An open caisson bottom support state intelligent analysis method and system based on correlation features

Abstract

The application discloses a kind of based on associated feature's caisson bottom support state intelligent analysis method and system, relating to caisson construction monitoring technical field, method includes: obtaining the strain time series data of target monitoring position on caisson bottom in drainage sinking stage;Strain time series data is preprocessed and the classification identification of support state, and construct sample data with identification, support state includes void and non-void;Multiple correlation analysis is carried out to sample data with identification, and multiple correlation information between sample data and support state is mined;For each kind of correlation information, a classification prediction model is constructed, and each classification prediction model is trained by each kind of correlation information;The trained each classification prediction model is deployed to undrained sinking stage application, and the support state of caisson bottom in undrained sinking stage is comprehensively judged according to the prediction result of each classification prediction model.The application can realize the accurate and efficient identification of the support state of caisson bottom in undrained sinking stage.

Description

Technical Field

[0001] This invention relates to the field of caisson construction monitoring technology, and in particular to an intelligent analysis method and system for the bottom support status of caissons based on correlation features. Background Technology

[0002] With the continuous construction of long-span heavy-load bridges, caisson foundations have become one of the important foundation forms in the field of bridge construction. However, the construction process still faces many key and difficult issues. Since the caisson foundation is dredged underwater during the non-drainage sinking stage, underwater operations are characterized by poor visibility and high difficulty. The support status at the bottom of the caisson foundation cannot be directly judged. How to carry out accurate and efficient identification work has become one of the important technical challenges in the non-drainage sinking stage. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent analysis method and system for the bottom support status of caissons based on correlation features, thereby solving the technical problem that the bottom support status of caissons cannot be accurately determined during the non-drainage sinking stage.

[0004] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0005] In a first aspect, the present invention provides an intelligent analysis method for the bottom support state of a caisson based on associated features, comprising:

[0006] Acquire strain time-series data at the target monitoring position at the bottom of the caisson during the drainage and sinking stage;

[0007] The strain time series data is preprocessed and the support states are classified and labeled, and labeled sample data is constructed. The support states include voided and non-voided states.

[0008] Multiple correlation analyses are performed on the labeled sample data to uncover various correlation information between the sample data and the support status;

[0009] A classification prediction model is constructed for each type of association information, and the classification prediction model is trained using each type of association information;

[0010] The trained classification prediction models are deployed to the non-drained sinking stage, and the bottom support status of the caisson is comprehensively judged based on the prediction results of each classification prediction model.

[0011] Optionally, the target monitoring locations include the cutting edge of the caisson and key stress points on the partition wall;

[0012] If a vibrating wire strain gauge is used for detection, it should be placed at the critical stress point.

[0013] If fiber optic gratings or distributed optical fibers are used for detection, they should be densely distributed to cover the surface of the partition wall and the bottom of the blade.

[0014] Optionally, the preprocessing of the strain time series data includes missing value processing, outlier processing, and filtering and noise reduction processing.

[0015] Optionally, the classification identifier for the support status includes:

[0016] The support condition was investigated on-site, and the investigation time and results were recorded each time.

[0017] Starting from each exploration time, preprocessed strain time series data of a preset duration are obtained as sample data, and the sample data are labeled according to the exploration results.

[0018] Optionally, the various association analyses performed on the labeled sample data include:

[0019] Select a preset number of sample data that are marked as non-empty, and denot them as non-empty sample data; concatenate each of the non-empty sample data as a column to generate a non-empty analysis matrix;

[0020] Select a preset number of sample data marked as "empty" and denot them as empty sample data; concatenate each empty sample data as a column to generate an empty analysis matrix;

[0021] Calculate multiple correlation coefficients between each column of data in the non-empty analysis matrix and each column of data in the non-empty analysis matrix, and generate a non-empty correlation coefficient matrix for each type of correlation coefficient;

[0022] Calculate various correlation coefficients between each column of the non-void analysis matrix and each column of the void analysis matrix, and generate a void correlation coefficient matrix for each type of correlation coefficient.

[0023] Optionally, the correlation coefficients include the grey correlation coefficient, the maximum mutual information number, the Pearson correlation coefficient, and the Spearman correlation coefficient.

[0024] Optionally, training the classification prediction model using each type of association information includes:

[0025] Flatten the non-void correlation coefficient matrix and the void correlation coefficient matrix from multidimensional to one-dimensional to obtain the non-void correlation coefficient vector and the void correlation coefficient vector.

[0026] The non-empty correlation coefficient vector and the empty correlation coefficient vector are extracted by sliding through a window of preset length to obtain the non-empty correlation coefficient array and the empty correlation coefficient array.

[0027] The classification identifiers of the non-empty association coefficient array and the empty association coefficient array are set to 1 and 0 respectively, representing non-empty and empty, thus forming a dataset;

[0028] The dataset is stratified and sampled according to a preset ratio to generate training and test sets. The corresponding classification prediction model is then trained and tested using the training and test sets until training is complete.

[0029] Optionally, deploying the trained classification prediction models to the non-drainage sinking stage includes:

[0030] Acquire strain time-series data at the target monitoring position at the bottom of the caisson during the non-drainage sinking stage;

[0031] The strain time series data is preprocessed, and sample data to be identified is constructed;

[0032] Select a preset number of sample data to be identified, denoted as sample data to be identified; concatenate each of the sample data to be identified as a column to generate an analysis matrix to be identified;

[0033] Calculate multiple correlation coefficients between each column of the non-empty analysis matrix and each column of the analysis matrix to be identified, and generate a correlation coefficient matrix to be identified for each type of correlation coefficient;

[0034] Flatten the multidimensional correlation coefficient matrix to be identified into a one-dimensional matrix to obtain the correlation coefficient vector to be identified;

[0035] The vector of correlation coefficients to be identified is extracted by sliding through a window of preset length to obtain an array of correlation coefficients to be identified;

[0036] The association coefficient data to be identified is input into each of the trained classification prediction models to obtain the prediction results.

[0037] Secondly, the present invention provides an intelligent analysis system for the bottom support status of caissons based on correlation features, comprising:

[0038] The data acquisition module is configured to acquire strain time-series data of the target monitoring position at the bottom of the caisson during the drainage sinking stage;

[0039] The data processing module is configured to preprocess the strain time series data and classify the support states, and construct labeled sample data, wherein the support states include voided and non-voided states;

[0040] The association analysis module is configured to perform various association analyses on the labeled sample data to mine various association information between the sample data and the supporting state;

[0041] The model building module is configured to build a classification prediction model for each type of association information, and to train the classification prediction model using each type of association information;

[0042] The application deployment module is configured to deploy the trained classification prediction models to the non-drainage sinking stage and comprehensively judge the bottom support status of the caisson in the non-drainage sinking stage based on the prediction results of each classification prediction model.

[0043] Thirdly, the present invention provides an electronic device, including a processor and a storage medium;

[0044] The storage medium is used to store instructions;

[0045] The processor is configured to operate according to the instructions to perform the steps according to the method described above.

[0046] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0047] This invention provides an intelligent analysis method and system for the bottom support status of caissons based on correlation features. By collecting strain time-series data from target monitoring locations on the caisson structure as the data foundation, and combining this with a correlation analysis design scheme suitable for judging the bottom support status of caissons, a classification prediction model is constructed and trained during the drainage and sinking stage. By integrating multiple correlation algorithms, it comprehensively considers linear, monotonic, similar, and nonlinear correlation features between data, enabling efficient analysis of the support status and achieving accurate and efficient prediction by the classification prediction model. The trained classification prediction model is applied to the non-drainage sinking stage, solving the technical problem of inaccurate judgment of the bottom support status of caissons during this stage. This invention overcomes the limitations of traditional methods relying on experience-based judgment, achieving quantitative determination of the support status, reducing construction costs and operational difficulty, and ensuring the construction quality and stability of the caisson foundation. Attached Figure Description

[0048] Figure 1 This is a flowchart of the intelligent analysis method for the bottom support status of a caisson based on associated features provided in an embodiment of the present invention;

[0049] Figure 2 This is a local heatmap of the non-void correlation coefficient matrix corresponding to the Spearman correlation coefficient provided in this embodiment of the invention;

[0050] Figure 3 This is a local heatmap of the void correlation coefficient matrix corresponding to the Spearman correlation coefficient provided in this embodiment of the invention;

[0051] Figure 4 This is a confusion matrix diagram of the classification prediction model corresponding to the Spearman correlation coefficient provided in this embodiment of the invention;

[0052] Figure 5 This is the ROC curve of the classification prediction model corresponding to the Spearman correlation coefficient provided in this embodiment of the invention. Detailed Implementation

[0053] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0054] Example 1

[0055] like Figure 1 As shown, this embodiment of the invention provides an intelligent analysis method for the bottom support status of a caisson based on associated features, including the following steps:

[0056] Step S1: Obtain strain time series data of the target monitoring position at the bottom of the caisson during the drainage and sinking stage.

[0057] Specifically, in this embodiment, the target monitoring locations include the cutting edge of the caisson and key stress points on the partition wall;

[0058] If a vibrating wire strain gauge is used for testing, it should be placed at the critical stress point;

[0059] If fiber optic gratings or distributed optical fibers are used for detection, they should be densely distributed to cover the surface of the partition wall and the bottom of the blade.

[0060] In other alternative implementations, technicians can freely set the target monitoring location as needed. This invention selects the cutting edge and partition wall to obtain the structural stress at the bottom of the caisson under different support states as construction progresses, thereby providing favorable data support for the intelligent analysis and prediction of the subsequent support state.

[0061] Reliable and accurate monitoring sensors are a crucial prerequisite for efficient identification work. Fiber Bragg grating (FBG) monitoring technology utilizes optical fiber cables as the carrier for monitoring and signal transmission. Leveraging the wavelength sensitivity of FBGs, it infers changes in external physical quantities by capturing shifts in the wavelengths of specific reflected, transmitted, or refracted light, thereby enabling the monitoring of structural mechanical behavior. Densely distributed FBGs form a distributed sensor array. The spacing between the gratings in this densely distributed array should be designed in conjunction with the monitoring requirements and the dimensions of the well, retaining the high precision of point monitoring while possessing the wide coverage of distributed sensors. This results in advantages such as high accuracy and high spatial resolution.

[0062] Step S2: Preprocess the strain time series data and classify and label the support states, and construct labeled sample data. The support states include voided and non-voided states.

[0063] Preprocessing of strain time series data includes, but is not limited to, handling missing values, handling outliers, and filtering and denoising. Preprocessing can make strain time series data cleaner and more complete, enabling subsequent classification models to make more reliable estimates, thereby significantly improving prediction accuracy and reducing prediction errors.

[0064] The classification identifiers for support status include:

[0065] Conduct on-site surveys of the support conditions and record the survey time and results for each survey.

[0066] Starting from each exploration time, preprocessed strain time series data of a preset duration are obtained as sample data, and labels are added to the sample data according to the exploration results.

[0067] Step S3: Perform various correlation analyses on the labeled sample data to uncover multiple correlation information between the sample data and the supporting status.

[0068] This invention is based on correlation algorithms to mine the correlation information between samples and features in monitoring data. Targeting the characteristics of caisson engineering, it designs a targeted correlation analysis scheme focusing on applicability and computational efficiency, and studies the time trend patterns under different support states. The correlation algorithms include, but are not limited to, grey correlation coefficient, maximum mutual information number, Pearson correlation coefficient, and Spearman correlation coefficient.

[0069] Specifically, various association analyses are performed on the labeled sample data, including:

[0070] Select a preset number of sample data marked as non-empty data, and denot them as non-empty sample data; concatenate each non-empty sample data as a column to generate a non-empty analysis matrix;

[0071] Select a preset number of sample data marked as "empty" and denot them as empty sample data; concatenate each empty sample data as a column to generate an empty analysis matrix;

[0072] Calculate the various correlation coefficients between each column of data in the non-empty analysis matrix and each column of data in the non-empty analysis matrix, and generate a non-empty correlation coefficient matrix for each type of correlation coefficient;

[0073] Calculate various correlation coefficients between each column of the non-void analysis matrix and each column of the void analysis matrix, and generate a void correlation coefficient matrix for each type of correlation coefficient.

[0074] Taking the Spearman correlation coefficient as an example, the heatmaps of the non-void correlation coefficient matrix and the void correlation coefficient matrix obtained from its calculation are as follows: Figure 2 and Figure 3 As shown, due to the large amount of data actually calculated, the resulting heatmap is also quite large; this embodiment only shows a portion of the heatmap. Figure 2The heatmap shown uses column indices of the non-void analysis matrix as its axes. Each heatmap value represents the Spearman correlation coefficient between the corresponding column data of the non-void analysis matrix; the closer the value is to 1, the stronger the correlation between the two columns. For example... Figure 3 The heatmap shown has the column number of the non-void analysis matrix on the vertical axis and the column number of the void analysis matrix on the horizontal axis. Each heatmap value is the Spearman correlation coefficient between the corresponding column data of the non-void analysis matrix and the corresponding column data of the void analysis matrix. The closer the value is to 1, the stronger the correlation between the two columns of data.

[0075] Step S4: Build a classification prediction model for each type of association information, and train the classification prediction model using each type of association information.

[0076] Specifically, in this embodiment, the classification prediction model uses a support vector machine (SVM), which constructs the model by calculating the mapping relationship between the correlation coefficient and the classification features.

[0077] Training the classification prediction model using each type of association information includes:

[0078] Flatten the non-void correlation coefficient matrix and the void correlation coefficient matrix from multidimensional to one-dimensional to obtain the non-void correlation coefficient vector and the void correlation coefficient vector.

[0079] By using a window of preset length to slide and extract the non-empty correlation coefficient vector and the empty correlation coefficient vector, the non-empty correlation coefficient array and the empty correlation coefficient array are obtained.

[0080] The classification labels for the non-empty association coefficient array and the empty association coefficient array are set to 1 and 0 respectively, representing non-empty and empty, thus forming the dataset;

[0081] The dataset is stratified and sampled according to a preset ratio to generate training and test sets. The corresponding classification prediction models are then trained and tested using the training and test sets until training is complete.

[0082] Stratified sampling ensures that the ratio of the two classes of samples in the test set and training set is consistent with the original data, avoiding evaluation bias caused by class imbalance. The grid search algorithm combined with cross-validation is used to search for hyperparameters, and the grid search calculation index is extracted to obtain a support vector machine (SVM) initialized with optimal parameters and trained.

[0083] Step S5: Deploy the trained prediction models of each category to the non-drained sinking stage and comprehensively judge the support status of the bottom of the caisson in the non-drained sinking stage based on the prediction results of each category prediction model.

[0084] Step S5.1: Deploy the trained prediction models for each category to the non-drainage sinking stage application, including:

[0085] Acquire strain time-series data at the target monitoring location at the bottom of the caisson during the non-drainage sinking stage;

[0086] The strain time series data are preprocessed, and sample data to be identified are constructed.

[0087] Select a preset number of sample data to be labeled, denoted as sample data to be labeled; concatenate each sample data as a column to generate an analysis matrix to be labeled;

[0088] Calculate multiple correlation coefficients between each column of the non-empty analysis matrix and each column of the analysis matrix to be identified, and generate a correlation coefficient matrix to be identified for each type of correlation coefficient;

[0089] Flatten the multidimensional correlation coefficient matrix to be identified into a one-dimensional matrix to obtain the correlation coefficient vector to be identified;

[0090] The correlation coefficient vector to be identified is extracted by sliding through a window of preset length to obtain an array of correlation coefficients to be identified;

[0091] The correlation coefficient data to be identified is input into the trained classification prediction models to obtain the prediction results.

[0092] Using Spearman's correlation coefficient as an example, the performance of its classification prediction model is evaluated through confusion matrix and ROC curve. The confusion matrix shows the distribution of the prediction results of the classification prediction model. The prediction results are cross-tabulated with the true labels, directly showing the number of true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN). The ROC curve depicts the model's performance at different decision thresholds, with the false positive rate (FPR) on the horizontal axis and the true positive rate (TPR / recall) on the vertical axis, showing the trade-off between the two when the threshold changes. The area under the curve is the AUC. Figure 4 and Figure 5 These are the confusion matrix and ROC curve, respectively. Class A is non-empty, and Class B is empty. AUC is the area under the curve. The larger the AUC value, the stronger the model's ability to distinguish between positive and negative samples.

[0093] Step S5.2: Based on the prediction results of each classification prediction model, comprehensively determine the bottom support status of the caisson during the non-drainage sinking stage, including:

[0094] Taking the four classification prediction models constructed using the four types of correlation information provided in this embodiment—grey relational coefficient, maximum mutual information number, Pearson correlation coefficient, and Spearman correlation coefficient—as examples, each classification prediction model outputs a probability of empty space and a probability of not empty space. Technical personnel can select the identifier corresponding to the highest probability among all probabilities, or they can select based on quantity; for example, if three classification prediction models predict empty space and one predicts not empty space, then empty space is ultimately selected. There are many ways to make a comprehensive judgment, and this invention does not impose too many limitations.

[0095] Table 1 shows the prediction results of the classification prediction model under the grey relational coefficient, maximum mutual information number, Pearson correlation coefficient, and Spearman correlation coefficient. Class A represents no missing data, and Class B represents missing data.

[0096] Table 1: Prediction results of four classification prediction models

[0097]

[0098] In summary, this invention provides an intelligent analysis method for the bottom support state of caissons based on correlation features. It aims to solve the problem of determining the bottom support state during the non-drainage sinking stage by integrating correlation analysis algorithms and classification prediction algorithms, based on strain monitoring sensors deployed on the caisson structure. By laying strain monitoring sensors along the bottom cutting edges of the outer wall and inner partition walls of the caisson, the structural stress at the bottom of the caisson changes with construction under different support states. The collected strain signal data is then preprocessed, and the time-series data under different support states during the drainage sinking stage are classified. Based on the correlation algorithm, a targeted correlation analysis scheme is designed considering the applicability and computational efficiency of caisson engineering. Finally, an intelligent classification model is constructed using the classification prediction algorithm to explore the relationship between the correlation coefficient and classification features, and the support state is identified and calculated based on the time trend analysis results of different support states.

[0099] Example 2

[0100] This invention provides an intelligent analysis system for the bottom support status of caissons based on correlation features, comprising:

[0101] The data acquisition module is configured to acquire strain time-series data of the target monitoring position at the bottom of the caisson during the drainage sinking stage;

[0102] The data processing module is configured to preprocess the strain time series data and classify the support states, and to construct labeled sample data, including support states that are not empty;

[0103] The association analysis module is configured to perform various association analyses on labeled sample data to uncover multiple association information between the sample data and the supporting status.

[0104] The model building module is configured to build a classification prediction model for each type of association information and train the classification prediction model using each type of association information;

[0105] The application deployment module is configured to deploy the trained prediction models of each category to the non-drainage sinking stage and comprehensively judge the support status of the bottom of the caisson in the non-drainage sinking stage based on the prediction results of each category prediction model.

[0106] Example 3

[0107] Based on the intelligent analysis method for the bottom support status of caissons based on associated features provided in Embodiment 1, this embodiment of the invention provides an electronic device, including a processor and a storage medium;

[0108] Storage media are used to store instructions;

[0109] The processor is used to perform operations according to instructions to execute the steps according to the method described above.

[0110] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0111] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0112] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0113] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0114] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent analysis of the bottom support state of a caisson based on correlation features, characterized in that, include: Acquire strain time-series data at the target monitoring position at the bottom of the caisson during the drainage and sinking stage; The strain time series data is preprocessed and the support states are classified and labeled, and labeled sample data is constructed. The support states include voided and non-voided states. Multiple correlation analyses are performed on the labeled sample data to uncover various correlation information between the sample data and the support status; A classification prediction model is constructed for each type of association information, and the classification prediction model is trained using each type of association information; The trained classification prediction models are deployed to the non-drainage sinking stage, and the bottom support status of the caisson is judged comprehensively based on the prediction results of each classification prediction model. The various association analyses performed on the labeled sample data include: Select a preset number of sample data that are marked as non-empty, and denot them as non-empty sample data; concatenate each of the non-empty sample data as a column to generate a non-empty analysis matrix; Select a preset number of sample data marked as "empty" and denot them as empty sample data; concatenate each empty sample data as a column to generate an empty analysis matrix; Calculate multiple correlation coefficients between each column of data in the non-empty analysis matrix and each column of data in the non-empty analysis matrix, and generate a non-empty correlation coefficient matrix for each type of correlation coefficient; Calculate various correlation coefficients between each column of the non-void analysis matrix and each column of the void analysis matrix, and generate a void correlation coefficient matrix for each type of correlation coefficient.

2. The intelligent analysis method for the bottom support state of a caisson based on correlation features according to claim 1, characterized in that, The target monitoring locations include the cutting edge of the caisson and key stress points on the partition wall; If a vibrating wire strain gauge is used for detection, it should be placed at the critical stress point. If fiber optic gratings or distributed optical fibers are used for detection, they should be densely distributed to cover the surface of the partition wall and the bottom of the blade.

3. The intelligent analysis method for the bottom support state of a caisson based on correlation features according to claim 1, characterized in that, The preprocessing of the strain time series data includes missing value processing, outlier processing, and filtering and noise reduction processing.

4. The intelligent analysis method for the bottom support state of a caisson based on correlation features according to claim 1, characterized in that, The classification identifiers for the support status include: The support condition was investigated on-site, and the investigation time and results were recorded each time. Starting from each exploration time, preprocessed strain time series data of a preset duration are obtained as sample data, and the sample data are labeled according to the exploration results.

5. The intelligent analysis method for the bottom support state of a caisson based on correlation features according to claim 1, characterized in that, The correlation coefficients include the grey correlation coefficient, the maximum mutual information number, the Pearson correlation coefficient, and the Spearman correlation coefficient.

6. The intelligent analysis method for the bottom support state of a caisson based on correlation features according to claim 1, characterized in that, The step of training the classification prediction model using each type of association information includes: Flatten the non-void correlation coefficient matrix and the void correlation coefficient matrix from multidimensional to one-dimensional to obtain the non-void correlation coefficient vector and the void correlation coefficient vector. The non-empty correlation coefficient vector and the empty correlation coefficient vector are extracted by sliding through a window of preset length to obtain the non-empty correlation coefficient array and the empty correlation coefficient array. The classification identifiers of the non-empty association coefficient array and the empty association coefficient array are set to 1 and 0 respectively, representing non-empty and empty, thus forming a dataset; The dataset is stratified and sampled according to a preset ratio to generate training and test sets. The corresponding classification prediction model is then trained and tested using the training and test sets until training is complete.

7. The intelligent analysis method for the bottom support state of a caisson based on correlation features according to claim 1, characterized in that, The step of deploying the trained classification prediction models to the non-drainage sinking stage includes: Acquire strain time-series data at the target monitoring position at the bottom of the caisson during the non-drainage sinking stage; The strain time series data is preprocessed, and sample data to be identified is constructed; Select a preset number of sample data to be identified, denoted as sample data to be identified; concatenate each of the sample data to be identified as a column to generate an analysis matrix to be identified; Calculate multiple correlation coefficients between each column of the non-empty analysis matrix and each column of the analysis matrix to be identified, and generate a correlation coefficient matrix to be identified for each type of correlation coefficient; Flatten the multidimensional correlation coefficient matrix to be identified into a one-dimensional matrix to obtain the correlation coefficient vector to be identified; The vector of correlation coefficients to be identified is extracted by sliding through a window of preset length to obtain an array of correlation coefficients to be identified; The association coefficient data to be identified is input into each of the trained classification prediction models to obtain the prediction results.

8. An intelligent analysis system for the bottom support status of a caisson based on correlation features, characterized in that, include: The data acquisition module is configured to acquire strain time-series data of the target monitoring position at the bottom of the caisson during the drainage sinking stage; The data processing module is configured to preprocess the strain time series data and classify the support states, and construct labeled sample data, wherein the support states include voided and non-voided states; The association analysis module is configured to perform various association analyses on the labeled sample data to mine various association information between the sample data and the supporting state; The various association analyses performed on the labeled sample data include: Select a preset number of sample data that are marked as non-empty, and denot them as non-empty sample data; concatenate each of the non-empty sample data as a column to generate a non-empty analysis matrix; Select a preset number of sample data marked as "empty" and denot them as empty sample data; concatenate each empty sample data as a column to generate an empty analysis matrix; Calculate multiple correlation coefficients between each column of data in the non-empty analysis matrix and each column of data in the non-empty analysis matrix, and generate a non-empty correlation coefficient matrix for each type of correlation coefficient; Calculate multiple correlation coefficients between each column of data in the non-void analysis matrix and each column of data in the void analysis matrix, and generate a void correlation coefficient matrix for each type of correlation coefficient; The model building module is configured to build a classification prediction model for each type of association information, and to train the classification prediction model using each type of association information; The application deployment module is configured to deploy the trained classification prediction models to the non-drainage sinking stage and comprehensively judge the bottom support status of the caisson in the non-drainage sinking stage based on the prediction results of each classification prediction model.

9. An electronic device, characterized in that, Including processor and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to perform the steps of the method according to any one of claims 1-7.

Images