A digital twin method for shield TBM equipment operation status based on coupling CAE simulation and machine learning

Abstract

The present invention relates to a digital twin method for the operating status of a shield TBM equipment based on the coupling of CAE simulation and machine learning, and the steps are as follows: S1: establishing a digital model of the shield TBM equipment; S2: applying constraint boundary conditions; S3: establishing an equipment load set; S4: extracting load samples and applying the loads in the model; S5: solving the finite element model; S6: acquiring data such as equipment structural stress and deformation results to form a data set; S7: machine learning model training; S8: outputting data such as shield TBM equipment operating stress and deformation; S9: visualizing the data twin model; the above process couples CAE simulation with machine learning to establish a nonlinear mapping relationship between load data and state data such as stress and strain, thereby realizing rapid prediction of stress, strain and other states, which is of great significance to improving the level of structural stress state assessment during the operation of the shield TBM equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of shield TBM equipment operating status assessment methods, and more specifically, it relates to a shield TBM equipment operating status digital twin method based on the coupling of CAE simulation and machine learning. Background Art

[0002] During shield TBM construction, the stress and strain state of the equipment structure is crucial for evaluating its operational status. Traditional CAE simulation analysis, due to the complex equipment model, consumes significant computational resources and takes a long time to perform finite element calculations. Furthermore, after inputting operating loads, it is impossible to display the equipment's stress and strain states with minimal latency.

[0003] In order to improve the timeliness and accuracy of the digital twin of the shield TBM operating status and give full play to the guiding role of digital twin technology in shield operation, especially to solve the problems of shield TBM equipment operation evaluation and construction parameter selection under complex working conditions, a digital twin method for the operating status of shield TBM equipment based on the coupling of CAE simulation and machine learning was proposed. Summary of the Invention

[0004] In view of the shortcomings of the existing technology, the purpose of the present invention is to propose a digital twin method for the operating status of shield TBM equipment based on the coupling of CAE simulation and machine learning.

[0005] The present invention provides the following technical solution: a digital twin method for the operating status of shield TBM equipment based on the coupling of CAE simulation and machine learning, comprising the following steps:

[0006] S1: Establish a digital model of the shield TBM equipment. Use 3D modeling software to build a 3D model of the shield TBM equipment. Use meshing tools to mesh the 3D model of the shield TBM equipment to form a digital model of the shield TBM equipment.

[0007] S2: Apply constraint boundary conditions; based on the structural characteristics of the shield TBM equipment and the load transfer path, apply the equipment operation process constraint boundary conditions on the digital model of the shield TBM equipment;

[0008] S3: Establishing an equipment load set; analyzing the combination of various loads, the direction of each load, and the numerical range of the load magnitude during the operation of the shield TBM equipment to form a shield TBM equipment load set;

[0009] S4: Extract load samples and apply them to the model; sample possible load combinations, directions, and magnitudes from the TBM equipment load set to form a load sample data set. Each load group in the sampled load sample data set is added to the digital model of the TBM equipment to form a finite element model with complete constraint boundary conditions and load boundary conditions.

[0010] S5: finite element model solution; calling the CAE solver to solve the finite element model with different constraint boundary conditions or load boundary conditions generated in step S4;

[0011] S6: Obtain data such as equipment structural stress and deformation results to form a data set; create a special data file, which includes constraint boundary conditions, load boundary conditions, stress, strain results and other data;

[0012] S7: Machine learning model training: The obtained data set consisting of constraint boundary conditions, load boundary conditions, etc. is divided into a training set and a test set. Different machine learning algorithms are used to train the model. The machine learning algorithm model with the highest accuracy in predicting the relationship between input and output is selected as the target model.

[0013] S8: Outputs operational stress, deformation, and other data of the shield TBM equipment. Based on the target model obtained through machine learning, the load data and constraint boundary conditions during the equipment operation are used as input, loaded into the target model for calculation and output, and the operational stress, deformation, and other data of the shield TBM equipment are obtained and stored as a result output file.

[0014] S9: Data twin model visualization; three-dimensional visualization of the stress, deformation and other data of the shield TBM equipment operation status, so as to intuitively obtain the stress and strain status of the equipment during operation.

[0015] Furthermore, in step S1, a three-dimensional model is constructed using the shield TBM equipment drawings. When conditions permit, the three-dimensional model is directly obtained from the equipment manufacturer and used. When the three-dimensional model file cannot be directly obtained from the equipment manufacturer, a reverse engineering method such as three-dimensional laser scanning is used to obtain a three-dimensional model of the parts or assemblies of the shield TBM equipment. The three-dimensional model of the shield TBM equipment is discretized by finite element mesh to obtain a finite element mesh model.

[0016] Furthermore, in step S2, the types of constraint boundary conditions and the positions of the constraints are analyzed according to the structural characteristics of the digital twin object, and the constraint boundary conditions are applied to the finite grid model.

[0017] Furthermore, in step S3, a set of shield TBM equipment loads is established based on the big data platform and professional knowledge such as mechanisms; specifically, the direction or theoretical value of a specific load is determined through mechanism knowledge, and the size range of specific loads of similar equipment is obtained by querying data collected by the big data platform.

[0018] Furthermore, in step S4, orthogonal arrays, Monte Carlo sampling, Latin square sampling, etc. are used to sample possible combinations, directions, and sizes of loads to form a load sample data set, and the sampled loads are sequentially added to the finite element mesh model through programming methods.

[0019] Furthermore, in step S5, a CAE cloud platform is used to perform calculations to solve the finite element model.

[0020] Furthermore, in step S7, when the machine learning model is trained, the proportion P1 of the training set in the data set is determined, and a training set is randomly selected from the data set according to the proportion P1, and the remaining data in the data set is used as a test set;

[0021] The portion of the training set containing constraint boundary conditions and load boundary condition data is classified as the test set data input, and the remaining data is used as the test set data output. The test set data input and output are normalized separately and mapped to the interval [0, 1] or [-1, 1] according to the data range. The obtained dataset consisting of constraint boundary conditions, load boundary conditions, etc. is divided into a training set and a test set. Different machine learning algorithms are used to train the model, and the model with the highest accuracy in predicting the relationship between input and output is selected as the target model.

[0022] Convolutional neural network, RNN neural network and BP neural network algorithms were selected for model training and verified with test set data. Finally, the relative standard deviation between the actual value and the predicted value was used to evaluate the model, and the model with high prediction accuracy was selected as the final target model.

[0023] Furthermore, in step S8, the stress, deformation and other data outputs of the shield TBM equipment operating status are obtained in step S1, and the load data collected during the operation of the shield TBM or the data highly correlated with the load are converted into load data through specific operations, and then input into the target model for calculation after normalization, and the output is stress, deformation and other data. Then, the denormalization is carried out in combination with the normalization parameters of the same category of data, and the output data is restored to the real range, and the stress, deformation and other data of the shield TBM operating status are output.

[0024] Furthermore, in step S9, the obtained data on the operating status of the shield TBM, such as stress and deformation, are visualized through the ParaView open source program or programming.

[0025] In summary, the present invention has the following beneficial effects: through the digital twin method of the operating status of shield TBM equipment coupled with CAE simulation and machine learning, a nonlinear mapping relationship between load data and stress, strain and other state data is established, so that after collecting load data or other data with strong correlation with load data, the stress, strain and other states can be quickly predicted, eliminating the time-consuming finite element calculation simulation process, greatly improving the calculation efficiency, and making the calculation accuracy meet the engineering requirements. It can quickly display the stress, strain and other states to guide the development of shield TBM construction and fully explore the value of data to the project; it can improve the timeliness and accuracy of the digital twin of the operating status of shield TBM, especially solve the problems of shield TBM equipment operation evaluation and construction parameter selection under complex working conditions, which is of great significance to improving the level of structural stress state evaluation during the operation of shield TBM equipment, giving full play to the guiding role of data in construction production activities, and improving the scientific level and refinement of shield construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flowchart of a specific embodiment 1 of the digital twin method of the operating status of shield TBM equipment based on the coupling of CAE simulation and machine learning of the present invention. DETAILED DESCRIPTION

[0027] The present invention is described in detail below with reference to the accompanying drawings.

[0028] Specific embodiment 1 of the present invention of a digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning:

[0029] like Figure 1 As shown in the figure, a digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning includes the following steps:

[0030] S1: Establish a digital model of the shield TBM equipment. Use 3D modeling software to build a 3D model of the shield TBM equipment. Use meshing tools to mesh the 3D model of the shield TBM equipment to form a digital model of the shield TBM equipment.

[0031] Specifically, the 3D model is constructed using the drawings of the shield TBM equipment. When conditions permit, the 3D model is obtained directly from the equipment manufacturer and used. When it is impossible to obtain the 3D model file directly from the equipment manufacturer, reverse engineering methods such as 3D laser scanning are used to obtain the 3D model of the parts or assemblies of the shield TBM equipment. The 3D model of the shield TBM equipment is discretized by finite element mesh to obtain a finite element mesh model.

[0032] S2: Apply constraint boundary conditions; based on the structural characteristics of the shield TBM equipment and the load transfer path, apply the equipment operation process constraint boundary conditions on the digital model of the shield TBM equipment;

[0033] Specifically, the types of constraint boundary conditions and the locations of constraints are analyzed according to the structural characteristics of the digital twin object, and the constraint boundary conditions are applied to the finite grid model.

[0034] S3: Establishing an equipment load set; analyzing the combination of various loads, the direction of each load, and the numerical range of the load magnitude during the operation of the shield TBM equipment to form a shield TBM equipment load set;

[0035] Specifically, a set of TBM equipment loads is established based on the big data platform and professional knowledge such as mechanisms; specifically, the direction or theoretical value range of a specific load is determined through mechanism knowledge, and the size range of specific loads of similar equipment is obtained by querying data collected by the big data platform.

[0036] S4: Extract load samples and apply them to the model; sample possible load combinations, directions, and magnitudes from the TBM equipment load set to form a load sample data set. Each load group in the sampled load sample data set is added to the digital model of the TBM equipment to form a finite element model with complete constraint boundary conditions and load boundary conditions.

[0037] Specifically, orthogonal arrays, Monte Carlo sampling, Latin square sampling, etc. are used to sample possible combinations, directions, and sizes of loads to form a load sample data set. The sampled loads are sequentially added to the finite element mesh model through programming methods.

[0038] S5: Finite element model solution; call the CAE solver to solve the finite element model with different constraint boundary conditions or load boundary conditions generated in step S4 (specifically, after inputting the constraint boundary conditions and load boundary conditions, a complete finite element model is obtained, the finite element model solution is started, the data is stored in the subsequent steps, and then the next input constraint boundary conditions and load boundary conditions are input, and the process is repeated in sequence); the CAE cloud platform is preferably used to perform calculations to solve the finite element model to give full play to the computing power and storage advantages of the CAE cloud platform.

[0039] S6: Obtain data such as equipment structure stress and deformation results to form a data set; establish a special data file, which includes constraint boundary conditions, load boundary conditions, stress, strain results and other data.

[0040] S7: Machine learning model training: The obtained data set consisting of constraint boundary conditions, load boundary conditions, etc. is divided into a training set and a test set. Different machine learning algorithms are used to train the model. The machine learning algorithm model with the highest accuracy in predicting the relationship between input and output is selected as the target model.

[0041] Specifically, when training a machine learning model, determine the proportion P1 of the training set in the data set, randomly select the training set from the data set according to the proportion P1, and use the rest of the data in the data set as the test set;

[0042] The portion of the training set containing constraint boundary conditions and load boundary condition data is classified as the test set data input, and the remaining data is used as the test set data output. The test set data input and output are normalized separately and mapped to the interval [0, 1] or [-1, 1] according to the data range. The obtained dataset consisting of constraint boundary conditions, load boundary conditions, etc. is divided into a training set and a test set. Different machine learning algorithms are used to train the model, and the model with the highest accuracy in predicting the relationship between input and output is selected as the target model.

[0043] Convolutional neural network, RNN neural network and BP neural network algorithms were selected for model training and verified with test set data. Finally, the relative standard deviation between the actual value and the predicted value was used to evaluate the model, and the model with high prediction accuracy was selected as the final target model.

[0044] S8: Outputs operational stress, deformation, and other data of the shield TBM equipment. Based on the target model obtained through machine learning, the load data and constraint boundary conditions during the equipment operation are used as input, loaded into the target model for calculation and output, and the operational stress, deformation, and other data of the shield TBM equipment are obtained and stored as a result output file.

[0045] Specifically, in step S8, the stress, deformation and other data outputs of the shield TBM equipment operating status are obtained in step S1, and the load data collected during the operation of the shield TBM or the data highly correlated with the load are converted into load data through specific operations, and then input into the target model for calculation after normalization, and the output is stress, deformation and other data (at this time, the data is the stress, deformation and other data before normalization). Then, the denormalization is carried out in combination with the normalization parameters of the same category of data, and the output data is restored to the real range, and the stress, deformation and other data of the shield TBM operating status are output.

[0046] S9: Data twin model visualization: 3D visualization of the stress, deformation and other data of the shield TBM equipment during operation, so as to intuitively obtain the stress and strain status of the equipment during operation;

[0047] Specifically, the obtained data on the operating status of the shield TBM, such as stress and deformation, are visualized through the ParaView open source program or programming.

[0048] The present invention provides a digital twin method for the operating status of shield TBM equipment based on the coupling of CAE simulation and machine learning: first, a digital model of the shield TBM equipment is established, and the boundary conditions for the operating status analysis of the shield TBM equipment are applied to obtain the combination mode of its operating status load, the direction of the load and the range of the load size; then, a sampling method is used to obtain several groups of working loads to establish corresponding CAE analysis models; the several groups of CAE analysis models are sequentially loaded into the CAE solver for solving, and the stress and deformation output data corresponding to the shield TBM equipment are obtained; a data set containing parameters such as load, stress, and strain is established; part of the data containing load, stress, and strain is selected from the data set as a training set; a machine learning method is used to train the model; the load data during the actual operation of the shield TBM equipment or several mechanism fusion data are used as load input; the equipment structure stress, deformation and other data are calculated and output through the trained model; finally, the stress, deformation and other data of the equipment operating status are visualized to realize the digital twin of the shield TBM equipment operation process. The above process solves the technical problems of existing technologies such as large computing power resources and long delays in the operation of shield TBM equipment, improves the ability to provide instant feedback on mechanical states such as stress and deformation during equipment operation, and has certain guiding significance for on-site construction activities of shield TBM.

[0049] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that improvements and modifications that do not depart from the principles of the present invention are within the scope of protection of the present invention.

Claims

1. A digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning, characterized by: The following steps are involved: S1: Establish a digital model of the shield TBM equipment. Use 3D modeling software to build a 3D model of the shield TBM equipment. Use meshing tools to mesh the 3D model of the shield TBM equipment to form a digital model of the shield TBM equipment. S2: Apply constraint boundary conditions; based on the structural characteristics of the shield TBM equipment and the load transfer path, apply the equipment operation process constraint boundary conditions on the digital model of the shield TBM equipment; S3: Establishing an equipment load set; analyzing the combination of various loads, the direction of each load, and the numerical range of the load magnitude during the operation of the shield TBM equipment to form a shield TBM equipment load set; S4: Extract load samples and apply them to the model; sample possible load combinations, directions, and magnitudes from the TBM equipment load set to form a load sample data set. Each load group in the sampled load sample data set is added to the digital model of the TBM equipment to form a finite element model with complete constraint boundary conditions and load boundary conditions. S5: finite element model solution; calling the CAE solver to solve the finite element model with different constraint boundary conditions or load boundary conditions generated in step S4; S6: Obtain data on equipment structure stress and deformation results to form a data set; create a dedicated data file, which includes but is not limited to constraint boundary conditions, load boundary conditions, stress, and strain results; S7: Machine learning model training: Divide the acquired data set into a training set and a test set. The data set includes but is not limited to constraint boundary conditions and load boundary conditions. Use different machine learning algorithms to train the model. Select the machine learning algorithm model with the highest accuracy in predicting the relationship between input and output as the target model. S8: Outputting shield TBM operating status data, including but not limited to stress and deformation; based on the target model obtained through machine learning, using load data and constraint boundary conditions during equipment operation as input, loading them into the target model for calculation and output, obtaining shield TBM equipment operating status data, and storing them as a result output file; the obtained shield TBM operating status data includes but is not limited to stress and deformation; S9: Data twin model visualization; the acquired shield TBM equipment operating status data is visualized in three dimensions to intuitively obtain the stress and strain status of the equipment during operation; the acquired shield TBM operating status data includes but is not limited to stress and deformation.

2. A digital twin method for the operation status of shield TBM equipment based on CAE simulation and machine learning coupling as claimed in claim 1, characterized in that the steps In S1, the shield TBM equipment drawings are used to construct a three-dimensional model. When conditions permit, the three-dimensional model is obtained from the equipment manufacturer and used directly. When the three-dimensional model file cannot be directly obtained from the equipment manufacturer, reverse engineering methods are used to obtain the three-dimensional model of the parts or assemblies of the shield TBM equipment. The finite element mesh discretization of the three-dimensional model of the shield TBM equipment is performed to obtain a finite element mesh model. Reverse engineering methods include but are not limited to three-dimensional laser scanning.

3. The digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning according to claim 1 is characterized in that: In step S2, the types of constraint boundary conditions and the locations of the constraints are analyzed according to the structural characteristics of the digital twin object, and the constraint boundary conditions are applied to the finite grid model.

4. The digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning according to claim 1 is characterized in that: In step S3, a shield TBM equipment load set is established based on the big data platform and mechanism; Specifically, the direction or theoretical value of a specific load is determined through mechanism knowledge, and the size range of specific loads of similar equipment is obtained by querying data collected by the big data platform.

5. The digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning according to claim 1 is characterized in that: In step S4, orthogonal array, Monte Carlo sampling, and Latin square sampling are used to sample possible combinations, directions, and magnitudes of loads to form a load sample data set. The sampled loads are sequentially added to the finite element mesh model through programming methods.

6. The digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning according to claim 1 is characterized in that: In step S5, the CAE cloud platform is used to perform calculations to solve the finite element model.

7. The digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning according to claim 1 is characterized in that: In step S7, when the machine learning model is trained, the proportion P1 of the training set in the data set is determined, and a training set is randomly selected from the data set according to the proportion P1, and the remaining data in the data set is used as a test set; The part of the training set containing constraint boundary conditions and load boundary condition data is classified as the test set data input part. The rest of the data is used as the test set data output; Normalize the test set data input and output separately, and map them to the interval [0, 1] or [-1, 1] depending on the data range. Divide the acquired data set into a training set and a test set. The data set composition includes but is not limited to constraint boundary conditions and load boundary conditions. Use different machine learning algorithms to train the model, and select the model with the highest accuracy in predicting the relationship between input and output as the target model. Convolutional neural network, RNN neural network and BP neural network algorithms were selected for model training and verified with test set data. Finally, the relative standard deviation between the actual value and the predicted value was used to evaluate the model, and the model with high prediction accuracy was selected as the final target model.

8. The digital twin method for shield TBM equipment operating status based on CAE simulation and machine learning coupling according to claim 1 is characterized in that: In step S8, the shield TBM operation status data output is obtained, and the obtained shield TBM operation status data includes but is not limited to stress and deformation. The load data collected during the operation of the shield TBM or data highly correlated with the load is converted into load data through specific operations, and input into the target model for calculation after normalization to output data. The output data includes but is not limited to stress and deformation. Then, denormalization is performed based on the parameters of normalization of data of the same category, and the output data is restored to the real range. The shield TBM operation status data is output, and the output shield TBM operation status data includes but is not limited to stress and deformation.

9. The digital twin method for the operation status of shield TBM equipment based on the coupling of CAE simulation and machine learning according to claim 1 is characterized in that: In step S9, the obtained shield TBM operating status data is visualized through the ParaView open source program or programming; the obtained shield TBM operating status data includes but is not limited to stress and deformation.

Images