Post-construction settlement prediction and risk assessment system for soft soil foundation

Abstract

A post-construction settlement prediction and risk assessment system for a soft soil foundation comprises a data acquisition module which is arranged at a key position of the soft soil foundation and comprises a soil humidity sensor, a pressure sensor and a displacement sensor; the data analysis module is used for preprocessing the data acquired by the data acquisition module, extracting key features from the preprocessed original data and taking the key features as model input; the settlement prediction module comprises a finite element model and a prediction model, inputs the model into the finite element model, outputs settlement results under different load conditions, and outputs corresponding settlement curves and prediction trends in combination with the prediction model; and the risk assessment module is used for outputting a corresponding risk level according to a set threshold value based on the prediction trend of the settlement prediction module. An efficient technical means is provided for monitoring and predicting the settlement risk, a scientific basis is provided for decision making in the construction process, and the safety and economical efficiency of engineering are improved.

Description

Technical Field

[0001] The invention relates to the technical field of roadbed settlement, and in particular to a soft soil foundation post-construction settlement prediction and risk assessment system. Background Art

[0002] Globally, a large amount of infrastructure construction and urban development takes place on soft soil foundations. The physical and mechanical properties of soft soil make it easy to consolidate and settle after load is applied, affecting the safety and function of buildings. Existing settlement prediction methods mostly rely on empirical formulas, which are often unable to adapt to the complexity of different soil properties and construction conditions. In addition, traditional monitoring methods such as water level gauges and settlement observation wells often have lags and are difficult to reflect soil changes in a timely manner. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides a soft soil foundation post-construction settlement prediction and risk assessment system.

[0004] To achieve the above object, the present invention provides the following technical solutions: A soft soil foundation post-construction settlement prediction and risk assessment system, comprising: The data acquisition module is arranged at a key position of the soft soil foundation, and includes a soil moisture sensor for detecting the moisture inside the soil, a pressure sensor for detecting the pressure inside the soil, and a displacement sensor for detecting the displacement inside the soil; The data analysis module preprocesses the data collected by the data acquisition module and extracts key features from the preprocessed raw data as model input; The settlement prediction module includes a finite element model and a prediction model. The model input is input into the finite element model, the settlement results under different load conditions are output, and the corresponding settlement curve and prediction trend are output in combination with the prediction model; The risk assessment module outputs the corresponding risk level based on the prediction trend of the settlement prediction module and the set threshold.

[0005] The data acquisition module is also provided with a temperature sensor for collecting the internal pressure temperature of the soil.

[0006] The data analysis module realizes key feature extraction based on the following steps: Data preprocessing, including data denoising, outlier detection, and interpolation and filling operations; Feature extraction, including: Correlation analysis was used to screen out characteristic variables that were closely related to settlement; Principal component analysis is used to reduce the dimensionality of high-dimensional data and obtain the main influencing factors.

[0007] The data analysis module also includes a feature extraction model, which outputs model inputs required by the settlement prediction module based on the data collected by the data collection module.

[0008] The model inputs include soil cohesion, friction angle, and elastic modulus.

[0009] The finite element model is a three-dimensional finite element model. The specific operation steps of the three-dimensional finite element model are as follows: 1. By setting boundary conditions and initial conditions; 2. Select the corresponding material model according to the properties of the soil, and input the corresponding physical and mechanical parameters of the soil; 3. Simulate different load conditions during construction, conduct static or dynamic analysis, and obtain the stress-strain state of the soil; 4. The consolidation and settlement process of soil under applied load is obtained through finite element calculation.

[0010] The prediction model is based on the random forest model, Firstly, a 10-fold cross-validation random forest model was established based on the land subsidence and main controlling factors; From the training sample set Z = {( x 1 , y 1 ),(x 2 ,y 2 ),…,(x n ,y n )}, randomly generate N training sets M 1 ,M 2 ,…,M N ; Generate a corresponding decision tree T for each training set 1 ,T 2 ,…,T N ; Randomly select l attributes from all L attributes, l≪L as the split attribute set of the current node, and select the best split attribute as the node to split; The generated multiple decision trees form a random forest. For the test sample set Z, each decision tree is used for testing. Get the prediction result T 1 ( x),T 2 (x),…,T N (x), where T(x) represents N T N The set of (x), i.e., the random forest prediction result; T N (x) represents the prediction result of a single decision tree.

[0011] It also has a visual interface for presenting settlement prediction results.

[0012] The settlement prediction results include predicted settlement curves, risk assessment results, and sensitivity analysis diagrams.

[0013] The beneficial effects of the invention are as follows: it not only provides an efficient technical means to monitor and predict settlement risks, but also provides a scientific basis for decision-making during the construction process, improving the safety and economy of the project. Through continuous technical iteration and optimization, the system has the potential to play an important role in a wider range of engineering applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a land subsidence prediction model based on random forest.

[0015] Figure 2 Schematic diagram of the comparison between the prediction results and actual values ​​of different models on the test set, including a, the comparison and verification of the predicted value of random forest for land subsidence in coastal areas and b, the comparison and verification of the predicted value of multivariate linear regression and the observed value. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] The present invention discloses a soft soil foundation post-construction settlement prediction and risk assessment system based on comprehensive data analysis and real-time monitoring, which comprises: The data acquisition module is arranged at a key position of the soft soil foundation, and includes a soil moisture sensor for detecting the moisture inside the soil, a pressure sensor for detecting the pressure inside the soil, and a displacement sensor for detecting the displacement inside the soil; The data analysis module preprocesses the data collected by the data acquisition module and extracts key features from the preprocessed raw data as model input; The settlement prediction module includes a finite element model and a prediction model. The model input is input into the finite element model, the settlement results under different load conditions are output, and the corresponding settlement curve and prediction trend are output in combination with the prediction model; The risk assessment module outputs the corresponding risk level based on the prediction trend of the settlement prediction module and the set threshold.

[0018] The data acquisition module includes several sensors, which are set at key locations of the soft soil foundation, including soil moisture sensors, pressure sensors, displacement sensors and temperature sensors, to comprehensively and real-time monitor the physical state of the soil.

[0019] All sensor data are transmitted to the central processing unit in real time via wireless networks to ensure the timeliness and accuracy of the data.

[0020] The data analysis module first preprocesses the data collected by the data acquisition module, cleans the collected data, removes outliers, processes missing data, and ensures data quality. The preprocessing steps include: Denoising: Applying signal processing techniques (such as filters) to remove noise from sensor data.

[0021] Outlier detection: Identify and handle outliers through statistical methods (such as Z-score, IQR).

[0022] Interpolation and filling: For missing data, interpolation methods (such as linear interpolation and spline interpolation) are used to fill the missing values.

[0023] After the preprocessing, key features such as soil compressibility and water content change rate are extracted from the raw data as model input.

[0024] Feature selection uses correlation analysis (such as Pearson correlation coefficient) to screen out characteristic variables that are closely related to settlement.

[0025] Principal component analysis (PCA) is used to reduce the dimensionality of high-dimensional data and find out the main influencing factors.

[0026] At the same time, new features can be constructed, such as soil compressibility, water content change rate, etc., to improve the performance of the model.

[0027] Standardize or normalize the features to suit subsequent model training.

[0028] The data analysis module also includes a feature extraction model, which outputs model inputs required by the settlement prediction module based on the data collected by the data collection module.

[0029] The feature extraction model selects an appropriate machine learning algorithm, such as linear regression, random forest, support vector machine (SVM), or neural network.

[0030] Evaluate the performance of the model through cross-validation (such as k-fold cross-validation).

[0031] The preprocessed data is divided into a training set and a test set, and the training set is used to train the model.

[0032] The model training process includes the following steps: (1) Data preprocessing The input data is cleaned, denoised, and normalized to improve the training effect of the model.

[0033] (2) Divide the data set The data set is divided into a training set and a test set, generally 70%-80% is used for training and 20%-30% is used for testing.

[0034] (3) Model training The training set is input into the selected machine learning model for training, and the model learns the relationship between sedimentation and features based on the input features.

[0035] The model parameters are adjusted through an optimization algorithm (such as gradient descent) to minimize the loss function (such as mean squared error).

[0036] (4) Model evaluation Use the test set to evaluate the trained model and calculate the accuracy of the prediction results (such as RMSE, R², etc.) to determine the performance of the model.

[0037] Tune hyperparameters (e.g., through grid search or random search) to optimize model performance.

[0038] The model performance is evaluated using a variety of evaluation indicators (such as root mean square error RMSE and determination coefficient R²).

[0039] The accuracy of the classification model was evaluated using methods such as confusion matrix and ROC curve.

[0040] The analysis results are presented in a visual way, including data distribution graphs, feature importance graphs, and prediction result graphs, to facilitate subsequent decision-making.

[0041] The settlement prediction module integrates two functions, one is finite element analysis, the other is real-time prediction, that is, the purpose of the settlement prediction module is to perform settlement prediction based on the analyzed data.

[0042] Finite element analysis: Construct a finite element model to perform mechanical calculations on the soft soil foundation and evaluate the settlement under different load conditions through numerical simulation.

[0043] Real-time prediction: Combine machine learning models with finite element analysis results to update settlement predictions in real time and output settlement curves and future trends.

[0044] Finite Element Analysis (1) Model establishment A three-dimensional finite element model is established based on the geological data and soil compression characteristics of the project site.

[0045] Set boundary conditions (such as fixed boundaries, free boundaries) and initial conditions.

[0046] (2) Material model selection Select an appropriate material model (such as Mohr-Coulomb model, Cam-clay model, etc.) according to the properties of the soil.

[0047] Input the physical and mechanical parameters of the soil, such as elastic modulus, Poisson's ratio, internal friction angle, etc.

[0048] (3) Loading condition setting Simulate different load conditions during the construction process (such as building weight, construction load, groundwater level changes, etc.).

[0049] Perform static or dynamic analysis to obtain the stress-strain state of the soil.

[0050] (4) Analysis and solution Finite element software (such as ANSYS, ABAQUS, etc.) is used to calculate and analyze the consolidation and settlement process of soil under applied load.

[0051] Machine Learning Prediction (1) Prediction model selection Combining historical settlement data and monitoring data, a regression model (such as support vector regression SVR, random forest regression, etc.) is selected for modeling.

[0052] Select a suitable prediction model based on the characteristics of the input features. Commonly used machine learning models include: Linear regression: Simple and easy to interpret, suitable for linear relationships.

[0053] Random Forest: It constructs multiple decision trees for integrated learning and is suitable for processing complex nonlinear relationships.

[0054] Support Vector Machine (SVM): Suitable for processing high-dimensional data and can efficiently handle nonlinear regression problems.

[0055] Neural network: Suitable for processing large-scale complex data, especially when there is a complex relationship between input features and output.

[0056] (2) Feature input The features extracted from the data analysis module are used as model inputs. The model inputs include: 1. Soil parameters Moisture content: The proportion of water in the soil, which affects the compressibility of the soil.

[0057] Soil type: Such as clay, sand, etc. Different soil types have different physical properties.

[0058] Density: The compactness of soil, which affects its bearing capacity and settlement characteristics.

[0059] Mechanical properties parameters such as elastic modulus and Poisson's ratio.

[0060] 2. Construction load Self-weight of a building: vertical load affecting the soil.

[0061] Dynamic loads: such as the weight of construction equipment and other loads applied during the construction process.

[0062] 3. Groundwater level Changes in groundwater levels affect the effective stress of the soil, which in turn affects the settlement behavior.

[0063] 4. Time Factor Construction time and stage, the state and response of soil at different time points.

[0064] (3) Real-time prediction The trained model is used to make real-time predictions of the current soil state and obtain future settlement trends.

[0065] Set a time window and regularly update the forecast results to reflect changes in real-time monitoring data.

[0066] That is, after the model training is completed and verified, prediction can be performed. The specific steps are as follows: (1) Enter new data New real-time monitoring data (such as soil parameters, loads, groundwater levels, etc.) are input into the trained model.

[0067] (2) Generate predictions The model calculates the input features and generates corresponding settlement prediction values. Depending on the type of model used, the prediction output may be: ① Numerical output: represents the amount of settlement at a specific time point (such as millimeters).

[0068] ②Time series output: predict the settlement change trend within a certain period of time in the future.

[0069] The prediction model in the present invention adopts a random forest model. Firstly, a 10-fold cross-validation random forest model was established based on the land subsidence amount and the main controlling factors; The core idea of ​​random forest is to automatically sample the training set to form multiple training sets. Each training set generates a decision tree, and all decision trees form a random forest, thereby training and predicting the samples. It is a machine learning method.

[0070] From the training sample set Z = {( x 1 , y 1 ),(x 2 ,y 2 ),…,(x n ,y n)}, randomly generate N training sets M 1 ,M 2 ,…,M N ; Generate a corresponding decision tree T for each training set 1 ,T 2 ,…,T N ; Randomly select l attributes from all L attributes, l≪L as the split attribute set of the current node, and select the best split attribute as the node to split; The generated multiple decision trees form a random forest. For the test sample set Z, each decision tree is used for testing. Get the prediction result T 1 ( x),T 2 (x),…,T N (x), where T(x) represents N T N The set of (x), i.e., the random forest prediction result; T N (x) represents the prediction result of a single decision tree.

[0071] Based on the point data obtained by spatial interpolation and Pearson correlation analysis, the data corresponding to the main controlling factors of land subsidence (including groundwater level and hydrogeological condition parameters, etc.) are determined. In order to eliminate the impact of the dimensions between different indicators, the data is normalized by calling the normalization function, and all input data are normalized to 0 ~ 1. The normalization formula is:

[0072] Where: x i is the measured sample value at point i; x min is the minimum sample value; x max is the maximum sample value; max and min are the maximum and minimum values ​​of a given scaling range; x∗ is the normalized data.

[0073] In addition, the feature importance evaluation method of random forest is used to screen out the features that have a greater contribution to ground subsidence prediction from the original features, and some combined features are constructed according to actual conditions to increase the expressive power of the model.

[0074] Model validation and results analysis Some commonly used evaluation indicators, such as root mean squared error (RMSE) and coefficient of determination (R2), are used to evaluate the prediction effect of the random forest model. The mathematical expression of RMSE is as follows:

[0075] Where: RMSE is the root mean square error, the closer its value is to 0, the smaller the overall error of the prediction; x i is the observed data; y i is the measured data; N is the number of experimental data; similarly, R 2 The closer it is to 1, the smaller the overall prediction error is, and vice versa.

[0076] Figure 2 a shows the comparison between the prediction results of the random forest model on the test set and the actual values. It can be seen that the predicted values ​​are basically consistent with the actual values, indicating that the model can capture the trend of land subsidence over time. 2 =0.98, RMSE=0.52 mm, indicating that the model has high accuracy. The random forest method can effectively capture the spatiotemporal distribution of land subsidence, and its simulation effect is significantly better than other methods, such as multivariate linear regression ( Figure 2 b), neural networks, support vector machines, etc. This shows that the random forest method can make full use of the observed data, and at the same time has strong generalization ability and adaptability, and is suitable for processing high-dimensional and nonlinear data.

[0077] Risk Assessment Module Risk identification: Based on the settlement prediction results, set thresholds, identify potential settlement risks and classify them (such as low risk, medium risk, high risk).

[0078] Fault tree analysis: construct a fault tree to quantitatively analyze the factors that may cause settlement and determine their impact on the overall project.

[0079] Emergency plan generation: Generate corresponding emergency plans based on risk assessment results and provide them to the construction management team.

[0080] (1) Sensitivity analysis By changing the input parameters (such as soil properties, load size, etc.), the impact on the settlement prediction results is analyzed and the key influencing factors are identified.

[0081] (2) Risk assessment The probability analysis method is used to evaluate the subsidence risk under different conditions and generate corresponding risk levels.

[0082] Visualize risk maps to quickly identify high-risk areas.

[0083] user interface Visual display: Real-time monitoring data, settlement prediction results and risk assessment information are displayed in the form of charts and dashboards to facilitate quick decision-making by engineering managers.

[0084] Interactive function: Users can adjust model parameters as needed and view settlement prediction results under different conditions.

[0085] Use the user interface to achieve result output and feedback.

[0086] The settlement prediction results are presented through a visual interface, including predicted settlement curves, risk assessment results, sensitivity analysis diagrams, etc.

[0087] Provide real-time reporting function to regularly send settlement prediction and risk assessment content to relevant managers.

[0088] Application of the present invention Example 1: Urban underground parking lot project During the construction of an underground parking lot in a certain city, this system was used to monitor and predict settlement. Sensors were placed at different depths in the construction area, and through real-time data collection, rapid changes in groundwater levels were monitored. Using machine learning to analyze historical data, the system predicted the settlement risks that might occur under specific loads. Based on the prediction results, the construction team adjusted the construction sequence and reinforcement design, and ultimately successfully controlled the settlement within a safe range.

[0089] Example 2: High-rise building project In a high-rise building project, the system monitored the compaction of the soil and the structural load in real time. In the early stages of construction, the system predicted the risk of settlement due to rising water levels. Through fault tree analysis, key risk factors were identified and targeted emergency plans were developed. During the actual construction process, reinforcement measures were taken in advance to ensure the safety of the project. The embodiments should not be regarded as limitations of the present invention, but any improvements based on the spirit of the present invention should be within the scope of protection of the present invention.

Claims

1. A soft soil foundation post-construction settlement prediction and risk assessment system, characterized by: It includes: The data acquisition module is arranged at a key position of the soft soil foundation, and includes a soil moisture sensor for detecting the moisture inside the soil, a pressure sensor for detecting the pressure inside the soil, and a displacement sensor for detecting the displacement inside the soil; The data analysis module preprocesses the data collected by the data acquisition module and extracts key features from the preprocessed raw data as model input; The settlement prediction module includes a finite element model and a prediction model. The model input is input into the finite element model, the settlement results under different load conditions are output, and the corresponding settlement curve and prediction trend are output in combination with the prediction model; The risk assessment module outputs the corresponding risk level based on the prediction trend of the settlement prediction module and the set threshold.

2. The soft soil foundation post-construction settlement prediction and risk assessment system according to claim 1, characterized in that: The data acquisition module is also provided with a temperature sensor for collecting the internal pressure temperature of the soil.

3. The soft soil foundation post-construction settlement prediction and risk assessment system according to claim 1 or 2, characterized in that: The data analysis module realizes key feature extraction based on the following steps: Data preprocessing, including data denoising, outlier detection, and interpolation and filling operations; Feature extraction, including: Correlation analysis was used to screen out characteristic variables that were closely related to settlement; Principal component analysis is used to reduce the dimensionality of high-dimensional data and obtain the main influencing factors.

4. The soft soil foundation post-construction settlement prediction and risk assessment system according to claim 3, characterized in that: The data analysis module also includes a feature extraction model, which outputs model inputs required by the settlement prediction module based on the data collected by the data collection module.

5. The soft soil foundation post-construction settlement prediction and risk assessment system according to claim 4, characterized in that: The model inputs include soil cohesion, friction angle, and elastic modulus.

6. The soft soil foundation post-construction settlement prediction and risk assessment system according to claim 1, characterized in that: The finite element model is a three-dimensional finite element model. The specific operation steps of the three-dimensional finite element model are as follows:

1. By setting boundary conditions and initial conditions; 2. Select the corresponding material model according to the properties of the soil, and input the corresponding physical and mechanical parameters of the soil; 3. Simulate different load conditions during construction, conduct static or dynamic analysis, and obtain the stress-strain state of the soil; Fourth, the consolidation and settlement process of soil under applied load is obtained through finite element calculation.

7. The soft soil foundation post-construction settlement prediction and risk assessment system according to claim 6, characterized in that: The prediction model is based on the random forest model, Firstly, a 10-fold cross-validation random forest model was established based on the land subsidence and main controlling factors; From the training sample set Z = {(x1, y1), (x2, y2), …, (x n ,y n )}, randomly generate N training sets M1, M2, ..., M N ; Generate corresponding decision trees T1, T2, ..., T for each training set N ; Randomly select l attributes from all L attributes, l≪L as the split attribute set of the current node, and select the best split attribute as the node to split; The generated multiple decision trees form a random forest. For the test sample set Z, each decision tree is used for testing. Get the prediction results T1(x),T2(x),…,T N (x), where T(x) represents N T N The set of (x), i.e., the random forest prediction result; T N (x) represents the prediction result of a single decision tree.

8. The soft soil foundation post-construction settlement prediction and risk assessment system according to claim 1, characterized in that: It also has a visual interface for presenting settlement prediction results.

9. The soft soil foundation post-construction settlement prediction and risk assessment system according to claim 1, characterized in that: The settlement prediction results include predicted settlement curves, risk assessment results, and sensitivity analysis diagrams.

Images