Performance test method of foundation anti-seismic liquefaction system of long and large underground structure
By integrating multi-source data and machine learning algorithms, the convolutional neural network model is constructed, which solves the problem of low test accuracy of seismic liquefaction performance of underground structure foundations in the existing technology, and accurately predicts and prevents the foundation liquefaction performance, and optimizes the underground Structural design and construction ensure safety.
Patent Information
- Application Number
- CN202411280769.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-12
- Publication Date
- 2025-01-07
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the prior art, the seismic liquefaction performance test method of the foundation of the Grown underground structure relies on a single data source and empirical formula, has low prediction accuracy, lacks intuitive and reliable experimental methods, and mostly relies on numerical simulation and lacks experimental verification.
By integrating multi-source data such as geology, earthquake, hydrology, etc., using machine learning algorithms to build a multi-modal convolutional neural network model, combining experimental data for feature extraction and performance prediction, establishing a support vector machine model, and achieving accurate prediction of foundation liquefaction performance .
It provides intuitive and accurate performance testing methods for foundation seismic liquefaction system, realizes accurate prediction of foundation liquefaction performance, and can provide preventive measures and design optimization solutions for underground structures, reduce economic losses and ensure safety.
Smart Images
Figure CN119268984A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of civil engineering, and in particular to a performance testing method for a foundation anti-seismic liquefaction system of a long underground structure. Background Art
[0002] With the acceleration of urbanization, long underground structures such as subway tunnels, underground utility corridors, and large underground parking lots play an increasingly important role in urban infrastructure. However, these structures are susceptible to foundation liquefaction under earthquakes, resulting in decreased structural stability or even destruction. Therefore, accurate testing of the performance of the seismic liquefaction system of the foundation of long underground structures is of great significance to ensure structural safety.
[0003] At present, scholars at home and abroad have conducted extensive research and exploration on the seismic liquefaction problem of underground structures and proposed a variety of performance testing methods and evaluation standards. However, the existing technology still has the following deficiencies:
[0004] (1) Traditional performance testing methods rely on a single data source and empirical formulas, which have problems such as low prediction accuracy;
[0005] (2) The testing methods for the seismic liquefaction resistance of underground structure foundations mostly rely on numerical simulation and theoretical analysis, such as finite element simulation, and lack intuitive and reliable experimental means.
[0006] Therefore, it is necessary to propose a performance testing method for the foundation seismic liquefaction system of a long underground structure that is different from the prior art. Summary of the invention
[0007] The purpose of the present invention is to provide a performance testing method for the foundation seismic liquefaction system of a long underground structure, by integrating multi-source data such as geology, seismology, and hydrology, and using machine learning algorithms to mine the complex relationships between data, so as to achieve accurate prediction of foundation liquefaction performance.
[0008] To achieve the above-mentioned object, the present invention provides a performance test method for a foundation seismic liquefaction system of a long underground structure, comprising a test preparation stage, a test implementation stage and a data analysis and evaluation stage;
[0009] The test preparation phase includes the following steps:
[0010] a1. Construct a simulation test model based on the key indicators of the long underground structure to be tested;
[0011] a2. Arrange the foundation liquefaction monitoring system;
[0012] The trial implementation phase includes the following steps:
[0013] b1. Install accelerometer and displacement sensor on the simulation test model and connect the data collector;
[0014] b2. Use a shaking table to apply simulated seismic waves to the simulation test model, gradually increase the intensity and time of the simulated seismic waves and modify the source location to simulate different earthquake effects, record the response data of the long underground structure and the foundation response data, and calculate the liquefaction index of the foundation after the earthquake;
[0015] The analysis and evaluation phase includes the following steps:
[0016] c1. Preprocessing key indicators of long underground structures, underground structure response data, foundation response data, and simulated seismic wave data to obtain preprocessed data;
[0017] Preprocessing includes data cleaning, denoising, and normalization;
[0018] c2. Extract features from the preprocessed data based on a multimodal convolutional neural network model, then perform correlation analysis between the extracted features and the foundation liquefaction index, screen out features that are significantly correlated with the foundation liquefaction index, and concatenate the screened features to obtain fused features;
[0019] c3. Establish a performance test model for the SVM foundation anti-seismic liquefaction system, use the fused features and the calculated foundation liquefaction index as input, train the model, and improve the generalization ability of the model through evaluation indicators;
[0020] c4. Input the test data into the trained model and output the prediction results.
[0021] Preferably, in step a1, a simulation test model is constructed according to key indicators of the long underground structure to be tested, including the following sub-steps:
[0022] a11. Test the key indicators of the long underground structure to be tested;
[0023] Key indicators include underground structure data, foundation soil data, and hydrological data; underground structure data include underground structure size and material characteristics; foundation soil data include soil type, soil layer thickness and permeability; hydrological data include groundwater flow rate and groundwater level;
[0024] a12. Based on the key indicators of step a11, a scaled model of the long underground structure is constructed using 3D modeling software;
[0025] a13. Based on the scaled model constructed in step a12, a simulation test model is made using the same material as the long underground structure to be tested.
[0026] Preferably, in step a2, the foundation liquefaction monitoring system includes a vibration table, an accelerometer, a displacement sensor and a data collector.
[0027] Preferably, in step b2, the response data of the long underground structure includes deformation and cracks of the underground structure; and the foundation response data includes displacement, acceleration, liquefaction depth, liquefaction range and liquefaction duration.
[0028] Preferably, in step c1, the simulated seismic wave data includes seismic wave velocity, magnitude, frequency and source location.
[0029] Preferably, in step c2, feature extraction is performed on the preprocessed data based on a deep learning method, including the following sub-steps:
[0030] c21. Construct a one-dimensional convolutional neural network model to extract features of the preprocessed key indicators of underground structures, foundation response data, and simulated seismic wave data to obtain convolution features of key indicators of underground structures, convolution features of underground structure responses, and convolution features of simulated seismic waves;
[0031] c22. Construct a two-dimensional convolutional neural network model to extract features of the preprocessed underground structure response data to obtain underground structure response convolution features;
[0032] c23. Use the torch.cat function to concatenate the features extracted in step c21 and step c22 to obtain the fused multimodal convolutional features.
[0033] Preferably, in step c23, the selected features are spliced, and the specific operation is as follows:
[0034] Z = torch.cat(Z1,Z2,Z3,Z4)
[0035] Among them, Z is the fused feature; torch.cat is the torch.cat function; Z1 is the convolution feature of the key indicator of the underground structure; Z2 is the convolution feature of the underground structure response; Z3 is the convolution feature of the simulated seismic wave; Z4 is the convolution feature of the underground structure response.
[0036] Preferably, in step c21, the one-dimensional convolutional neural network model includes five one-dimensional convolutional modules, namely one-dimensional convolutional module one, one-dimensional convolutional module two, one-dimensional convolutional module three, one-dimensional convolutional module four, and one-dimensional convolutional module five;
[0037] Among them, the one-dimensional convolution module 1, the one-dimensional convolution module 2, and the one-dimensional convolution module 3 are respectively composed of a Conv1d, a BatchNorm1d, a ReLU, and a MaxPool1d. The number of convolution kernels is 4, 8, and 16 respectively, the convolution kernel size is 3×3, and the kernel size of MaxPool1d is 1×1;
[0038] The one-dimensional convolution module 4 consists of a Conv1d, BatchNorm1d, ReLU and AveragePool1d, where the number of convolution kernels is 32, the size of the convolution kernel is 3×3, and the size of the AveragePool1d kernel is 1×1;
[0039] The one-dimensional convolution module five consists of a Conv1d, BatchNorm1d, ReLU and AdaptiveAvgPool1d, where the number of convolution kernels is 64, the size of the convolution kernel is 3×3, and the size of the AdaptiveAvgPool1d kernel is 1×1.
[0040] Preferably, in step c22, the two-dimensional convolutional neural network model includes two two-dimensional convolution blocks, and the two two-dimensional convolution modules are respectively composed of a Conv2d, a BatchNorm2d, a ReLU and a MaxPool2d, wherein the number of convolution kernels is 32 and 64 respectively, the size of the convolution kernel is 3×3, and the kernel size of MaxPool2d is 1×1.
[0041] Preferably, in step c4, the evaluation index is the determination coefficient R 2 , mean square error MSE and root mean square error RMSE.
[0042] Therefore, the present invention adopts the above-mentioned performance testing method of the foundation seismic liquefaction system of a long underground structure, and the technical effects are as follows:
[0043] (1) The present invention provides an intuitive, accurate and reliable performance testing method for the foundation anti-seismic liquefaction system by constructing a simulation test model;
[0044] (2) The present invention integrates geological, seismic, hydrological and other multi-source data and uses machine learning algorithms to mine the complex relationships between data to achieve accurate prediction of foundation liquefaction performance;
[0045] (3) The present invention can provide preventive measures and guidance programs for existing long and large underground structures through performance testing. Preventive measures refer to the fact that potential problems of underground structures in terms of earthquake resistance and liquefaction can be discovered in time through performance testing, so that preventive measures can be taken in advance to avoid serious damage in natural disasters such as earthquakes; guidance programs refer to the fact that the test results provide a scientific basis for the design optimization, construction reinforcement and daily maintenance of underground structures, which helps to guide engineers and designers to formulate more reasonable and effective earthquake-resistant design programs;
[0046] (4) The present invention has good economic benefits. Through performance testing, the design and construction of underground structures can be optimized, and the economic losses and maintenance costs caused by insufficient earthquake resistance can be reduced;
[0047] (5) The present invention has good social benefits. Through performance testing, it improves the seismic performance of underground structures, helps to protect the safety of people's lives and property and the normal operation of cities, and has important social benefits. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 The present invention is a flow chart of a performance testing method of a foundation seismic liquefaction system for a long underground structure. DETAILED DESCRIPTION
[0049] In order to make the purpose, technical scheme and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention are further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions.
[0050] It should be noted that the terms "include", "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or devices.
[0051] Like reference numerals and letters denote similar items in the following drawings, and thus, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.
[0052] In the description of the present invention, it should be noted that the terms "upper", "lower", "inside", "outside", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, or are directions or positional relationships in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as a limitation on the present invention.
[0053] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "setting", "installation" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0054] Embodiment 1
[0055] like Figure 1 As shown, it is a flow chart of a performance testing method of a foundation seismic liquefaction system of a long underground structure of the present invention, which includes three stages in total: test preparation stage, test implementation stage and data analysis and evaluation stage.
[0056] The trial preparation phase includes the following steps:
[0057] a1. Construct a simulation test model based on the key indicators of the long underground structure to be tested, including the following sub-steps:
[0058] a11. Test the key indicators of the long underground structure to be tested;
[0059] The key indicators include underground structure data, foundation soil data, and hydrological data. In addition to the key indicators of large and long underground structures, the present invention may also consider the influence of ground factors, such as buildings, pedestrians, and vehicles above the large and long underground structures.
[0060] Underground structure data include the size of underground structures and the material characteristics of underground structures; foundation soil data include soil type, soil layer thickness and permeability, etc.; hydrological data include groundwater flow rate and groundwater level, etc.;
[0061] a12. Based on the key indicators of the long underground structure to be tested in step a11, a scaled model of the long underground structure is constructed using three-dimensional modeling software;
[0062] 3D modeling software can use GoCad, Geosoft, etc.
[0063] a13. Based on the scaled model constructed in step a12, a simulation test model is made using the same material as the long underground structure to be tested.
[0064] a2. Arrange the foundation liquefaction monitoring system;
[0065] The foundation liquefaction monitoring system includes a vibration table, an accelerometer, a displacement sensor and a data collector.
[0066] The trial implementation phase includes the following steps:
[0067] b1. Install accelerometer and displacement sensor on the simulation test model and connect the data collector;
[0068] The specific operations are as follows:
[0069] First, determine the layout of monitoring points (first determine the liquefaction risk area, select appropriate monitoring point locations in the liquefaction risk area, and the monitoring points need to be representative and able to reflect the overall situation of foundation liquefaction), the selection of monitoring equipment and the data transmission method. The selection of monitoring equipment and the data transmission method are determined according to the actual situation and are not specifically limited here;
[0070] Then install the monitoring equipment and connect the data logger;
[0071] Then debug the monitoring system and check the working status and data transmission of each device; if there is a problem with the monitoring system, make timely adjustments and optimizations; if there is no problem, perform subsequent operations.
[0072] b2. Use a shaking table to apply simulated seismic waves to the simulation test model, gradually increase the intensity and time of the simulated seismic waves and modify the source location to simulate different earthquake effects, record the response data of the long underground structure and the foundation response data, and calculate the liquefaction index of the foundation after the earthquake;
[0073] The response data of large and long underground structures include deformation and cracks of underground structures; the response data of foundations include displacement, acceleration, liquefaction depth, liquefaction range and liquefaction duration.
[0074] The deformation and cracks of underground structures can be photographed with a high-precision camera to obtain the original images of the deformation and cracks of underground structures, and then feature extraction can be performed through a subsequent two-dimensional convolutional neural network model;
[0075] The analysis and evaluation phase includes the following steps:
[0076] c1. Preprocessing key indicators of long underground structures, underground structure response data, foundation response data, and simulated seismic wave data to obtain preprocessed data;
[0077] The simulated seismic wave data include seismic wave velocity, magnitude, frequency and source location.
[0078] Preprocessing includes data cleaning, denoising, and normalization;
[0079] c2. Extract features from the preprocessed data based on a multimodal convolutional neural network model, then perform correlation analysis between the extracted features and the foundation liquefaction index, screen out features that are significantly correlated with the foundation liquefaction index, and concatenate the screened features to obtain fused features;
[0080] c21. Construct a one-dimensional convolutional neural network model to extract features of the preprocessed key indicators of underground structures, foundation response data, and simulated seismic wave data to obtain convolution features of key indicators of underground structures, convolution features of underground structure responses, and convolution features of simulated seismic waves;
[0081] The one-dimensional convolutional neural network model includes five one-dimensional convolution modules, namely one-dimensional convolution module one, one-dimensional convolution module two, one-dimensional convolution module three, one-dimensional convolution module four, and one-dimensional convolution module five;
[0082] Among them, the one-dimensional convolution module 1, the one-dimensional convolution module 2, and the one-dimensional convolution module 3 are respectively composed of a Conv1d, a BatchNorm1d, a ReLU, and a MaxPool1d. The number of convolution kernels is 4, 8, and 16 respectively, the convolution kernel size is 3×3, and the kernel size of MaxPool1d is 1×1;
[0083] The one-dimensional convolution module 4 consists of a Conv1d, BatchNorm1d, ReLU and AveragePool1d, where the number of convolution kernels is 32, the size of the convolution kernel is 3×3, and the size of the AveragePool1d kernel is 1×1;
[0084] The one-dimensional convolution module five consists of a Conv1d, BatchNorm1d, ReLU and AdaptiveAvgPool1d, where the number of convolution kernels is 64, the size of the convolution kernel is 3×3, and the size of the AdaptiveAvgPool1d kernel is 1×1.
[0085] c22. Construct a two-dimensional convolutional neural network model to extract features of the preprocessed underground structure response data to obtain underground structure response convolution features;
[0086] The two-dimensional convolutional neural network model includes two two-dimensional convolutional blocks. The two two-dimensional convolutional modules are composed of a Conv2d, a BatchNorm2d, a ReLU and a MaxPool2d respectively. The number of convolution kernels is 32 and 64 respectively, the size of the convolution kernel is 3×3, and the kernel size of MaxPool2d is 1×1.
[0087] c23, concatenate the features extracted in step c21 and step c22 through the torch.cat function to obtain the fused multimodal convolutional features;
[0088] The selected features are spliced. The specific operations are as follows:
[0089] Z = torch.cat(Z1,Z2,Z3,Z4)
[0090] Among them, Z is the fused feature; torch.cat is the torch.cat function; Z1 is the convolution feature of the key indicator of the underground structure; Z2 is the convolution feature of the underground structure response; Z3 is the convolution feature of the simulated seismic wave; Z4 is the convolution feature of the underground structure response.
[0091] c3. Establish a performance test model of the SVM (support vector machine) foundation anti-seismic liquefaction system, use the fused features and the calculated foundation liquefaction index as input, train the model, and improve the generalization ability of the model through evaluation indicators;
[0092] This step can also be modeled by other machine learning algorithms.
[0093] Evaluation indicators include the coefficient of determination R 2 , mean square error MSE and root mean square error RMSE.
[0094] c4. Input the test data into the trained model and output the prediction results.
[0095] After obtaining the test results, the staff can make advance rectifications to the problematic areas of the long and large underground structure based on the test results in order to cope with real earthquake disasters.
[0096] Through the above method, the performance of the seismic liquefaction system of the foundation of a large underground structure can be tested when an earthquake occurs, and which areas may liquefy can be predicted in advance. However, there is also a situation where the foundation does not liquefy when an earthquake occurs, but cracks and deformations appear in the underground structure. Therefore, the present invention also tests the response data of the underground structure. It has been proved through experiments that over time, cracks and deformations in the underground structure will also affect the liquefaction of the foundation. The reason for the analysis may be related to the topography of the underground structure.
[0097] It is worth noting that the contents not elaborated in detail in the present invention are all prior art and are well known to those skilled in the art.
[0098] Therefore, the present invention adopts the above-mentioned performance testing method of the foundation seismic liquefaction system of a long underground structure, integrates multi-source data such as geology, earthquake, and hydrology, and uses machine learning algorithms to mine the complex relationship between data to achieve accurate prediction of foundation liquefaction performance; in addition, the present invention also has good economic and social benefits, can optimize the design and construction of underground structures, reduce economic losses caused by insufficient seismic resistance, and protect people’s lives and property safety and the normal operation of the city.
[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.
Claims
1. A performance testing method for a seismic liquefaction system of a long underground structure, characterized in that: It includes the trial preparation stage, trial implementation stage and data analysis and evaluation stage; The test preparation phase includes the following steps: a1. Construct a simulation test model based on the key indicators of the long underground structure to be tested; a2. Arrange the foundation liquefaction monitoring system; The trial implementation phase includes the following steps: b1. Install accelerometer and displacement sensor on the simulation test model and connect the data collector; b2. Use a shaking table to apply simulated seismic waves to the simulation test model, gradually increase the intensity and time of the simulated seismic waves and modify the source location to simulate different earthquake effects, record the response data of the long underground structure and the foundation response data, and calculate the liquefaction index of the foundation after the earthquake; The analysis and evaluation phase includes the following steps: c1. Preprocessing key indicators of long underground structures, underground structure response data, foundation response data, and simulated seismic wave data to obtain preprocessed data; c2. Extract features from the preprocessed data based on a multimodal convolutional neural network model, then perform correlation analysis between the extracted features and the foundation liquefaction index, screen out features that are significantly correlated with the foundation liquefaction index, and concatenate the screened features to obtain fused features; c3. Establish a performance test model for the SVM foundation anti-seismic liquefaction system, use the fused features and the calculated foundation liquefaction index as input, train the model, and improve the generalization ability of the model through evaluation indicators; c4. Input the test data into the trained model and output the prediction results.
2. The performance testing method of a foundation seismic liquefaction system for a long underground structure according to claim 1, characterized in that: In step a1, a simulation test model is constructed according to key indicators of the long underground structure to be tested, including the following sub-steps: a11. Test the key indicators of the long underground structure to be tested; Key indicators include underground structure data, foundation soil data, and hydrological data; underground structure data include underground structure size and material characteristics; foundation soil data include soil type, soil layer thickness and permeability; hydrological data include groundwater flow rate and groundwater level; a12. Based on the key indicators of step a11, a scaled model of the long underground structure is constructed using 3D modeling software; a13. Based on the scaled model constructed in step a12, a simulation test model is made using the same material as the long underground structure to be tested.
3. The performance testing method of a foundation seismic liquefaction system for a long underground structure according to claim 2, characterized in that: In step a2, the foundation liquefaction monitoring system includes a shaking table, an accelerometer, a displacement sensor and a data collector.
4. The performance testing method of a foundation seismic liquefaction system for a long underground structure according to claim 3, characterized in that: In step b2, the response data of the large underground structure includes deformation and cracks of the underground structure; the foundation response data includes displacement, acceleration, liquefaction depth, liquefaction range and liquefaction duration.
5. The performance testing method of the seismic liquefaction system of the foundation of a long underground structure according to claim 4, characterized in that: In step c1, the simulated seismic wave data includes seismic wave velocity, magnitude, frequency and earthquake source location.
6. The performance testing method of a foundation seismic liquefaction system for a long underground structure according to claim 5, characterized in that: In step c2, feature extraction is performed on the preprocessed data based on a deep learning method, including the following sub-steps: c21. Construct a one-dimensional convolutional neural network model to extract features of the preprocessed key indicators of underground structures, foundation response data, and simulated seismic wave data to obtain convolution features of key indicators of underground structures, convolution features of underground structure responses, and convolution features of simulated seismic waves; c22. Construct a two-dimensional convolutional neural network model to extract features of the preprocessed underground structure response data to obtain underground structure response convolution features; c23. Use the torch.cat function to concatenate the features extracted in step c21 and step c22 to obtain the fused multimodal convolutional features.
7. The performance testing method of the seismic liquefaction system of the foundation of a long underground structure according to claim 6 is characterized in that: In step c23, the selected features are spliced. The specific operations are as follows: Z = torch.cat(Z1,Z2,Z3,Z4) Among them, Z is the fused feature; torch.cat is the torch.cat function; Z1 is the convolution feature of the key indicator of the underground structure; Z2 is the convolution feature of the underground structure response; Z3 is the convolution feature of the simulated seismic wave; Z4 is the convolution feature of the underground structure response.
8. The performance testing method of the seismic liquefaction system of the foundation of a long underground structure according to claim 7, characterized in that: In step c21, the one-dimensional convolutional neural network model includes five one-dimensional convolutional modules, namely, one-dimensional convolutional module one, one-dimensional convolutional module two, one-dimensional convolutional module three, one-dimensional convolutional module four, and one-dimensional convolutional module five; Among them, the one-dimensional convolution module 1, the one-dimensional convolution module 2, and the one-dimensional convolution module 3 are respectively composed of a Conv1d, a BatchNorm1d, a ReLU, and a MaxPool1d. The number of convolution kernels is 4, 8, and 16 respectively, the convolution kernel size is 3×3, and the kernel size of MaxPool1d is 1×1; The one-dimensional convolution module 4 consists of a Conv1d, BatchNorm1d, ReLU and AveragePool1d, where the number of convolution kernels is 32, the size of the convolution kernel is 3×3, and the size of the AveragePool1d kernel is 1×1; The one-dimensional convolution module five consists of a Conv1d, BatchNorm1d, ReLU and AdaptiveAvgPool1d, where the number of convolution kernels is 64, the size of the convolution kernel is 3×3, and the size of the AdaptiveAvgPool1d kernel is 1×1.
9. The performance testing method of the seismic liquefaction system of the foundation of a long underground structure according to claim 8, characterized in that: In step c22, the two-dimensional convolutional neural network model includes two two-dimensional convolution blocks, and the two two-dimensional convolution modules are respectively composed of a Conv2d, a BatchNorm2d, a ReLU and a MaxPool2d, wherein the number of convolution kernels is 32 and 64 respectively, the size of the convolution kernel is 3×3, and the kernel size of MaxPool2d is 1×1.
10. The performance testing method of the seismic liquefaction system of the foundation of a long underground structure according to claim 9, characterized in that: In step c4, the evaluation index is the determination coefficient R 2 , mean square error MSE and root mean square error RMSE.