A method and system for calculating damage performance of submarine tunnel structures based on deep learning

Abstract

The present invention provides a method and system for calculating the damage performance of an undersea tunnel structure based on deep learning, and relates to the technical field of tunnel damage performance calculation. The present invention uses a convolutional neural network in combination with a long short-term memory network to perform deep learning model training on collected data, thereby realizing the fusion of multi-source data and the extraction of time series features; by calculating the average change trend coefficient and the real-time change trend coefficient of sediment density change data in past external microseismic activities, and performing comparative analysis, an impact adjustment coefficient is generated, and finally a first evaluation index for evaluating the structural damage performance status of each immersed tube connection is generated; the accuracy and efficiency of damage assessment are improved, and the system also has the ability of real-time monitoring and rapid response, thereby providing strong support for the structural health monitoring and maintenance of undersea tunnels.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel damage performance calculation, and specifically to a method and system for calculating damage performance of an undersea tunnel structure based on deep learning. Background Art

[0002] Subsea tunnels are located in a complex environment and have long been affected by multiple factors such as seawater erosion, geological movement, and human activities. Structural health monitoring and damage assessment have become urgent issues to be addressed. Traditional subsea tunnel structural damage assessment methods rely on regular manual inspections and instrument monitoring, which have many limitations such as long time spans, high costs, and the need to stop operations for inspections. In addition, the detection results are often delayed, making it difficult to reflect the true condition of the tunnel structure in a timely manner. In recent years, with the rapid development of deep learning technology, the use of artificial intelligence technology to conduct real-time monitoring and damage assessment of subsea tunnel structures has become a hot research direction.

[0003] In the prior art, the publication number is CN115840984B, and the name is a method, device and terminal device for calculating the damage performance of a tunnel structure. The above-mentioned method for calculating the damage performance of a tunnel structure includes: obtaining the basic parameters of the tunnel to be calculated, and constructing a tunnel structure model based on the basic parameters of the tunnel to be calculated; determining the seismic wave simulation parameters based on the basic parameters of the tunnel to be calculated, and generating simulated seismic waves according to the seismic wave simulation parameters; applying the simulated seismic waves to the tunnel structure model to obtain structural damage performance data; calculating the tunnel structure damage performance based on the structural damage performance data. This method takes into account the tensile damage problem of the structure at different stages, and realizes a method for simply and accurately calculating the damage performance of the tunnel structure under earthquake action.

[0004] Existing damage monitoring of submarine tunnel structures mainly relies on sensor networks and finite element analysis. Although these methods have improved the accuracy and efficiency of monitoring to a certain extent, they still have some shortcomings: first, the deployment and maintenance costs of sensor networks are high, and in complex environments such as submarine tunnels, sensor failures and data loss are common, affecting the continuity of monitoring and data quality; second, the finite element analysis method is highly dependent on the model, requires a lot of prior knowledge and manual intervention, and has high computational complexity, making it difficult to achieve real-time monitoring and rapid response. In addition, existing methods have limitations in processing multi-source data fusion and time series feature extraction, making it difficult to accurately capture the long-term impact of microseismic activity on tunnel structures. Therefore, there is an urgent need for a method for calculating the damage performance of submarine tunnel structures that can efficiently process complex multi-source data and has real-time monitoring capabilities.

[0005] The above information disclosed in the above Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for calculating the damage performance of an undersea tunnel structure based on deep learning, so as to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for calculating damage performance of a submarine tunnel structure based on deep learning is applied to a scaled submarine tunnel structure model specimen of a physical submarine tunnel structure. The submarine tunnel structure model specimen is provided with a simulated seabed area. The specific steps include:

[0009] Step S1: obtaining historical microseismic activity data of the physical submarine tunnel structure that has experienced e1 microseismic activities in the past, and sediment density change data of the selected seabed area above each immersed tube connection;

[0010] Mark the immersed tube connections of the submarine tunnel structure model specimen to form an immersed tube connection sequence set;

[0011] And conduct M times of incremental microseismic activity simulation experiments on the submarine tunnel structure model specimen, M>e1, and monitor the sediment density change data of the selected seabed area above each immersed tube connection in real time in the incremental microseismic activity data of each immersed tube connection;

[0012] Step S2: setting the time period before and after the submarine tunnel structure model specimen is subjected to external microseismic activity as the abnormal monitoring time period, and collecting key area monitoring data of each immersed tube connection within the abnormal monitoring time period;

[0013] Step S3: A deep learning model is constructed based on a convolutional neural network (CNN) and a long short-term memory (LSTM) network, and the microseismic activity data and sediment density change data of each microseismic activity simulation experiment are used as input features, and the key area monitoring data are used as output labels to train the deep learning model;

[0014] The convolutional neural network (CNN) is used to extract the features of microseismic activity data, sediment density change data, and key area monitoring data. The extracted features of microseismic activity data and sediment density change data are used as input, and the features of key area monitoring data are used as output labels to train the long short-term memory network (LSTM).

[0015] Step S4: if the physical submarine tunnel structure is subjected to external microseismic activity at the current moment, the historical microseismic activity data and sediment density change data of the physical submarine tunnel structure collected in step S1 are input into the trained deep learning model for calibration;

[0016] The real-time monitored microseismic activity data and sediment density change data are input into the calibrated deep learning model to obtain the prediction results of the monitoring data of each key area in the immersed tube connection sequence set;

[0017] Step S5: Calculate the average change trend coefficient of the sediment density change data in step S1 during the past e1 external microseismic activities, and calculate the real-time change trend coefficient of the sediment density change data during the current and last external microseismic activities, compare and analyze the real-time change trend coefficient with the average change trend coefficient, and generate an impact adjustment coefficient;

[0018] Step S6: The prediction results of the monitoring data of each key area are combined with the impact adjustment coefficient for analysis to generate a first evaluation index for evaluating the structural damage performance status of each immersed tube connection.

[0019] A system for calculating damage performance of an undersea tunnel structure based on deep learning, the system being used to execute the method for calculating damage performance of an undersea tunnel structure based on deep learning, comprising:

[0020] Marking module: obtain historical microseismic activity data of the physical submarine tunnel structure that has experienced e1 microseismic activities in the past, as well as sediment density change data of the selected seabed area above each immersed tube connection;

[0021] Mark the immersed tube connections of the submarine tunnel structure model specimen to form an immersed tube connection sequence set;

[0022] And conduct M times of incremental microseismic activity simulation experiments on the submarine tunnel structure model specimen, M>e1, and monitor the sediment density change data of the selected seabed area above each immersed tube connection in real time in the incremental microseismic activity data of each immersed tube connection;

[0023] Key area monitoring data acquisition module: used to set the time period before and after the submarine tunnel structure model specimen is subjected to external microseismic activity as the abnormal monitoring time period, and collect key area monitoring data of each immersed tube connection within the abnormal monitoring time period;

[0024] Deep learning module: A deep learning model is constructed based on convolutional neural network (CNN) and long short-term memory (LSTM) networks. The microseismic activity data and sediment density change data of each microseismic activity simulation experiment are used as input features, and the key area monitoring data are used as output labels to train the deep learning model.

[0025] The convolutional neural network (CNN) is used to extract the features of microseismic activity data, sediment density change data, and key area monitoring data. The extracted features of microseismic activity data and sediment density change data are used as input, and the features of key area monitoring data are used as output labels to train the long short-term memory network (LSTM).

[0026] Prediction result acquisition module: when the physical submarine tunnel structure is subjected to external microseismic activity at the current moment, the collected historical microseismic activity data and sediment density change data of the physical submarine tunnel structure are input into the trained deep learning model for calibration;

[0027] The real-time monitored microseismic activity data and sediment density change data are input into the calibrated deep learning model to obtain the prediction results of the monitoring data of each key area in the immersed tube connection sequence set;

[0028] Impact adjustment coefficient generation module: used to calculate the average change trend coefficient of sediment density change data in the past e1 external microseismic activities, and calculate the real-time change trend coefficient of sediment density change data during the current and last external microseismic activities, compare and analyze the real-time change trend coefficient with the average change trend coefficient, and generate the impact adjustment coefficient;

[0029] The first evaluation index generation module is used to combine the prediction results of the monitoring data of each key area with the impact adjustment coefficient for analysis, and generate the first evaluation index for evaluating the structural damage performance status of each immersed tube connection.

[0030] Compared with the prior art, the invention has the following beneficial effects: by marking the immersed tube connection of the submarine tunnel structure model specimen, an immersed tube connection sequence set is formed, and the incremental microseismic activity data and sediment density change data of each immersed tube connection are monitored in real time, and a deep learning model is trained on the collected data using a convolutional neural network combined with a long short-term memory network, thereby realizing the fusion of multi-source data and the extraction of time series features;

[0031] By calculating the average change trend coefficient and real-time change trend coefficient of sediment density change data in the past external microseismic activities, and conducting comparative analysis, the impact adjustment coefficient is generated, and finally the first evaluation index is generated to evaluate the structural damage performance status of each immersed tube connection; the accuracy and efficiency of damage assessment are improved, and the ability of real-time monitoring and rapid response is also possessed, which provides strong support for the structural health monitoring and maintenance of undersea tunnels. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0033] Figure 2This is a block diagram of the overall system module of the present invention. DETAILED DESCRIPTION

[0034] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0035] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0036] Embodiment 1:

[0037] See also Figure 1 , the present invention provides a technical solution:

[0038] A method for calculating damage performance of submarine tunnel structures based on deep learning, the specific steps include:

[0039] Step S1: obtaining historical microseismic activity data of the physical submarine tunnel structure that has experienced e1 microseismic activities in the past, and sediment density change data of the selected seabed area above each immersed tube connection;

[0040] Mark the immersed tube connections of the submarine tunnel structure model specimen to form an immersed tube connection sequence set;

[0041] And conduct M times of incremental microseismic activity simulation experiments on the submarine tunnel structure model specimen, M>e1, and monitor the sediment density change data of the selected seabed area above each immersed tube connection in real time in the incremental microseismic activity data of each immersed tube connection;

[0042] Step S2: setting the time period before and after the submarine tunnel structure model specimen is subjected to external microseismic activity as the abnormal monitoring time period, and collecting key area monitoring data of each immersed tube connection within the abnormal monitoring time period;

[0043] Step S3: A deep learning model is constructed based on a convolutional neural network (CNN) and a long short-term memory (LSTM) network, and the microseismic activity data and sediment density change data of each microseismic activity simulation experiment are used as input features, and the key area monitoring data are used as output labels to train the deep learning model;

[0044] The convolutional neural network (CNN) is used to extract the features of microseismic activity data, sediment density change data, and key area monitoring data. The extracted features of microseismic activity data and sediment density change data are used as input, and the features of key area monitoring data are used as output labels to train the long short-term memory network (LSTM).

[0045] Step S4: if the physical submarine tunnel structure is subjected to external microseismic activity at the current moment, the historical microseismic activity data and sediment density change data of the physical submarine tunnel structure collected in step S1 are input into the trained deep learning model for calibration;

[0046] The real-time monitored microseismic activity data and sediment density change data are input into the calibrated deep learning model to obtain the prediction results of the monitoring data of each key area in the immersed tube connection sequence set;

[0047] Step S5: Calculate the average change trend coefficient of the sediment density change data in step S1 during the past e1 external microseismic activities, and calculate the real-time change trend coefficient of the sediment density change data during the current and last external microseismic activities, compare and analyze the real-time change trend coefficient with the average change trend coefficient, and generate an impact adjustment coefficient;

[0048] Step S6: The prediction results of the monitoring data of each key area are combined with the impact adjustment coefficient for analysis to generate a first evaluation index for evaluating the structural damage performance status of each immersed tube connection.

[0049] It is further explained that the submarine tunnel structure model specimen is composed of several scaled immersed tube tunnel structure model specimens, and the submarine tunnel structure damage performance calculation is carried out;

[0050] Test piece design:

[0051] A 1:4 scaled immersed tube tunnel structure model specimen was used.

[0052] The concrete strength grade is LC50, and the steel bar is HPB300. During the casting process of each specimen, concrete is collected to make 3 compressive strength test cube blocks, 3 tensile strength test blocks, and 3 concrete elastic modulus test blocks, which are cured for 28 days under the same conditions as the immersed tube tunnel specimens.

[0053] According to standard test methods, the concrete test blocks are subjected to horizontal and vertical load tests under different external microseismic activities, and the microseismic activity data of each immersed tube connection area and the sediment density change data of the selected seabed area above the immersed tube connection are monitored in real time;

[0054] Test device and loading system:

[0055] The foundation beam is anchored to the test bench by ground anchor bolts. The horizontal actuator is connected to the immersed tube tunnel to apply horizontal load. The vertical load is applied by a hydraulic jack. A roller is set between the hydraulic jack and the loading beam to ensure the normal movement of the immersed tube tunnel during horizontal reciprocation.

[0056] The horizontal load is applied by load and displacement. The load is controlled before the specimen yields. Each loading is 1 / 5 to 1 / 10 of the expected ultimate load of the specimen. At the initial loading stage, 20k is cycled once, and the load is increased step by step. In the later stage, 40kN is cycled once until the specimen yields. After the specimen yields, the displacement is controlled. The displacement applied is a multiple of the yield displacement △y of the specimen. Each stage is 1 to 3 times the yield displacement. Loading is performed step by step until the structure is completely destroyed or the bearing capacity drops to 85% of the ultimate load. The test ends.

[0057] Measuring point arrangement and measurement plan:

[0058] The test measurements include horizontal displacement of the specimen, concrete strain, etc. The displacement is measured by laser displacement meters placed at the top and bottom of the specimen, and the concrete strain is measured by embedded resistance strain gauges. The measuring points are arranged symmetrically on both sides of the wall, with 10 points on each side.

[0059] Monitoring equipment for additional environmental parameters: Microseismic sensors: deployed in the tunnel area to monitor the frequency and intensity of microseismic activity.

[0060] Sediment density meter: installed around the model to monitor the changes in sediment density in real time.

[0061] The position data of each immersed tube connection is recorded by using sensors to form an immersed tube connection sequence set;

[0062] The microseismic activity data include different combinations of horizontal and vertical loads, and each combination represents one experimental number;

[0063] In each microseismic experiment, the combination of different horizontal and vertical loads is expressed as:

[0064] Horizontal load: ;

[0065] Vertical load: ;

[0066] Where N and K represent the number of different load conditions for horizontal and vertical loads, respectively; each experiment, i.e., the jth microseismic activity, uses the combination express, Indicates The index of the horizontal load; Indicates The index of the vertical load; , , the total number of experiments ;

[0067] The number of microseismic experiments is represented as {1, 2, ..., j, ..., M}, where j represents the index of the jth microseismic experiment and M represents the total number of experiments. The abnormal monitoring time period of the jth microseismic experiment is set as ,and ;in, is the time of the jth microseismic activity experiment The time period before and after;

[0068] The immersed tube connection sequence set is represented as {1, 2, ..., i, ..., n}, where i represents the index of the i-th immersed tube connection and n represents the total number of immersed tube connections.

[0069] Various types of sensors are arranged at each immersed tube connection and the selected seabed area above it, including microseismic sensors and sediment density sensors; the microseismic sensors are used to monitor the microseismic activity data in the area, and the sediment density sensors are used to monitor the sediment density change data;

[0070] Sensors are used to collect real-time microseismic activity data and sediment density change data in the area where each immersed tube connection is located; the data collection frequency is once per second.

[0071] It is further explained that the key area monitoring data includes the stress change data and displacement change data of the immersed tube joint during the abnormal monitoring period, as well as the water pressure change data of the waterproof structure on the current immersed tube connection; these collected data are transmitted to the data processing unit for unified standardized preprocessing, specifically using the minimum-maximum normalization Min-MaxNormalization; the standardized stress change data, displacement change data, water pressure change data and sediment density change data of the i-th immersed tube connection during the j-th microseismic activity are marked as ;

[0072] The monitoring range of the seabed area includes a depth of d1 extending from the simulated seabed surface to capture the changes in sediment density within the entire S1 area.

[0073] The stress change data, displacement change data, water pressure change data and sediment density change data of each immersed tube connection are set as average values. Specifically, at least three monitoring points are set at the top and both sides of each immersed tube connection. The stress change, displacement change, water pressure change and sediment density change of the three monitoring points are monitored respectively, and the monitored data are averaged to obtain the following , , , , , , ;

[0074] Stress variation data of the i-th immersed tube connection during the j-th microseismic activity The calculation formula is as follows:

[0075] ;

[0076] in, Indicates the abnormal monitoring time period corresponding to the jth microseismic activity at the ith immersed tube connection The stress value of

[0077] Displacement change data of the i-th immersed tube connection during the j-th microseismic activity The calculation formula is as follows:

[0078] ;

[0079] in, represents the time after the jth microseismic activity The horizontal displacement value of represents the time before the jth microseismic activity The horizontal displacement value of

[0080] represents the time after the jth microseismic activity The vertical displacement value of represents the time before the jth microseismic activity The vertical displacement value of

[0081] Water pressure variation data at the i-th immersed tube connection during the j-th microseismic activity The calculation formula is as follows:

[0082] ;

[0083] in, Indicates the abnormal monitoring time period corresponding to the jth microseismic activity at the ith immersed tube connection Water pressure value;

[0084] Sediment density variation data at the i-th immersed tube connection during the j-th microseismic activity The calculation formula is as follows:

[0085] ;

[0086] in, Indicates the abnormal monitoring time period corresponding to the jth microseismic activity at the ith immersed tube connection The sediment density value.

[0087] Further explanation: CNN is used to extract the features of microseismic activity data, sediment density change data, and key area monitoring data, and LSTM is used to analyze the changes of these features over time, including:

[0088] 1.1) Data segmentation: The standardized key area monitoring data, microseismic activity data, and corresponding sediment density change data are segmented into a training set (70%), a validation set (15%), and a test set (15%) to ensure the generalization ability of the model;

[0089] 1.2) Use the convolutional neural network CNN to extract features from the segmented data and obtain the extracted feature number f;

[0090] The model architecture of CNN is as follows:

[0091] Input layer: The input data dimension is (n, t, c), where n is the number of immersed tube connections, t is the time step, and c is the number of features;

[0092] Convolutional layer 1: 32 3×3 convolution kernels, ReLU activation function;

[0093] Pooling layer 1: 2×2 maximum pooling;

[0094] Convolutional layer 2: 64 3×3 convolution kernels, ReLU activation function;

[0095] Pooling layer 2: 2×2 maximum pooling;

[0096] Convolutional layer 3: 128 3×3 convolution kernels, ReLU activation function;

[0097] Pooling layer 3: 2×2 maximum pooling;

[0098] Fully connected layer: 256 neurons, ReLU activation function;

[0099] Output layer: The feature dimension is (n, t, f), where f is the number of extracted features, and a linear activation function is used;

[0100] 1.3) Use the long short-term memory network LSTM to perform time series analysis on the extracted feature number f;

[0101] The LSTM model architecture is as follows:

[0102] LSTM layer 1: 64 memory units, input is the f feature extracted by CNN;

[0103] LSTM layer 2: 64 memory units, output is time series features;

[0104] Fully connected layer: The output dimension is (n, t, m), where m is the number of monitoring data, and a linear activation function is used;

[0105] 1.4) Model training;

[0106] Loss function: mean square error (MSE);

[0107] Optimizer: Adam optimizer with a learning rate of 0.001;

[0108] Training parameters: the number of training rounds is 100 rounds, and the batch size is 32;

[0109] Early stopping mechanism: monitor the validation error. If the validation error does not decrease for 10 consecutive rounds, stop training early.

[0110] Further explanation: if the physical submarine tunnel structure is subjected to external microseismic activity at the current moment, the real-time monitored microseismic activity data and sediment density change data are input into the calibrated deep learning model to obtain the prediction results of the monitoring data of each key area in the immersed tube connection sequence set, including:

[0111] Using the same standardized parameters used in training, the real-time acquisition of microseismic activity data and sediment density change data of the physical submarine tunnel structure subjected to microseismic activity is standardized;

[0112] The standardized real-time data is input into a calibrated convolutional neural network (CNN) to extract the features of microseismic activity data and sediment density change data;

[0113] The extracted features are input into the calibrated long short-term memory network LSTM model to perform time series prediction to obtain the prediction results of the monitoring data of each key area; the specific acquisition of the prediction results is as follows:

[0114] Combine the horizontal load, vertical load and microseismic activity of the physical submarine tunnel structure at the current moment Perform matching analysis to find the closest j+u-th combination experiment, where j+u∈{1,2,…,j,…,M};

[0115] The sediment density change data of the j+uth combined experiment is recorded as ;

[0116] Output the prediction results of the key area monitoring data corresponding to each immersed tube connection at the current moment. The output value of the prediction result is limited to the range of (0,1]. The prediction result includes the stress change data of the ith immersed tube connection at the j+uth microseismic activity. , displacement change data , water pressure change data .

[0117] Further explanation: generating the impact adjustment coefficient includes:

[0118] Before calculating the j+uth combined experiment, the average change trend coefficient of the sediment density change data of the physical submarine tunnel structure in the past e1 microseismic activities is calculated in advance, which is recorded as , the formula is:

[0119] ;

[0120] in, and They are the sediment density variation data of the i-th immersed tube connection during the j+uk and j+uk-1 microseismic activities;

[0121] Calculate the real-time trend coefficient of sediment density change data during the current and previous external microseismic activity , the formula is:

[0122] ;

[0123] in, is the sediment density change data of the i-th immersed tube connection at the j+u-th microseismic activity at the current moment, is the sediment density change data of the i-th immersed tube connection during the last j+u-1 microseismic activity;

[0124] Compare and analyze the real-time change trend coefficient with the average change trend coefficient to generate the impact adjustment coefficient , the formula is:

[0125] ;

[0126] in, Represents the impact adjustment coefficient; exp() is an exponential function;

[0127] The value range of is (0,1]; when and As we get closer, When it is closer to 1, it means that the current sediment density change trend is more consistent with the historical average trend; on the contrary, when and The greater the difference, The closer it is to 0, the greater the difference between the current sediment density trend and the historical average trend;

[0128] Setting the impact adjustment factor The judgment threshold is ;

[0129] like The smaller the ,and When , it means that the sediment density change caused by the current microseismic activity is more drastic than the historical average, and the impact on the physical submarine tunnel structure is greater;

[0130] like The smaller the ,and , it means that the sediment density changes caused by the current microseismic activity are more moderate than the historical average, and the impact on the physical submarine tunnel structure is smaller.

[0131] Further, generating a first evaluation index for evaluating the structural damage performance status of each immersed tube connection specifically includes:

[0132] The first evaluation index of the i-th immersed tube connection is recorded as , the formula is:

[0133] ;

[0134] in, represents the first evaluation index of the ith immersed tube connection; represents the impact adjustment coefficient;

[0135] The value range is (0,1]; based on the expert group system and experimental data analysis, the first evaluation index is set The judgment threshold is ; and stress change data , displacement change data , water pressure change data The warning thresholds are ;

[0136] in,

[0137] when When, and , , When any value in exceeds the corresponding warning threshold, the structural damage performance status of the i-th immersed tube connection is level one damage; level one damage means that the damage degree of the i-th immersed tube connection is less than 50%;

[0138] Prediction results include stress change data , displacement change data , water pressure change data The maximum value in exceeds { } is within 50% of the corresponding warning threshold;

[0139] when When, and , , When any value exceeds the corresponding warning threshold, the structural damage performance status of the i-th immersed tube connection is level 2 damage; level 2 damage means that the damage degree of the i-th immersed tube connection is between 50% and 70%;

[0140] Prediction results include stress change data , displacement change data , water pressure change data The maximum value in exceeds { } corresponds to 50%-70% of the warning threshold;

[0141] when , , ,and , , When any value exceeds the corresponding warning threshold, the structural damage performance status of the i-th immersed tube connection is level 3 damage; level 3 damage means that the degree of damage of the i-th immersed tube connection is between 70% and 100%.

[0142] Prediction results include stress change data , displacement change data , water pressure change data The maximum value in exceeds { } corresponds to more than 70% of the warning threshold.

[0143] Embodiment 2:

[0144] See also Figure 2 , a system for calculating damage performance of an undersea tunnel structure based on deep learning, the system being used to execute the method for calculating damage performance of an undersea tunnel structure based on deep learning, comprising:

[0145] Marking module: obtain historical microseismic activity data of the physical submarine tunnel structure that has experienced e1 microseismic activities in the past, as well as sediment density change data of the selected seabed area above each immersed tube connection;

[0146] Mark the immersed tube connections of the submarine tunnel structure model specimen to form an immersed tube connection sequence set;

[0147] And conduct M times of incremental microseismic activity simulation experiments on the submarine tunnel structure model specimen, M>e1, and monitor the sediment density change data of the selected seabed area above each immersed tube connection in real time in the incremental microseismic activity data of each immersed tube connection;

[0148] Key area monitoring data acquisition module: used to set the time period before and after the submarine tunnel structure model specimen is subjected to external microseismic activity as the abnormal monitoring time period, and collect key area monitoring data of each immersed tube connection within the abnormal monitoring time period;

[0149] Deep learning module: A deep learning model is constructed based on convolutional neural network (CNN) and long short-term memory (LSTM) networks. The microseismic activity data and sediment density change data of each microseismic activity simulation experiment are used as input features, and the key area monitoring data are used as output labels to train the deep learning model.

[0150] The convolutional neural network (CNN) is used to extract the features of microseismic activity data, sediment density change data, and key area monitoring data. The extracted features of microseismic activity data and sediment density change data are used as input, and the features of key area monitoring data are used as output labels to train the long short-term memory network (LSTM).

[0151] Prediction result acquisition module: when the physical submarine tunnel structure is subjected to external microseismic activity at the current moment, the collected historical microseismic activity data and sediment density change data of the physical submarine tunnel structure are input into the trained deep learning model for calibration;

[0152] The real-time monitored microseismic activity data and sediment density change data are input into the calibrated deep learning model to obtain the prediction results of the monitoring data of each key area in the immersed tube connection sequence set;

[0153] Impact adjustment coefficient generation module: used to calculate the average change trend coefficient of sediment density change data in the past e1 external microseismic activities, and calculate the real-time change trend coefficient of sediment density change data during the current and last external microseismic activities, compare and analyze the real-time change trend coefficient with the average change trend coefficient, and generate the impact adjustment coefficient;

[0154] The first evaluation index generation module is used to combine the prediction results of the monitoring data of each key area with the impact adjustment coefficient for analysis, and generate the first evaluation index for evaluating the structural damage performance status of each immersed tube connection.

[0155] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0156] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0157] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0158] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for calculating damage performance of an undersea tunnel structure based on deep learning, applied to a scaled undersea tunnel structure model specimen of an actual undersea tunnel structure, wherein the undersea tunnel structure model specimen is provided with a simulated seabed area, characterized in that: The specific steps include: Step S1: obtaining historical microseismic activity data of the physical submarine tunnel structure that has experienced e1 microseismic activities in the past, and sediment density change data of the selected seabed area above each immersed tube connection; Mark the immersed tube connections of the submarine tunnel structure model specimen to form an immersed tube connection sequence set; And conduct M times of incremental microseismic activity simulation experiments on the submarine tunnel structure model specimen, M>e1, and monitor the sediment density change data of the selected seabed area above each immersed tube connection in real time in the incremental microseismic activity data of each immersed tube connection; Step S2: setting the time period before and after the submarine tunnel structure model specimen is subjected to external microseismic activity as the abnormal monitoring time period, and collecting key area monitoring data of each immersed tube connection within the abnormal monitoring time period; Step S3: A deep learning model is constructed based on a convolutional neural network (CNN) and a long short-term memory (LSTM) network, and the microseismic activity data and sediment density change data of each microseismic activity simulation experiment are used as input features, and the key area monitoring data are used as output labels to train the deep learning model; The convolutional neural network (CNN) is used to extract the features of microseismic activity data, sediment density change data, and key area monitoring data. The extracted features of microseismic activity data and sediment density change data are used as input, and the features of key area monitoring data are used as output labels to train the long short-term memory network (LSTM). Step S4: if the physical submarine tunnel structure is subjected to external microseismic activity at the current moment, the historical microseismic activity data and sediment density change data of the physical submarine tunnel structure collected in step S1 are input into the trained deep learning model for calibration; The real-time monitored microseismic activity data and sediment density change data are input into the calibrated deep learning model to obtain the prediction results of the monitoring data of each key area in the immersed tube connection sequence set; Step S5: Calculate the average change trend coefficient of the sediment density change data in step S1 during the past e1 external microseismic activities, and calculate the real-time change trend coefficient of the sediment density change data during the current and last external microseismic activities, compare and analyze the real-time change trend coefficient with the average change trend coefficient, and generate an impact adjustment coefficient; Step S6: The prediction results of the monitoring data of each key area are combined with the impact adjustment coefficient for analysis to generate a first evaluation index for evaluating the structural damage performance status of each immersed tube connection.

2. The method for calculating damage performance of submarine tunnel structure based on deep learning according to claim 1 is characterized in that: The submarine tunnel structure model specimen is composed of several scaled immersed tube tunnel structure model specimens, and the submarine tunnel structure damage performance calculation is carried out; The microseismic activity data include different combinations of horizontal and vertical loads, and each combination represents one experimental number; In each microseismic experiment, the combination of different horizontal and vertical loads is expressed as: Horizontal load: ; Vertical load: ; Where N and K represent the number of different load conditions for horizontal and vertical loads, respectively; each experiment, i.e., the jth microseismic activity, uses the combination express, Indicates The index of the horizontal load; Indicates the index of the vertical load; and , , the total number of experiments ; The number of microseismic experiments is represented as {1, 2, ..., j, ..., M}, where j represents the index of the jth microseismic experiment and M represents the total number of experiments. The abnormal monitoring time period of the jth microseismic experiment is set as ,and ;in, is the time of the jth microseismic activity experiment The time period before and after; The immersed tube connection sequence set is represented as {1, 2, ..., i, ..., n}, where i represents the index of the i-th immersed tube connection and n represents the total number of immersed tube connections.

3. The method for calculating damage performance of submarine tunnel structure based on deep learning according to claim 2 is characterized in that: The key area monitoring data include stress change data and displacement change data of the immersed tube joint during the abnormal monitoring period, as well as the water pressure change data of the waterproof structure on the current immersed tube joint; The immersed tube joint is the junction between adjacent immersed tubes. The key area is to set at least three monitoring points on the top and both sides of each immersed tube junction; The standardized stress change data, displacement change data, water pressure change data and sediment density change data of the ith immersed tube connection during the jth microseismic activity are marked as ; The monitoring range of the seabed area includes a depth of d1 extending from the simulated seabed surface to capture the changes in sediment density within the entire S1 area.

4. The method for calculating damage performance of submarine tunnel structure based on deep learning according to claim 3 is characterized in that: The features of microseismic activity data, sediment density change data and key area monitoring data extracted by CNN include: 1.1) Data segmentation: The standardized key area monitoring data, microseismic activity data, and corresponding sediment density change data are segmented into a training set of 70%, a validation set of 15%, and a test set of 15%; 1.2) Use the convolutional neural network CNN to extract features from the segmented data and obtain the extracted feature number f; 1.3) Use the long short-term memory network LSTM to perform time series analysis on the extracted feature number f; 1.4) Model training.

5. The method for calculating damage performance of submarine tunnel structure based on deep learning according to claim 4 is characterized in that: If the physical submarine tunnel structure is subjected to external microseismic activity at the current moment, the real-time monitored microseismic activity data and sediment density change data are input into the calibrated deep learning model to obtain the prediction results of the monitoring data of each key area in the immersed tube connection sequence set, including: Using the same standardized parameters used in training, the real-time acquisition of microseismic activity data and sediment density change data of the physical submarine tunnel structure subjected to microseismic activity is standardized; The standardized real-time data is input into a calibrated convolutional neural network (CNN) to extract the features of microseismic activity data and sediment density change data; The extracted features are input into the calibrated long short-term memory network LSTM model to perform time series prediction to obtain the prediction results of the monitoring data of each key area; the specific acquisition of the prediction results is as follows: Combine the horizontal load, vertical load and microseismic activity of the physical submarine tunnel structure at the current moment Perform matching analysis to find the closest j+u-th combination experiment, where j+u∈{1,2,…,j,…,M}; The sediment density change data of the j+uth combined experiment is recorded as ; Output the prediction results of the key area monitoring data corresponding to each immersed tube connection at the current moment. The output value of the prediction result is limited to the range of (0,1]. The prediction result includes the stress change data of the ith immersed tube connection at the j+uth microseismic activity. , displacement change data , water pressure change data .

6. The method for calculating damage performance of submarine tunnel structure based on deep learning according to claim 5 is characterized in that: Generate impact adjustment factors, including: Before calculating the j+uth combined experiment, the average change trend coefficient of the sediment density change data of the physical submarine tunnel structure in the past e1 microseismic activities is calculated in advance, which is recorded as , the formula is: ； in, and They are the sediment density variation data of the i-th immersed tube connection during the j+uk and j+uk-1 microseismic activities; Calculate the real-time trend coefficient of sediment density change data during the current and previous external microseismic activity , the formula is: ； in, is the sediment density change data of the i-th immersed tube connection at the j+u-th microseismic activity at the current moment, is the sediment density change data of the i-th immersed tube connection during the last j+u-1 microseismic activity; Compare and analyze the real-time change trend coefficient with the average change trend coefficient to generate the impact adjustment coefficient , the formula is: ； in, Represents the impact adjustment coefficient; exp() is an exponential function; The value range of is (0,1]; when and As we get closer, When it is closer to 1, it means that the current sediment density change trend is more consistent with the historical average trend; on the contrary, when and The greater the difference, The closer it is to 0, the greater the difference between the current sediment density trend and the historical average trend; Setting the impact adjustment factor The judgment threshold is ; like The smaller the ,and The greater the impact on the physical undersea tunnel structure; like The smaller the ,and The smaller the impact on the physical undersea tunnel structure.

7. The method for calculating damage performance of submarine tunnel structure based on deep learning according to claim 6 is characterized in that: Generate a first evaluation index for evaluating the structural damage performance status of each immersed tube connection, specifically including: The first evaluation index of the i-th immersed tube connection is recorded as , the formula is: ； in, represents the first evaluation index of the ith immersed tube connection; represents the impact adjustment coefficient; set up The value range is (0,1]; based on the expert group system and experimental data analysis, the first evaluation index is set The judgment threshold is ; and stress change data , displacement change data , water pressure change data The warning thresholds are ; when When, and , , When any value in exceeds the corresponding warning threshold, the structural damage performance status of the i-th immersed tube connection is level one damage; level one damage means that the damage degree of the i-th immersed tube connection is less than 50%; when When, and , , When any value exceeds the corresponding warning threshold, the structural damage performance status of the i-th immersed tube connection is level 2 damage; level 2 damage means that the damage degree of the i-th immersed tube connection is between 50% and 70%; when , , ,and , , When any value exceeds the corresponding warning threshold, the structural damage performance status of the i-th immersed tube connection is level 3 damage; level 3 damage means that the degree of damage of the i-th immersed tube connection is between 70% and 100%.

8. A deep learning-based calculation system for damage performance of submarine tunnel structures, characterized by: The system is used to execute the deep learning-based damage performance calculation method for submarine tunnel structure according to any one of claims 1 to 7, comprising: Marking module: obtain historical microseismic activity data of the physical submarine tunnel structure that has experienced e1 microseismic activities in the past, as well as sediment density change data of the selected seabed area above each immersed tube connection; Mark the immersed tube connections of the submarine tunnel structure model specimen to form an immersed tube connection sequence set; And conduct M times of incremental microseismic activity simulation experiments on the submarine tunnel structure model specimen, M>e1, and monitor the sediment density change data of the selected seabed area above each immersed tube connection in real time in the incremental microseismic activity data of each immersed tube connection; Key area monitoring data acquisition module: used to set the time period before and after the submarine tunnel structure model specimen is subjected to external microseismic activity as the abnormal monitoring time period, and collect key area monitoring data of each immersed tube connection within the abnormal monitoring time period; Deep learning module: A deep learning model is constructed based on convolutional neural network (CNN) and long short-term memory (LSTM) networks. The microseismic activity data and sediment density change data of each microseismic activity simulation experiment are used as input features, and the key area monitoring data are used as output labels to train the deep learning model. The convolutional neural network (CNN) is used to extract the features of microseismic activity data, sediment density change data, and key area monitoring data. The extracted features of microseismic activity data and sediment density change data are used as input, and the features of key area monitoring data are used as output labels to train the long short-term memory network (LSTM). Prediction result acquisition module: when the physical submarine tunnel structure is subjected to external microseismic activity at the current moment, the collected historical microseismic activity data and sediment density change data of the physical submarine tunnel structure are input into the trained deep learning model for calibration; The real-time monitored microseismic activity data and sediment density change data are input into the calibrated deep learning model to obtain the prediction results of the monitoring data of each key area in the immersed tube connection sequence set; Impact adjustment coefficient generation module: used to calculate the average change trend coefficient of sediment density change data in the past e1 external microseismic activities, and calculate the real-time change trend coefficient of sediment density change data during the current and last external microseismic activities, compare and analyze the real-time change trend coefficient with the average change trend coefficient, and generate the impact adjustment coefficient; The first evaluation index generation module is used to combine the prediction results of the monitoring data of each key area with the impact adjustment coefficient for analysis, and generate the first evaluation index for evaluating the structural damage performance status of each immersed tube connection.

Images