Old tunnel structure health diagnosis and utilization feasibility grading method and system

Abstract

The invention provides an old tunnel structure health diagnosis and utilization feasibility grading method and system, and relates to the technical field of tunnel infrastructure detection and evaluation. According to the invention, data is collected through multi-source equipment such as an Internet of Things sensor, a three-dimensional laser scanning device and a geological radar, a standardized evaluation process is constructed by combining data fusion and machine learning technologies, a traditional manual evaluation mode is converted into digital intelligent analysis, the evaluation efficiency and the result accuracy are greatly improved, and the method is suitable for large-scale popularization and application. And technical support is provided for rapid evaluation of a large number of old tunnels.

Description

Technical Field

[0001] This invention relates to the field of tunnel infrastructure inspection and assessment technology, and more specifically, to a method and system for diagnosing the health of old tunnel structures and classifying their feasibility of utilization. Background Technology

[0002] In the field of tunnel infrastructure inspection and assessment, the feasibility assessment of existing tunnels is a crucial step in ensuring the sustainable use of infrastructure. Currently, most existing methods for assessing the feasibility of existing tunnels rely on manual inspections combined with single inspection data, which has significant limitations: on the one hand, manual inspections are greatly influenced by the subjective experience of experts, and the judgment standards of different assessors are difficult to unify, resulting in insufficient objectivity of the assessment results; on the other hand, single inspection data cannot comprehensively reflect the overall condition of the tunnel structure, easily creating information silos, and the inspection and assessment process is inefficient, making it difficult to meet the actual needs of rapid screening and potential assessment of a large number of existing tunnels. Therefore, we propose an improvement, namely a method and system for the health diagnosis and utilization feasibility classification of existing tunnel structures. Summary of the Invention

[0003] This invention provides a method for health diagnosis and utilization feasibility classification of old tunnel structures, including: Step 1: Data Acquisition. Structural data of the old tunnel is collected using multi-source equipment. Step 2: Data preprocessing, which involves cleaning and standardizing the collected structural data; Step 3: Feature extraction. Key features are extracted from the preprocessed data for use as model input. Step 4: Data fusion. A feature-level fusion method is used to combine multi-source features into a unified feature set. Step 5: Train the machine learning model and establish a mapping model from features to feasibility classification; Step Six: Feasibility Classification. Based on the model output, generate a quantitative feasibility classification index, including: Select the level with the highest confidence level as the initial classification, and calculate the classification index S: ; Where i is the rank value (1, 2, 3), p i To correspond to the confidence level, the grading index S is a continuous value; The grading results are presented in chart form, including bar charts of feature importance and pie charts of confidence, to assist in decision-making; Step 7: System Deployment and Updates, Integrating the Trained Model into the Decision Support System, including: Real-time assessment: After new tunnel data is input, the system automatically executes steps one through six and outputs the classification results; Model updates: The model is incrementally learned every quarter using new data.

[0004] As a preferred technical solution in this application, step one includes the multi-source device and the collected structural data: Internet of Things (IoT) sensors are deployed inside the tunnel to monitor parameters in real time, including vibration, strain, displacement, temperature, and humidity. 3D laser scanning uses 3D laser scanning to acquire point cloud data of the tunnel's inner wall, which is used to identify cracks, spalling, and deformation. Ground-penetrating radar (GPR) uses high-frequency ground-penetrating radar to detect internal defects in the surrounding rock of tunnels, including loose zones, cavities, and water content.

[0005] As a preferred technical solution of this application, the structural data of the old tunnel in step one also includes historical data, which includes the tunnel's design drawings, construction records, and annual inspection reports. After data collection, the data is stored in a central database, and each tunnel sample is assigned a unique identifier.

[0006] As a preferred technical solution of this application, the data preprocessing in step two includes: Noise filtering involves using Kalman filtering to remove random noise from sensor data and a statistical outlier removal algorithm to eliminate noise points from point cloud data. Data alignment unifies multi-source data into the same coordinate system and timestamp, using time synchronization and spatial registration techniques. For missing values, multiple imputation or K-Nearest Neighbor (K-NN) based imputation algorithms are used to fill in missing data. Normalization performs min-max normalization on numerical data, scaling the values ​​to the [0,1] range.

[0007] As a preferred technical solution of this application, the specific steps of feature extraction in step three include: Geometric features are extracted from point cloud data: Principal component analysis and curvature calculation are used to extract crack features, including length, width, depth, and deformation rate. Extracting dynamic features from sensor data: calculating mean, variance, peak factor, and spectral features; Internal features are extracted from ground-penetrating radar data: internal defects are identified by using reflected wave amplitude, phase difference, and attenuation coefficient. Extract static features from historical data, including tunnel age, material strength, and past maintenance records; After feature extraction, a feature vector is formed, with each tunnel sample having a feature dimension of 50.

[0008] As a preferred technical solution in this application, the feature-level fusion method in step four includes: Fusion algorithm: Using a weighted fusion strategy, weights are assigned according to the feature importance; The weights are determined by the entropy weight method: Calculate the information entropy of each feature. The smaller the entropy value, the larger the weight. The fused feature set is expressed as: ; where, f i is the i-th feature, w i is the weight, ∑w i = 1, and n is the total number of features; Dimensionality reduction processing: To avoid the curse of dimensionality, the fused feature set is reduced in dimension using principal component analysis (PCA), retaining 95% of the variance and reducing the feature dimension to 30 dimensions.

[0009] As a preferred technical solution of this application, in step five, the method for establishing a mapping model from features to feasibility grading includes: Model selection: Adopt a hybrid model of support vector machine (SVM) and multi-layer perceptron (MLP); SVM model: Used to handle the linearly separable part, and the radial basis function (RBF) is selected as the kernel function; MLP model: Used to handle non-linear relationships, and the network structure includes an input layer, two hidden layers, and an output layer; Training data: Use the historical data set, which contains 500 labeled tunnel samples, and each sample has a feasibility level evaluated by experts. The feasibility level is divided into three levels; Training process: Divide the data set into a training set and a test set according to the ratio of 70:30. For the SVM model, use the training set for training; Model fusion: Perform weighted averaging on the prediction results of the SVM model and the MLP model. The weights are assigned according to the test set accuracy. The final output probability distribution P = [p1, p2, p3], corresponding to the confidence levels of the three levels respectively.

[0010] As a preferred technical solution of this application, in step five, the three levels of the feasibility level are: Level 1 is "can be directly utilized", Level 2 is "needs to be strengthened before utilization", and Level 3 is "not recommended for utilization"; Step six also includes threshold setting: According to actual needs, set the threshold interval: If S ≤ 1.5, output "can be directly utilized"; if 1.5 < S ≤ 2.5, output "needs to be strengthened before utilization"; if S > 2.5, output "not recommended for utilization".

[0011] As a preferred technical solution of this application, in step five, the training process uses 5-fold cross-validation to adjust the penalty parameter C and the kernel parameter γ. For the MLP model, the training set is used for training, the optimizer is Adam, the loss function is classification cross-entropy, the training period is 100, and the batch size is 32.

[0012] A system for diagnosing the health of old tunnel structures and classifying their utilization feasibility, used to implement methods for diagnosing the health of old tunnel structures and classifying their utilization feasibility, includes: Data acquisition module: integrates IoT sensors, 3D laser scanner, ground-penetrating radar, and historical data storage functions; Preprocessing and fusion module: used to perform data cleaning, standardization, weighted fusion, and dimensionality reduction; Model inference module: Used to deploy the trained hybrid model and output hierarchical results; Visualization module: used to generate feature importance bar charts and confidence pie charts; Update module: Train the model incrementally with new data each quarter and optimize parameters.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: In the scheme of this application: 1. This invention collects data from multiple sources, including IoT sensors, 3D laser scanning, and ground-penetrating radar. By combining data fusion and machine learning technologies, it constructs a standardized evaluation process, transforming the traditional manual evaluation model into digital intelligent analysis. This significantly improves evaluation efficiency and accuracy, providing technical support for the rapid evaluation of a large number of old tunnels. 2. This invention abandons the traditional method's over-reliance on expert experience, eliminates information bias through multi-source data complementarity, and establishes a mapping relationship between features and feasibility levels using quantitative analysis methods such as entropy weighting and machine learning algorithms. This makes the evaluation results data-driven, reduces subjective intervention, and enhances the credibility and consistency of the evaluation conclusions. 3. This invention can output standardized feasibility classification results and quantitative indices, providing unified data specifications and technical support for building a regional old tunnel database, thereby supporting the construction of a decision support system, helping to achieve overall planning and sustainable utilization of old tunnel resources, and promoting the scientific and refined management of infrastructure. 4. This invention adopts a multi-source data fusion strategy and a hybrid machine learning model, which can comprehensively cover multi-dimensional information such as surface defects, internal hidden dangers, dynamic response and historical status of tunnel structures, and adapt to the evaluation needs of old tunnels of different types and service years; at the same time, it supports incremental learning of the model, which can continuously optimize the model performance by incorporating new data, and adapt to the needs of tunnel engineering technology development and evaluation standard updates. Attached Figure Description

[0014] Figure 1 A flowchart of the method for health diagnosis and utilization feasibility classification of old tunnel structures provided in this application; Figure 2 The data preprocessing provided for this application includes flowcharts; Figure 3 A flowchart for feature extraction provided for this application. Detailed Implementation

[0015] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0016] It should be noted that, unless otherwise specified, the embodiments and features and technical solutions in the present invention can be combined with each other.

[0017] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0018] Example 1, please refer to Figure 1 A method for health diagnosis and utilization feasibility classification of old tunnel structures, including: Step 1: Data Acquisition. Structural data of the old tunnel is collected using multi-source equipment. Step 2: Data preprocessing, which involves cleaning and standardizing the collected structural data; Step 3: Feature extraction. Key features are extracted from the preprocessed data for use as model input. Step 4: Data fusion. A feature-level fusion method is used to combine multi-source features into a unified feature set. Step 5: Train the machine learning model and establish a mapping model from features to feasibility classification; Step Six: Feasibility Classification. Based on the model output, generate a quantitative feasibility classification index, including: Select the level with the highest confidence level as the initial classification, and calculate the classification index S: ; Where i is the rank value (1, 2, 3), p i To correspond to the confidence level, the grading index S is a continuous value, which facilitates fine-grained comparison; The grading results are presented in chart form, including bar charts of feature importance and pie charts of confidence, to assist in decision-making; Step 7: System Deployment and Updates, Integrating the Trained Model into the Decision Support System, including: Real-time assessment: After new tunnel data is input, the system automatically executes steps one through six and outputs the classification results; Model updates: The model is incrementally learned using new data every quarter to maintain its accuracy.

[0019] Furthermore, in step one, the multi-source devices and the acquired structural data include: The Internet of Things (IoT) sensors are deployed inside the tunnel to monitor parameters in real time, including vibration, strain, displacement, temperature, and humidity. The sampling frequency is 1Hz. The 1Hz sampling frequency of the IoT sensors can capture the dynamic structural response in real time, ensuring the timeliness of dynamic parameters. 3D laser scanning uses 3D laser scanning to acquire point cloud data of the tunnel inner wall with a resolution of no less than 5mm, which is used to identify cracks, spalling, and deformation. 3D laser scanning with a resolution of ≥5mm can accurately identify surface defects such as tiny cracks, providing high-precision data for appearance assessment. Ground-penetrating radar (GPR) uses high-frequency GPR (center frequency ≥ 1 GHz) to detect internal defects in the surrounding rock of tunnels. Internal defects include loose zones, cavities, and water content. The ≥ 1 GHz high-frequency GPR can penetrate the surrounding rock to detect internal defects, overcoming the limitations of surface detection. Different devices are specifically designed to cover different dimensions such as dynamic monitoring, surface inspection, and internal detection. The settings of key parameters (sampling frequency, resolution, center frequency) meet the accuracy and timeliness requirements of engineering inspection, and the complementary data from multiple dimensions ensures the integrity of the collected information.

[0020] Furthermore, the structural data of the old tunnels in Step 1 also includes historical data, which includes tunnel design drawings, construction records, and annual inspection reports. After data collection, the data is stored in a central database, and each tunnel sample is assigned a unique identifier. Historical data can reflect the tunnel's "inherent conditions" and performance evolution process. Combined with real-time data, it can more accurately determine the current health status of the structure. The central database and unique identifiers ensure centralized data storage and efficient retrieval, avoid data scattering and loss, and facilitate subsequent sample traceability and model training.

[0021] Furthermore, such as Figure 2 As shown, the data preprocessing in step two includes: Noise filtering involves using Kalman filtering to remove random noise from sensor data and a statistical outlier removal algorithm to eliminate noise points from point cloud data. Kalman filtering and statistical outlier removal are used to specifically filter noise from different types of data to ensure data authenticity. Data alignment unifies multi-source data into the same coordinate system (such as the local coordinate system of a tunnel) and timestamp, employing time synchronization and spatial registration techniques. Time and spatial alignment resolves the "spatiotemporal misalignment" problem of multi-source data, enabling data association and fusion. For missing value handling, multiple imputation or K-Nearest Neighbor (K-NN) based imputation algorithms are used to fill in missing data; multiple imputation or K-NN imputation avoids information loss caused by missing data; Normalization is a minimum-maximum normalization process used to scale numerical data (such as sensor readings) to the range [0,1].

[0022] Furthermore, such as Figure 3 As shown, the specific steps of feature extraction in step three include: Geometric features were extracted from point cloud data: Principal component analysis (PCA) and curvature calculation were used to extract crack features, including length, width, depth, and deformation rate. Extracting dynamic features from sensor data: calculating mean, variance, peak factor, and spectral features; Internal features are extracted from ground-penetrating radar data: internal defects are identified by using reflected wave amplitude, phase difference, and attenuation coefficient. Extract static features from historical data, including tunnel age, material strength, and past maintenance records; After feature extraction, a feature vector is formed, with each tunnel sample having a feature dimension of 50 (e.g., including 10 geometric features, 15 sensor features, 15 radar features, and 10 historical features).

[0023] Furthermore, the feature-level fusion method in step four includes: Fusion algorithm: Uses a weighted fusion strategy, assigning weights according to feature importance; The weights are determined using the entropy weighting method: the information entropy of each feature is calculated, and the smaller the entropy value, the larger the weight. The fused feature set is represented as follows: ; Among them, f i For the i-th feature, w i For the weights, ∑w i =1, where n is the total number of features; Dimensionality reduction: To avoid the curse of dimensionality, principal component analysis (PCA) is used to reduce the dimensionality of the fused feature set, retaining 95% of the variance and reducing the feature dimension to 30.

[0024] Furthermore, in step five, the methods for establishing a mapping model from features to feasibility classification include: Model Selection: A hybrid model of Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP) is adopted to balance classification accuracy and generalization ability; SVM Model: Used to handle linearly separable parts. The kernel function selects the Radial Basis Function (RBF), and the parameters are optimized through grid search; MLP Model: Used to handle non-linear relationships. The network structure includes an input layer (30 nodes), two hidden layers (64 and 32 nodes respectively, using the ReLU activation function), and an output layer (3 nodes, using the Softmax activation function); Training Data: Use the historical dataset, which contains 500 labeled tunnel samples. Each sample has a feasibility level evaluated by experts, and the feasibility level is divided into three levels; Training Process: Divide the dataset into a training set and a test set in a 70:30 ratio. For the SVM model, use the training set for training; Model Fusion: Perform weighted averaging on the prediction results of the SVM model and the MLP model. The weights are assigned according to the accuracy on the test set (for example, if the SVM accuracy is 85% and the MLP accuracy is 90%, the weights are 0.45 and 0.55 respectively). Finally, output the probability distribution P = [p1, p2, p3], corresponding to the confidence levels of the three levels respectively.

[0025] Furthermore, in step five, the three levels divided for the feasibility level are: level 1 is "can be directly utilized", level 2 is "needs to be strengthened before utilization", and level 3 is "not recommended for utilization"; Step six also includes threshold setting: According to actual requirements, set the threshold range: If S ≤ 1.5, output "can be directly utilized"; if 1.5 < S ≤ 2.5, output "needs to be strengthened before utilization"; if S > 2.5, output "not recommended for utilization".

[0026] Furthermore, in step five, the 5-fold cross-validation is adopted in the training process to adjust the penalty parameter C and the kernel parameter γ. For the MLP model, use the training set for training. The optimizer selects Adam, the loss function is categorical cross-entropy, the training period is 100, and the batch size is 32.

[0027] Example 2, used to implement the method for health diagnosis and utilization feasibility grading of old tunnel structures, includes: Data Acquisition Module: Integrates the functions of Internet of Things sensors, 3D laser scanners, ground-penetrating radar, and historical data storage; Preprocessing and Fusion Module: Used to perform data cleaning, standardization, weighted fusion, and dimensionality reduction; Model Inference Module: Used to deploy the trained hybrid model and output the grading results; Visualization Module: Used to generate bar charts of feature importance and pie charts of confidence levels; Update module: Train the model incrementally with new data each quarter and optimize parameters.

[0028] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a connection that allows communication between them; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0029] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.

Claims

1. A method for health diagnosis and utilization feasibility classification of old tunnel structures, characterized in that, include: Step 1: Data Acquisition. Structural data of the old tunnel is collected using multi-source equipment. Step 2: Data preprocessing, which involves cleaning and standardizing the collected structural data; Step 3: Feature extraction. Key features are extracted from the preprocessed data for use as model input. Step 4: Data fusion. A feature-level fusion method is used to combine multi-source features into a unified feature set. Step 5: Train the machine learning model and establish a mapping model from features to feasibility classification; Step Six: Feasibility Classification. Based on the model output, generate a quantitative feasibility classification index, including: Select the level with the highest confidence level as the initial classification, and calculate the classification index S: ； Where i is the rank value (1, 2, 3), p i To correspond to the confidence level, the grading index S is a continuous value; The grading results are presented in chart form, including bar charts of feature importance and pie charts of confidence, to assist in decision-making; Step 7: System Deployment and Updates, Integrating the Trained Model into the Decision Support System, including: Real-time assessment: After new tunnel data is input, the system automatically executes steps one through six and outputs the classification results; Model updates: The model is incrementally learned every quarter using new data.

2. The method for health diagnosis and utilization feasibility classification of old tunnel structures according to claim 1, characterized in that, In step one, the multi-source devices and the acquired structural data include: Internet of Things (IoT) sensors are deployed inside the tunnel to monitor parameters in real time, including vibration, strain, displacement, temperature, and humidity. 3D laser scanning is used to acquire point cloud data of the tunnel's inner wall to identify cracks, spalling, and deformation. Ground-penetrating radar (GPR) uses high-frequency ground-penetrating radar to detect internal defects in the surrounding rock of tunnels, including loose zones, cavities, and water content.

3. The method for health diagnosis and utilization feasibility classification of old tunnel structures according to claim 1, characterized in that, The structural data of the old tunnel in step one also includes historical data, which includes the tunnel's design drawings, construction records, and annual inspection reports. After data collection, the data is stored in a central database, and each tunnel sample is assigned a unique identifier.

4. The method for health diagnosis and utilization feasibility classification of old tunnel structures according to claim 1, characterized in that, Step two, data preprocessing, includes: Noise filtering involves using Kalman filtering to remove random noise from sensor data and a statistical outlier removal algorithm to eliminate noise points from point cloud data. Data alignment unifies multi-source data into the same coordinate system and timestamp, using time synchronization and spatial registration techniques. For missing values, multiple imputation or K-Nearest Neighbor (K-NN) based imputation algorithms are used to fill in missing data. Normalization performs min-max normalization on numerical data, scaling the values ​​to the [0,1] range.

5. The method for health diagnosis and utilization feasibility classification of old tunnel structures according to claim 1, characterized in that, The specific steps of feature extraction in step three include: Geometric features are extracted from point cloud data: Principal component analysis and curvature calculation are used to extract crack features, including length, width, depth, and deformation rate. Extracting dynamic features from sensor data: calculating mean, variance, peak factor, and spectral features; Internal features are extracted from ground-penetrating radar data: internal defects are identified by using reflected wave amplitude, phase difference, and attenuation coefficient. Extract static features from historical data, including tunnel age, material strength, and past maintenance records; After feature extraction, a feature vector is formed, with each tunnel sample having a feature dimension of 50.

6. The method for health diagnosis and utilization feasibility classification of old tunnel structures according to claim 1, characterized in that, The feature-level fusion methods in step four include: Fusion algorithm: Uses a weighted fusion strategy, assigning weights according to feature importance; The weights are determined by the entropy weight method: Calculate the information entropy of each feature. The smaller the entropy value, the greater the weight. The fused feature set is expressed as: ； Among them, f i For the i-th feature, w i For the weights, ∑w i =1, where n is the total number of features; Dimensionality reduction processing: To avoid the curse of dimensionality, the fused feature set is reduced in dimension using principal component analysis (PCA), retaining 95% of the variance and reducing the feature dimension to 30 dimensions.

7. The method for health diagnosis and utilization feasibility classification of old tunnel structures according to claim 1, characterized in that, In step five, the methods for establishing the mapping model from features to feasibility grading include: Model selection: Adopt a hybrid model of support vector machine (SVM) and multi-layer perceptron (MLP); SVM model: Used to handle the linearly separable part, and the radial basis function (RBF) is selected as the kernel function; MLP model: Used to handle non-linear relationships, and the network structure includes an input layer, two hidden layers, and an output layer; Training data: Use the historical data set, which contains 500 labeled tunnel samples, and each sample has a feasibility level evaluated by experts. The feasibility level is divided into three levels; Training process: Divide the data set into a training set and a test set in a ratio of 70:

30. For the SVM model, use the training set for training; Model fusion: Perform weighted averaging on the prediction results of the SVM model and the MLP model. The weights are assigned according to the test set accuracy. The final output probability distribution is P = [p1, p2, p3], corresponding to the confidence levels of the three levels respectively.

8. The method for health diagnosis and utilization feasibility classification of old tunnel structures according to claim 7, characterized in that, In step five, the three levels of feasibility grading are: Level 1 is "can be directly utilized", Level 2 is "needs to be strengthened before utilization", and Level 3 is "not recommended for utilization"; Step six also includes threshold setting: According to actual requirements, set the threshold interval: If S ≤ 1.5, output "can be directly utilized"; if 1.5 < S ≤ 2.5, output "needs to be strengthened before utilization"; if S > 2.5, output "not recommended for utilization".

9. The method for health diagnosis and utilization feasibility classification of old tunnel structures according to claim 7, characterized in that, In step five, the 5-fold cross-validation is used in the training process to adjust the penalty parameter C and the kernel parameter γ. For the MLP model, use the training set for training. The optimizer is selected as Adam, the loss function is categorical cross-entropy, the training period is 100, and the batch size is 32.

10. A system for diagnosing the health of old tunnel structures and classifying their utilization feasibility, used to implement the method for diagnosing the health of old tunnel structures and classifying their utilization feasibility as described in any one of claims 1-9, characterized in that, Include: Data acquisition module: Integrate the functions of Internet of Things sensors, 3D laser scanners, ground-penetrating radar, and historical data storage; Preprocessing and fusion module: Used to perform data cleaning, standardization, weighted fusion, and dimensionality reduction; Model inference module: Used to deploy the trained hybrid model and output the grading results; Visualization module: Used to generate a bar chart of feature importance and a pie chart of confidence; Update module: Incrementally train the model with new data every quarter and optimize the parameters.

Images