipelines, flutter of flying wings, etc. The current methods for studying fluid-solid coupling problems include experimental research and numerical simulation. The experimental research cost of fluid-solid coupling problems is relatively expensive and it is difficult to obtain detailed flow field details data, making it difficult to be widely used in the design of engineering cases. The numerical simulation technology of fluid-solid coupling problems uses computational fluid dynamics methods and computational solid mechanics methods to solve the equations of the fluid part and the solid part respectively. For bidirectional fluid-solid coupling problems, it is necessary to iterate the fluid and solid calculations, so the calculation cost is extremely high. Especially when it comes to the calculation part of the computational fluid dynamics solver, because the calculation grid needs to ensure the boundary layer y + The requirements of the two-way fluid-structure coupling solution system make the number of grids large, which seriously restricts the computational efficiency and cost of the entire two-way fluid-structure coupling solution system. It is urgent to develop corresponding methods to accelerate the solution speed of the fluid part to improve the speed of the entire two-way fluid-structure coupling solution system.

[0003] In recent years, intelligent flow field research methods based on machine learning have developed rapidly, which can realize the rapid prediction of unsteady flow fields and provide a new method for obtaining flow field information. The idea is to use data-driven technology to capture the key features of the flow field and use very little computation to simulate the flow field. Compared with traditional flow field reduction methods based on intrinsic orthogonal decomposition and dynamic mode decomposition, deep learning methods in machine learning have stronger nonlinear description capabilities and are more suitable for the prediction of flow fields with strong unsteady and nonlinear characteristics in the process of solving bidirectional fluid-solid coupling.

[0004] The present invention fully combines the advantages of machine learning methods and proposes a bidirectional fluid-solid coupling calculation method based on a masked deep neural network to solve the problems of low calculation efficiency and high calculation cost of the fluid part in the traditional bidirectional fluid-solid coupling solution system. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] The purpose of the present invention is to provide a bidirectional fluid-solid coupling calculation method based on a masked deep neural network, which aims to replace the computational fluid dynamics solver part of the bidirectional fluid-solid coupling solution system by establishing a flow field reduction model based on a masked deep neural network, realize rapid prediction of the flow field solution part during the solution process, and couple with the computational structural mechanics solver to reduce the computational cost of the bidirectional fluid-solid coupling solution system, realize rapid calculation of bidirectional fluid-solid coupling problems while ensuring calculation accuracy, and solve the problems of high computational cost and long calculation cycle in engineering applications.

[0007] (II) Technical solution

[0008] In order to solve the above technical problems, the present invention provides a bidirectional fluid-solid coupling calculation method based on a masked deep neural network, comprising the following steps:

[0009] Step 1: Establish a database;

[0010] Step 2: construct training set and test set according to time series;

[0011] Step 3: Build and train a masked deep neural network;

[0012] Step 4: Combine the masked deep neural network with the computational structural mechanics solver to perform fluid-structure coupling calculations;

[0013] ① The establishment of a database comprises:

[0014] Determine the parameters that need to be generalized for the fluid-solid coupling problem under study, use a fluid-solid coupling calculation system based on a computational fluid dynamics solver and a computational structural mechanics solver to perform multi-time series numerical simulation calculations on the fluid-solid coupling problem example under study, and obtain flow field data at each time step and solid field data at each time step; wherein the spatial position of the fluid-solid coupling problem example under study includes the flow field position and the solid field position; the time series is composed of time step 1 to time step t, and the value of the time step t is a positive integer;

[0015] Projecting the flow field data of each time step to a fluid matrix F, and projecting the solid field data of each time step to a solid matrix S; wherein the fluid matrix F and the solid matrix S have the same dimension;

[0016] The flow field position corresponding to the solid matrix S is assigned a value of 1, and the solid field position corresponding to the solid matrix S is assigned a value of 0; the database is composed of the fluid matrix F and the solid matrix S;

[0017] ② The construction of training set and test set according to time series includes:

[0018] The fluid matrix F is normalized to obtain a normalized fluid matrix F n , the normalized fluid matrix F n Multiply the solid matrix S by the dot to get the mask fluid matrix F m , that is, F m =F n ·S;

[0019] The mask fluid matrix at time step t-1 and the solid matrix S with time step t t Perform splicing processing on the time dimension to obtain the variable matrix T with a time step of t t ; The mask fluid matrix with time step t As the variable matrix T t The label matrix of each time step is t and the variable matrix T at each time step t The label matrices are concatenated to obtain the masked deep neural network training set and the masked deep neural network test set;

[0020] ③ The construction and training of the masked deep neural network includes:

[0021] The masked deep neural network is constructed by using a convolutional layer, a convolutional long-short-time neural network layer and a deconvolutional layer; the convolutional layer is used to capture the spatial features of the flow field data at each time step and reduce the dimension; the convolutional long-short-time neural network layer is based on the masked fluid matrix of the time step t-1 and the solid matrix S with time step t t Predict the low-dimensional flow field features with a time step of t; use the deconvolution layer to reconstruct the low-dimensional flow field features with a time step of t to obtain a high-dimensional flow field with a time step of t The high-dimensional flow field and the solid matrix S at the time step t t Perform dot multiplication to obtain the predicted mask fluid matrix Right now The prediction mask fluid matrix Perform anti-normalization processing to obtain the real flow field;

[0022] The masked deep neural network training set and the masked deep neural network test set constructed in step 2 are used to train the constructed masked deep neural network to obtain an available masked deep neural network model; the available masked deep neural network model can be used according to the fluid matrix F at time step t-1 t-1 and the solid matrix S with time step t t Complete the prediction of flow field data with time step t;

[0023] ④ The fluid-solid coupling calculation by combining the masked deep neural network with the computational structural mechanics solver includes:

[0024] A fluid-solid coupling computing system based on machine learning is constructed using a computational structural mechanics solver and the available masked deep neural network model obtained in step 3; the fluid-solid coupling computing system based on machine learning is composed of a solid module and a fluid module, the solid module and the fluid module use different execution kernels, and the solid module execution kernel and the fluid module execution kernel interact sequentially through a judgment file;

[0025] At the beginning of the calculation, according to the case of the fluid-solid interaction problem to be solved, the fluid matrix F corresponding to the initial location of the flow field is given 0 and the solid field initial field; the solid module obtains the solid boundary according to the solid field initial field, stores it as a solid boundary file, and generates a judgment file; the fluid module execution kernel converts the solid boundary file into a solid matrix S with a time step of 1 according to the judgment file 1 ; The solid matrix S with the time step of 1 1 The fluid matrix F corresponding to the initial position of the flow field 0 As the input of the available mask deep neural network model obtained in step 3, the fluid matrix F with time step 1 1 Make predictions;

[0026] Iterate according to the time series, and obtain the fluid matrix F at the time step t predicted by the available mask deep neural network model t The solid boundary pressure with a time step of t is obtained and stored as a solid boundary pressure file with a time step of t; when the time step is advanced from t to t+1, the judgment file is deleted, and when the solid module execution kernel detects that the judgment file does not exist, the data in the solid boundary pressure file with a time step of t is passed to the computational structural mechanics solver as the boundary condition for solving the solid field, and the computational structural mechanics solver is used to complete the solution of the solid deformation and displacement of the fluid-solid coupling problem case to be solved, and the solid boundary is obtained and stored as a solid matrix S with a time step of t+1. t+1 ; The solid matrix S with time step t+1 t+1 The fluid matrix F at time step t predicted by the available mask deep neural network model t As the input of the available masked deep neural network model obtained in step 3, the fluid matrix F at time step t+1 is t+1 Make predictions;

[0027] Iterate in order from time step 1 to time step t in the time series until the total number of iteration steps T is reached, thereby completing the bidirectional fluid-solid coupling calculation based on the masked deep neural network.

[0028] (III) Beneficial effects

[0029] A bidirectional fluid-solid coupling calculation method based on a masked deep neural network provided by the present invention has the following beneficial effects: by constructing a bidirectional fluid-solid coupling calculation system that combines a masked deep neural network with a computational structural mechanics solver, the problem that traditional bidirectional fluid-solid coupling solution systems are difficult to quickly and accurately solve complex geometric fluid-solid coupling cases is solved, the solution speed of bidirectional fluid-solid coupling cases is greatly improved, and the rapid solution of fluid-solid coupling problems in engineering cases is realized, which is helpful for the optimal design of fluid-solid coupling engineering cases.

[0030] The available masked deep neural network model obtained by training the method of the present invention can quickly predict the flow field according to the solid deformation and displacement conditions, and achieve the effect of accurately predicting the flow field within the parameter training range.

[0031] The method of the present invention uses a computational structural mechanics solver controlled by Python and the Tensorflow library, and has the characteristics of simple code, short training time, and easy extension to existing programs.

[0032] The method of the present invention has shown good application effects in multiple classic case tests, achieving the effect of shortening the calculation time of each time step of the flow field solution part to millisecond level, and has good generalization performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention;

[0034] Figure 2 A schematic diagram of a masked deep neural network of a bidirectional fluid-solid coupling calculation method based on a masked deep neural network according to the present invention;

[0035] Figure 3 A fluid-solid coupling calculation system and a file interaction roadmap for a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention;

[0036] Figure 4 It is a schematic diagram of the flow field grid when the vertical plate is deformed in the flexible vertical plate oscillation calculation example of the embodiment of the present invention;

[0037] Figure 5A comparison diagram of calculation results of the vertical plate top displacement of the flexible vertical plate oscillation example of the embodiment of the present invention using a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention and using an original bidirectional fluid-solid coupling solution system;

[0038] Figure 6 It is a comparison chart of the calculation results of the pressure at the top of the vertical plate of the flexible vertical plate oscillation example of the embodiment of the present invention using a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention and using the original bidirectional fluid-solid coupling solution system. DETAILED DESCRIPTION

[0039] The following is a further detailed description of the specific implementation of the present invention in combination with the accompanying drawings and embodiments, taking the fluid-solid coupling problem of the oscillation of a flexible vertical plate in a static fluid as an example. The following embodiments are only used to illustrate the present invention, but are not used to limit the scope of the present invention.

[0040] The present invention provides a bidirectional fluid-solid coupling calculation method based on a masked deep neural network, comprising the following steps:

[0041] Step 1: Establish a database;

[0042] In this step, the parameters that need to be generalized for the fluid-solid coupling problem under study are determined, and a fluid-solid coupling calculation system based on a computational fluid dynamics solver and a computational structural mechanics solver is used to perform multi-time series numerical simulation calculations on the fluid-solid coupling problem example under study, to obtain the flow field data for each time step and the solid field data for each time step; wherein, the spatial position of the fluid-solid coupling problem example under study includes the flow field position and the solid field position; the time series is composed of time step 1 to time step t, and the value of time step t is a positive integer. The flow field data for each time step is projected onto the fluid matrix F, and the solid field data for each time step is projected onto the solid matrix S; wherein, the fluid matrix F and the solid matrix S have the same dimension. The flow field position corresponding to the solid matrix S is assigned a value of 1, and the solid field position corresponding to the solid matrix S is assigned a value of 0; the database is composed of the fluid matrix F and the solid matrix S;

[0043] In the flexible vertical plate oscillation example of the embodiment of the present invention, the parameter that needs to be generalized is the Young's modulus E of the vertical plate. The fluid-solid coupling calculation system Abaqus2016 based on the computational fluid dynamics solver and the computational structural mechanics solver is used to perform multi-time series numerical simulation calculations on the flexible vertical plate oscillation example studied. Figure 4The figure shows a schematic diagram of the flow field grid when the vertical plate of the flexible vertical plate oscillation example of the embodiment of the present invention is deformed. The flow field grid adopts a structured grid and is encrypted around the solid wall, so that even after the solid is deformed, the grid can still ensure good quality. The Young's modulus E = 250MPa, E = 300MPa, and E = 350MPa are calculated respectively, with a calculation time step of 0.1s and a total calculation time of 50s; wherein each different Young's modulus E is calculated from time step 1 to time step 500, and 1500 flow field data are obtained. The flow field data of each time step is projected to the fluid matrix F, whose dimension is (224, 224, 1), and the solid field data of each time step is projected to the solid matrix S, whose dimension is (224, 224, 1), and the fluid matrix F has the same dimension as the solid matrix S. The flow field position corresponding to the solid matrix S is assigned to 1, and the solid field position corresponding to the solid matrix S is assigned to 0; the constructed database is composed of the fluid matrix F and the solid matrix S.

[0044] Step 2: construct training set and test set according to time sequence;

[0045] In this step, the fluid matrix F is normalized to obtain the normalized fluid matrix F n , the normalized fluid matrix F n Multiply the solid matrix S by the dot to get the mask fluid matrix F m , that is, F m =F n S; the mask fluid matrix at time step t-1 and the solid matrix S with time step t t Perform splicing processing on the time dimension to obtain the variable matrix T with a time step of t t ; The mask fluid matrix with time step t As the variable matrix T t The label matrix of each time step; the variable matrix T t and the variable matrix T at each time step t The label matrices are concatenated to obtain the masked deep neural network training set and the masked deep neural network test set.

[0046] In the flexible vertical plate oscillation calculation example of the embodiment of the present invention, the maximum and minimum values ​​of the fluid matrix F are obtained to normalize the fluid matrix F to obtain the normalized fluid matrix F n ; Because the solid matrix has only 0 and 1, no normalization is required. The normalized fluid matrix F n Multiply the solid matrix S by the dot to get the mask fluid matrix F m , that is, F m =F nS, this operation can make the solid boundary more obvious to enhance the boundary capture ability of the neural network. and the solid matrix S with time step t t Perform splicing processing on the time dimension to obtain the variable matrix T with a time step of t t , whose dimensions are (1500,2,224,224,1), and the mask fluid matrix with time step t As the variable matrix T t The label matrix has a dimension of (1500, 224, 224, 1); the variable matrix T of each time step t and the variable matrix T at each time step t The label matrices are spliced ​​to obtain the masked deep neural network training set and the masked deep neural network test set of the flexible vertical plate oscillation example of the embodiment of the present invention.

[0047] Step 3: Build and train a masked deep neural network;

[0048] In this step, the masked deep neural network is constructed using convolutional layers, convolutional long-short-time neural network layers, and deconvolutional layers; the convolutional layers are used to capture the spatial features of the flow field data at each time step and reduce the dimension; the convolutional long-short-time neural network layers are used to generate the masked fluid matrix according to the time step t-1. and the solid matrix S with time step t t Predict the low-dimensional flow field features with a time step of t; use the deconvolution layer to reconstruct the low-dimensional flow field features with a time step of t to obtain a high-dimensional flow field with a time step of t High dimensional flow field and the solid matrix S with time step t t Perform dot multiplication to obtain the predicted mask fluid matrix Right now Prediction mask fluid matrix The real flow field is obtained by performing anti-normalization processing.

[0049] The masked deep neural network training set and masked deep neural network test set constructed in step 2 are used to train the constructed masked deep neural network to obtain an available masked deep neural network model; the available masked deep neural network model can be used according to the fluid matrix F at time step t-1 t-1 and the solid matrix S with time step t t Complete the prediction of flow field data with time step t.

[0050] Figure 2The figure is a schematic diagram of a masked deep neural network of a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention. In the flexible vertical plate oscillation example of the embodiment of the present invention, the masked deep neural network constructed is divided into a convolution layer, a convolutional long-short-time neural network layer and a deconvolution layer. The fluid matrix of the previous time step and the solid matrix after deformation of the current time step are used as input, and the fluid matrix of the current time step is output. The convolution layer is used to capture the spatial characteristics of the flow field and reduce the dimension, and the low-dimensional flow field characteristics are used to predict the flow field changes to improve the calculation efficiency. Since the fluid-solid coupling problem of the flexible vertical plate oscillating in the static fluid is relatively complex, it is necessary to perform multi-layer convolution operations. The residual connection structure of the residual neural network can overcome the gradient vanishing problem of the deep network; the mask operation can improve the prediction accuracy of the flow field.

[0051] Step 4: Combine the masked deep neural network with the computational structural mechanics solver to perform fluid-structure coupling calculations;

[0052] In this step, the computational structural mechanics solver and the available masked deep neural network model obtained in step 3 are used to build a fluid-solid coupling computing system based on machine learning. The fluid-solid coupling computing system based on machine learning consists of a solid module and a fluid module. The solid module and the fluid module use different execution kernels, and the solid module execution kernel and the fluid module execution kernel interact sequentially through the judgment file. Figure 3 As shown, a fluid-solid coupling calculation system and a file interaction roadmap of a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention are provided. The core of the fluid-solid coupling calculation system is to realize the data interaction between the fluid module execution kernel and the solid module execution kernel through a judgment file, and the flow field and solid field are calculated respectively to complete the bidirectional fluid-solid coupling calculation.

[0053] At the beginning of the calculation, according to the case of the fluid-solid interaction problem to be solved, the fluid matrix F corresponding to the initial location of the flow field is given 0 and the initial field of the solid field; the solid module obtains the solid boundary according to the initial field of the solid field, stores it as a solid boundary file, and generates a judgment file; the execution kernel of the fluid module converts the solid boundary file into a solid matrix S with a time step of 1 according to the judgment file 1 ; The solid matrix S with time step 1 1 The fluid matrix F corresponding to the initial position of the flow field 0 As the input of the available masked deep neural network model obtained in step 3, the fluid matrix F with time step 1 1 Make predictions.

[0054] Iterate in time series, starting from the fluid matrix F at time step t predicted by the available masked deep neural network model tThe solid boundary pressure with time step t is obtained in the calculation, and stored as a solid boundary pressure file with time step t; when the time step advances from t to t+1, the judgment file is deleted. When the execution kernel of the solid module detects that there is no judgment file, the data in the solid boundary pressure file with time step t is passed to the computational structural mechanics solver as the boundary condition for solving the solid field. The computational structural mechanics solver is used to complete the solution of the solid deformation and displacement of the fluid-solid coupling problem case to be solved, and the solid boundary is obtained and stored as a solid matrix S with time step t+1. t+1 ; The solid matrix S with time step t+1 t+1 and the fluid matrix F at time step t predicted by the available masked deep neural network model t As the input of the available masked deep neural network model obtained in step 3, the fluid matrix F at time step t+1 is t+1 Make predictions.

[0055] Iterate in order from time step 1 to time step t according to the time series until the total number of iteration steps T is reached, and the bidirectional fluid-solid coupling calculation based on the masked deep neural network is completed.

[0056] like Figure 5 As shown, it is a comparison chart of the calculation results of the top displacement of the flexible vertical plate of the embodiment of the present invention using a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention and a traditional method. This figure corresponds to the generalized example of E=275MPa. In the legend, "CFD" represents the fluid-solid coupling calculation result of the traditional method based on the computational fluid dynamics solver, and "MDNN" represents the calculation result of the bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention. The shaded part represents the 50s training time of all training sets, and the time after 50s to 100s is not in the database.

[0057] Figure 6 This is a comparison chart of the calculation results of the flexible vertical plate oscillation example using a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention and a traditional method for the pressure of the monitoring point on the top of the flexible vertical plate. In the legend, "CFD" represents the fluid-solid coupling calculation result of the traditional method based on the computational fluid dynamics solver, and "MDNN" represents the calculation result using a bidirectional fluid-solid coupling calculation method based on a masked deep neural network of the present invention.

[0058] from Figure 5 , Figure 6 From the comparison of the results, it can be seen that the bidirectional fluid-solid coupling calculation method based on the masked deep neural network of the present invention can accurately capture parameters such as the pressure of the solid boundary and the solid displacement while quickly predicting the flow field, ensuring the accurate calculation of the solid deformation and having good generalization ability.

[0059] The above description 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.

[0060] In summary, the present invention establishes a flow field reduction model based on a masked deep neural network, replaces the computational fluid dynamics solver part of the bidirectional fluid-solid coupling solution system, and performs bidirectional fluid-solid coupling calculations in conjunction with the computational structural mechanics solver, thereby achieving rapid prediction of the flow field solution part during the solution process and reducing the computational cost of the bidirectional fluid-solid coupling solution system. The available masked deep neural network model obtained can quickly predict the flow field according to the solid deformation and displacement conditions, and achieve the effect of accurately predicting the flow field within the parameter training range. While ensuring the calculation accuracy, the rapid calculation of the bidirectional fluid-solid coupling problem is achieved, solving the problems of high computational cost and long calculation cycle of bidirectional fluid-solid coupling solutions in engineering applications, and providing a fast and accurate technical method for solving bidirectional fluid-solid coupling problems.

Claims

1. A bidirectional fluid-structure interaction calculation method based on masked deep neural network, characterized in that: The steps include: Step 1: Establish a database; Step 2: construct training set and test set according to time series; Step 3: Build and train a masked deep neural network; Step 4: Combine the masked deep neural network with the computational structural mechanics solver to perform fluid-structure coupling calculations; ① The establishment of a database comprises: Determine the parameters that need to be generalized for the fluid-solid coupling problem under study, use a fluid-solid coupling calculation system based on a computational fluid dynamics solver and a computational structural mechanics solver to perform multi-time series numerical simulation calculations on the fluid-solid coupling problem example under study, and obtain flow field data at each time step and solid field data at each time step; wherein the spatial position of the fluid-solid coupling problem example under study includes the flow field position and the solid field position; the time series is composed of time step 1 to time step t, and the value of the time step t is a positive integer; Projecting the flow field data of each time step to a fluid matrix F, and projecting the solid field data of each time step to a solid matrix S; wherein the fluid matrix F and the solid matrix S have the same dimension; The flow field position corresponding to the solid matrix S is assigned a value of 1, and the solid field position corresponding to the solid matrix S is assigned a value of 0; the database is composed of the fluid matrix F and the solid matrix S; ② The construction of training set and test set according to time series includes: The fluid matrix F is normalized to obtain a normalized fluid matrix F n , the normalized fluid matrix F n Multiply the solid matrix S by the dot to get the mask fluid matrix F m , that is, F m =F n ·S; The mask fluid matrix at time step t-1 and the solid matrix S with time step t t Perform splicing processing on the time dimension to obtain the variable matrix T with a time step of t t ; The mask fluid matrix with time step t As the variable matrix T t The label matrix of each time step is t and the variable matrix T at each time step t The label matrices are concatenated to obtain the masked deep neural network training set and the masked deep neural network test set; ③ The construction and training of the masked deep neural network includes: The masked deep neural network is constructed by using a convolutional layer, a convolutional long-short-time neural network layer and a deconvolutional layer; the convolutional layer is used to capture the spatial features of the flow field data at each time step and reduce the dimension; the convolutional long-short-time neural network layer is based on the masked fluid matrix of the time step t-1 and the solid matrix S with time step t t Predict the low-dimensional flow field features with a time step of t; use the deconvolution layer to reconstruct the low-dimensional flow field features with a time step of t to obtain a high-dimensional flow field with a time step of t The high-dimensional flow field and the solid matrix S at the time step t t Perform dot multiplication to obtain the predicted mask fluid matrix Right now The prediction mask fluid matrix Perform anti-normalization processing to obtain the real flow field; The masked deep neural network training set and the masked deep neural network test set constructed in step 2 are used to train the constructed masked deep neural network to obtain an available masked deep neural network model; the available masked deep neural network model can be used according to the fluid matrix F at time step t-1 t-1 and the solid matrix S with time step t t Complete the prediction of flow field data with time step t; ④ The fluid-solid coupling calculation by combining the masked deep neural network with the computational structural mechanics solver includes: A fluid-solid coupling computing system based on machine learning is constructed using a computational structural mechanics solver and the available masked deep neural network model obtained in step 3; the fluid-solid coupling computing system based on machine learning is composed of a solid module and a fluid module, the solid module and the fluid module use different execution kernels, and the solid module execution kernel and the fluid module execution kernel interact sequentially through a judgment file; At the beginning of the calculation, according to the case of the fluid-solid interaction problem to be solved, the fluid matrix F corresponding to the initial location of the flow field is given 0 and the solid field initial field; the solid module obtains the solid boundary according to the solid field initial field, stores it as a solid boundary file, and generates a judgment file; the fluid module execution kernel converts the solid boundary file into a solid matrix S with a time step of 1 according to the judgment file 1 ; The solid matrix S with the time step of 1 1 The fluid matrix F corresponding to the initial position of the flow field 0 As the input of the available mask deep neural network model obtained in step 3, the fluid matrix F with time step 1 1 Make predictions; Iterate according to the time series, and obtain the fluid matrix F at the time step t predicted by the available mask deep neural network model t The solid boundary pressure with a time step of t is obtained and stored as a solid boundary pressure file with a time step of t; when the time step is advanced from t to t+1, the judgment file is deleted, and when the solid module execution kernel detects that the judgment file does not exist, the data in the solid boundary pressure file with a time step of t is passed to the computational structural mechanics solver as the boundary condition for solving the solid field, and the computational structural mechanics solver is used to complete the solution of the solid deformation and displacement of the fluid-solid coupling problem case to be solved, and the solid boundary is obtained and stored as a solid matrix S with a time step of t+1. t+1 ; The solid matrix S with time step t+1 t+1 The fluid matrix F at time step t predicted by the available mask deep neural network model t As the input of the available mask deep neural network model obtained in step 3, the fluid matrix F at time step t+1 is t+1 Make predictions; Iterate in order from time step 1 to time step t in the time series until the total number of iteration steps T is reached, thereby completing the bidirectional fluid-solid coupling calculation based on the masked deep neural network.

Images