Urban road collapse safety risk intelligent diagnosis method

Abstract

The application provides a kind of urban road collapse safety risk intelligent diagnosis method, belongs to the technical field of urban lifeline engineering safety, comprising: determining the target area road network range, obtaining the detailed catalog of road collapse in target area and disaster factor data;Build road collapse disaster factor set, and carry out data preprocessing and multiple collinearity analysis;Build road collapse sample data set, divide it into training set and test set, and carry out data enhancement on training set;Using extreme gradient boosting algorithm to capture the nonlinear relationship between disaster factors, build urban road collapse prediction classification model;Based on the classification model, the probability of road collapse is predicted, and the road collapse risk distribution map of the target area is prepared;The constructed classification model is analyzed for interpretability, and the importance of each disaster factor and its effect on road collapse is evaluated.

Description

Technical Field

[0001] This invention belongs to the field of urban lifeline engineering safety technology, and in particular relates to an intelligent diagnostic method for urban road collapse safety risks. Background Technology

[0002] Urban road traffic infrastructure is a lifeline project ensuring normal daily operation and sustainable development. Currently, with the advancement of urbanization and global climate change, urban road collapse accidents are becoming increasingly frequent, posing a significant threat to urban public safety. The road collapse safety risk diagnosis technology proposed in this invention can provide strong technical support for management departments to formulate daily operation and maintenance and hazard investigation plans.

[0003] Due to the complexity and uncertainty of the urban environment, the causes of urban road collapse accidents are complex and diverse. Typically, collapse accidents are compound disasters—"urban ills"—caused by the coupled effects of natural and human factors. Currently, urban road collapse risk analysis mainly relies on physical methods and multi-criteria decision-making. However, existing methods are often limited by multiple assumptions, and multi-criteria decision-making rules rely excessively on expert subjective experience. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent diagnostic method for urban road collapse safety risks, which comprehensively and effectively captures the complex nonlinear relationships between various disaster-causing factors in urban road collapse accidents, and accurately and quickly identifies potential collapse risks in large-scale urban roads, so as to comprehensively explain the important influence and action law of each disaster-causing factor in the road collapse process, including the following steps: Step 1: Determine the scope of the road network in the target area and obtain detailed catalogs of road collapses and data on causative factors within the target area; Step 2: Construct a set of factors causing road collapse, and perform data preprocessing and multicollinearity analysis on the factors; Step 3: Construct a road collapse sample dataset, divide it into training and test sets according to the proportions, and perform data augmentation on the training set; Step 4: Use the extreme gradient boosting algorithm to capture the nonlinear relationship between disaster-causing factors and construct a prediction and classification model for urban road collapse; Step 5: Based on the constructed urban road collapse prediction and classification model, predict the probability of road collapse and compile a road collapse risk distribution map of the target area; Step 6: Perform interpretability analysis on the constructed classification model to assess the importance of each disaster-causing factor and its effect on road collapse.

[0005] Furthermore, step 1 specifically includes the following steps: Step 1.1: Divide the road network of the target area into multiple evaluation units according to intersections or road lengths in order to conduct a refined risk assessment; Step 1.2: Obtain a detailed catalog of road collapses; Step 1.3: Obtain a dataset of potential disaster factors that may cause road collapse.

[0006] Furthermore, in step 1.2, the detailed cataloging of road collapses includes, but is not limited to, the specific time, detailed location, and cause of the road collapse. It may further include information such as the accident process, consequences, and video data of the road collapse.

[0007] Furthermore, in step 1.3, the data on potential disaster-causing factors of road collapse includes, but is not limited to, one or more of the following: road network data, rail transit data, geological condition data, underground pipeline network data, underground engineering data, meteorological condition data, and surface water system data.

[0008] Furthermore, step 2 specifically includes the following steps: Step 2.1: Construct a set of disaster-causing factors for road collapse in the target area. The set of disaster-causing factors includes discrete disaster-causing factors and continuous disaster-causing factors. Step 2.2: Label and encode discrete disaster-causing factors to make them analyzable; standardize continuous disaster-causing factors to eliminate differences in scale and dimension. The standardization formula is as follows: ; in, X The value of the disaster-causing factor; This is the mean value of the disaster-causing factor; This represents the standard deviation of the causative factor. This is the standardized value of the disaster-causing factor.

[0009] Step 2.3: Evaluate the correlation between various disaster-causing factors and screen out strongly correlated disaster-causing factors to reduce the impact of multicollinearity on the model. The correlation coefficient is calculated using the following formula: ; in, The Spearman correlation coefficient; n The number of evaluation units; For the first i The difference in rank of disaster-causing factors in each evaluation unit.

[0010] Furthermore, discrete disaster-causing factors include road grade, soil type, drainage pipe type, and drainage pipe material.

[0011] Continuous disaster-causing factors include road length, road design speed, road network density, shallow sand layer thickness, shallow sand layer elevation, drainage pipe burial depth, drainage pipe service life, water supply pipe damage frequency, distance to nearby underground works, distance to adjacent subway lines, as well as distance to surrounding rivers and annual rainfall.

[0012] Furthermore, step 3 specifically includes the following steps: Step 3.1: Use the evaluation units that have experienced collapse accidents as positive samples and randomly sample the evaluation units that have not experienced collapse accidents as negative samples to construct a complete sample dataset; Step 3.2: Divide the complete sample dataset into a training set and a test set in a 4:1 ratio, with the training set accounting for 80% and the test set accounting for 20%. Step 3.3: The training set is augmented using a synthetic minority oversampling method. By synthesizing new collapsed positive samples, the number of positive and negative samples in the training set is balanced. The interpolation formula for the synthetic minority oversampling method is as follows: ; in, For new synthetic samples; For the selected minority class samples; for Neighboring samples; A random number in the range of 0 to 1.

[0013] Furthermore, step 4 specifically includes the following steps: Step 4.1: Train the extreme gradient boosting classification model using the enhanced training set, and optimize the model hyperparameters using Bayesian optimization; Step 4.2: Evaluate the trained classification model on the test set and calculate performance metrics based on the confusion matrix. Performance metrics include accuracy, precision, recall, and AUC value under the ROC curve.

[0014] Furthermore, step 5 specifically includes the following steps: Step 5.1: Use the probability value output by the classification model as the probability value of road collapse, and evaluate the collapse probability of all evaluation units in the target area. Step 5.2: Using the natural breakpoint method, all evaluation units are divided into Level I, Level II, Level III, Level IV and Level V based on their collapse probability from high to low; Step 5.3: Create a road collapse risk map based on the geographic information system, and assign different colors to each evaluation unit according to the collapse probability level for visualization display. Among them, Level I, Level II, Level III, Level IV and Level V correspond to red, orange, yellow, green and blue respectively.

[0015] Furthermore, step 6 specifically includes the following steps: Step 6.1: Calculate the gain value contributed by each disaster-causing factor at the split node of the extreme gradient boosting algorithm model, and quantitatively evaluate its importance in road collapse prediction. The gain importance of the disaster-causing factors is... The calculation formula is expressed as: ; in, Disaster-causing factors X The total number of times it is used as a split node in the model; Disaster-causing factors X In the i Information gain from each split node; Step 6.2: Plot the partial dependence curves of the disaster-causing factors to reveal the influence of these factors on the probability of road collapse. Partial dependency The calculation formula is expressed as: ; in, The total number of samples; For the first j Values ​​of each disaster-causing factor for each sample; Disaster-causing factors Specific values ​​that can be taken.

[0016] Compared with the prior art, the beneficial effects of the present invention are mainly reflected in: This invention integrates a multi-source heterogeneous dataset of road collapse causative factors and utilizes an extreme gradient boosting machine learning algorithm to effectively capture the complex nonlinear relationships between various road collapse causative factors, enabling accurate and rapid prediction of road collapse risks. It provides city managers with a detailed road collapse risk distribution map and guides proactive maintenance measures to improve the reliability and resilience of urban road traffic infrastructure.

[0017] This invention identifies the importance of various disaster-causing factors and their impact on the probability of road collapse, providing a practical method for city managers to identify key risk factors, optimize resource allocation, and conduct emergency response. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the intelligent diagnostic method for urban road collapse safety risks provided in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of the road network and road subsidence distribution provided in an embodiment of the present invention.

[0020] Figure 3A heat map showing the correlation between road collapse disaster-causing factors provided in an embodiment of the present invention.

[0021] Figure 4 This is a road collapse risk distribution map provided for an embodiment of the present invention.

[0022] Figure 5 This is a schematic diagram showing the importance ranking of road collapse disaster-causing factors provided in an embodiment of the present invention.

[0023] Figure 6 This is a schematic diagram of the partial dependence curve of the road collapse disaster-causing factors provided in an embodiment of the present invention. Detailed Implementation

[0024] The following will describe in more detail the intelligent diagnostic method for urban road collapse safety risks according to the present invention with reference to the schematic diagram, which illustrates the preferred embodiments of the present invention. It should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as being of general knowledge to those skilled in the art and is not intended to limit the present invention.

[0025] like Figure 1 As shown, a smart diagnostic method for urban road collapse safety risks includes the following steps: Step 1: Determine the scope of the road network in the target area and obtain detailed catalogs of road collapses and data on disaster-causing factors within the target area.

[0026] Taking the road network of a certain central urban area as an example, the road network is divided into 11,605 road segments, or 11,605 evaluation units, according to the intersections.

[0027] The road collapse catalogues obtained from the road traffic management department include the specific time, detailed location, and cause of the collapse. The road network and road collapse distribution provided in this embodiment of the invention are as follows: Figure 2 As shown.

[0028] Data on the factors contributing to road collapses were obtained from government departments and publicly available data sources, including road network data, rail transit data, geological condition data, underground pipeline network data, underground engineering data, meteorological data, and surface water system data.

[0029] Step 2: Construct a set of factors causing road collapse, and perform data preprocessing and multicollinearity analysis on the factors.

[0030] Based on the detailed cataloging of subsidence, a set of disaster-causing factors for road subsidence in the target area was constructed, including 4 discrete disaster-causing factors and 12 continuous disaster-causing factors, for a total of 16 disaster-causing factors. Among them, the discrete disaster-causing factors are road grade, soil type, drainage pipe type, and drainage pipe material, while the continuous disaster-causing factors are road length, road design speed, road network density, shallow sand layer thickness, shallow sand layer elevation, drainage pipe burial depth, drainage pipe service life, water supply pipe damage frequency, distance to adjacent underground works, distance to adjacent subway lines, distance to surrounding rivers, and annual rainfall.

[0031] Preprocessing of disaster-causing factor data includes labeling and encoding discrete disaster-causing factors to make them analyzable, and standardizing continuous disaster-causing factors to eliminate differences in scale and dimensions. The standardization formula is as follows: ; in, X The value of the disaster-causing factor; This is the mean value of the disaster-causing factor; This represents the standard deviation of the causative factor. This is the standardized value of the disaster-causing factor.

[0032] The correlation between various disaster-causing factors was assessed, and strongly correlated disaster-causing factors with an absolute correlation coefficient higher than 0.7 were screened out to reduce the impact of multicollinearity on the model. The correlation coefficient was calculated using the following formula: ; in, The Spearman correlation coefficient; n The number of evaluation units; For the first i The difference in rank of disaster-causing factors in each evaluation unit.

[0033] The correlation of disaster-causing factors provided in the embodiments of this invention is shown in the following figures. Figure 3 There is no strong correlation among the 16 disaster-causing factors.

[0034] Step 3: Construct a road collapse sample dataset, divide it into training and test sets according to the proportions, and perform data augmentation on the training set.

[0035] The 149 road sections that had experienced collapses were used as positive samples, and the 1000 road sections that had not experienced collapses were randomly sampled as negative samples, thus constructing a complete sample dataset with a sample size of 1149.

[0036] The complete sample dataset was divided into a training set and a test set in a 4:1 ratio. The number of positive samples and negative samples in the training set were 118 and 801, respectively, while the number of positive samples and negative samples in the test set were 31 and 199, respectively.

[0037] A synthetic minority class oversampling method was used to augment the training set. 683 new synthetic collapsed positive samples were generated through interpolation, balancing the number of positive and negative samples in the training set. The augmented training set now contains 801 positive and 801 negative samples. The interpolation formula for the synthetic minority class oversampling method is as follows: ; in, For new synthetic samples; For the selected minority class samples; for Neighboring samples; A random number in the range of 0 to 1.

[0038] Step 4: Use the extreme gradient boosting algorithm to capture the nonlinear relationship between disaster-causing factors and construct a prediction and classification model for urban road collapse.

[0039] Based on the enhanced training set, an extreme gradient boosting classification model was trained with a classification threshold of 0.5. The optimal model parameters after Bayesian optimization were: 100 trees, a maximum tree depth of 4, a learning rate of 0.2, a subsample ratio of 0.9, and a feature ratio of 0.8 per tree.

[0040] The trained extreme gradient boosting classification model was evaluated using a test set. Based on the confusion matrix, four performance metrics were calculated: accuracy, precision, recall, and AUC value under the ROC curve (the full value of the four performance metrics is 1). The results were 0.9608, 0.8235, 0.9032, and 0.9796, respectively. The evaluation results show that the constructed urban road collapse prediction classification model has excellent prediction performance.

[0041] Step 5: Use the constructed classification model to predict the probability of road collapse and compile a road collapse risk distribution map of the target area.

[0042] The probability values ​​output by the classification model were used as the probability values ​​of road collapse to assess the collapse probability of 11,605 evaluation units within the target area.

[0043] The natural breakpoint method was used to classify the collapse probability of all evaluation units into Level I, Level II, Level III, Level IV and Level V from high to low. The range of road collapse probability values ​​for each level is as follows: Level I (0.7877-1), Level II (0.5872-0.7877), Level III (0.3852-0.5872), Level IV (0.1972-0.3852) and Level V (0-0.1972). The number of evaluation units for each level is 1930 (Level I), 1882 (Level II), 1995 (Level III), 2487 (Level IV) and 3311 (Level V), respectively.

[0044] Road collapse risk mapping is performed based on a geographic information system. Different colors are assigned to each evaluation unit according to the collapse probability level for visualization: red (Level I), orange (Level II), yellow (Level III), green (Level IV), and blue (Level V). The road collapse risk distribution provided in this embodiment of the invention is as follows: Figure 4 As shown.

[0045] Step Six: Conduct interpretability analysis on the constructed classification model to assess the importance of each disaster-causing factor and its effect on road collapse.

[0046] The gain values ​​contributed by 16 disaster-causing factors at the split nodes of the extreme gradient boosting algorithm model are calculated to quantitatively evaluate their importance in road collapse prediction. The ranking of the importance of disaster-causing factor gains provided in this embodiment of the invention is shown below. Figure 5 The three most important disaster-causing factors are the number of times water supply pipelines are damaged, the service life of drainage pipelines, and the thickness of shallow sand layers. The importance of gain The calculation formula is as follows: ; in, Disaster-causing factors X The total number of times it is used as a split node in the model; Disaster-causing factors X In the i The information gain brought about by each split node.

[0047] Partial dependency curves were plotted for the three most important disaster-causing factors: the number of water supply pipeline damages, the service life of drainage pipelines, and the thickness of shallow sand layers. These curves reveal the patterns of their influence on the probability of road collapse. The partial dependency curves for these disaster-causing factors provided in this embodiment of the invention are shown below. Figure 6 The three most important disaster-causing factors were all significantly positively correlated with the probability of road collapse. Partial dependency The calculation formula is as follows: ; in, N The total number of samples; For the first j Values ​​of each disaster-causing factor for each sample; Disaster-causing factors Specific values ​​that can be taken.

[0048] In summary, the intelligent diagnostic method for urban road collapse safety risks provided by this invention is a practical and reliable technology for road collapse risk prevention and control. It does not rely on physical assumptions or expert experience, but utilizes an extreme gradient boosting intelligent algorithm to capture the complex nonlinear relationships between various disaster-causing factors of road collapse, generating a detailed road collapse risk distribution map, thus achieving accurate and rapid prediction of road collapse risks. Furthermore, this method accurately identifies the importance of various disaster-causing factors and their impact on the probability of road collapse, providing significant reference value for disaster prevention and mitigation of urban road infrastructure.

[0049] The above are merely preferred embodiments of the present invention and do not constitute any limitation on the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the technical solutions and content disclosed in the present invention without departing from the scope of the present invention shall be deemed to have remained within the protection scope of the present invention.

Claims

1. A method for intelligent diagnosis of safety risks associated with urban road collapse, characterized in that, Includes the following steps: S1: Determine the scope of the road network in the target area and obtain detailed catalogs of road collapses and data on disaster-causing factors within the target area; S2: Construct a set of factors causing road collapse, and perform data preprocessing and multicollinearity analysis on the factors; S3: Construct a road collapse sample dataset, divide it into training and test sets according to the proportions, and perform data augmentation on the training set; S4: The extreme gradient boosting algorithm is used to capture the nonlinear relationship between disaster-causing factors and to construct a prediction and classification model for urban road collapse. S5: Based on the constructed urban road collapse prediction and classification model, predict the probability of road collapse and compile a road collapse risk distribution map of the target area; S6: Conduct interpretability analysis on the constructed classification model to assess the importance of each disaster-causing factor and its effect on road collapse.

2. The intelligent diagnostic method for urban road collapse safety risks according to claim 1, characterized in that, S1 specifically includes the following steps: S11: Divide the road network of the target area into multiple evaluation units according to intersections or road lengths in order to conduct refined risk assessments; S12: Obtain a detailed catalog of road collapses; S13: Obtain data on potential disaster factors that could cause road collapse.

3. The intelligent diagnostic method for urban road collapse safety risks according to claim 2, characterized in that, In S12, the detailed catalog of road collapses includes the specific time, detailed location, and cause of the accident.

4. The intelligent diagnostic method for urban road collapse safety risks according to claim 2, characterized in that, In S13, the data on potential disaster-causing factors of road collapse are one or more sets of road network data, rail transit data, geological condition data, underground pipeline data, underground engineering data, meteorological data, and surface water system data.

5. The intelligent diagnostic method for urban road collapse safety risks according to claim 1, characterized in that, S2 specifically includes the following steps: S21: Construct a set of disaster-causing factors for road collapse in the target area, wherein the set of disaster-causing factors includes discrete disaster-causing factors and continuous disaster-causing factors; S22: Discrete disaster-causing factors are labeled and encoded to make them analyzable; continuous disaster-causing factors are standardized to eliminate differences in scale and dimension. The standardization formula is expressed as: ； in, X The value of the disaster-causing factor; This is the mean value of the disaster-causing factor; This represents the standard deviation of the causative factor. The value of the disaster-causing factor after standardization; S23: Assess the correlation between various disaster-causing factors, and screen out strongly correlated disaster-causing factors to reduce the impact of multicollinearity on the model. The correlation coefficient calculation formula is expressed as: ； in, The Spearman correlation coefficient; n The number of evaluation units; For the first i The difference in rank of disaster-causing factors in each evaluation unit.

6. The intelligent diagnostic method for urban road collapse safety risks according to claim 3, characterized in that, The discrete disaster-causing factors include road grade, soil type, drainage pipe type, and drainage pipe material; The continuous disaster-causing factors include road length, road design speed, road network density, shallow sand layer thickness, shallow sand layer elevation, drainage pipe burial depth, drainage pipe service life, water supply pipe damage frequency, distance to nearby underground works, distance to adjacent subway lines, as well as distance to surrounding rivers and annual rainfall.

7. The intelligent diagnostic method for urban road collapse safety risks according to claim 1, characterized in that, S3 specifically includes the following steps: S31: Use evaluation units that have experienced collapse accidents as positive samples and randomly sample evaluation units that have not experienced collapse accidents as negative samples to construct a complete sample dataset; S32: Divide the complete sample dataset into a training set and a test set in a 4:1 ratio, with the training set accounting for 80% and the test set accounting for 20%. S33: A synthetic minority oversampling method is used to augment the training set. By synthesizing new collapsed positive samples, the number of positive and negative samples in the training set is balanced. The interpolation formula for the synthetic minority oversampling method is expressed as: ； in, For new synthetic samples; For the selected minority class samples; for Neighboring samples; It is a random number in the range of 0-1.

8. The intelligent diagnostic method for urban road collapse safety risks according to claim 1, characterized in that, S4 specifically includes the following steps: S41: The enhanced training set is used to train the extreme gradient boosting classification model, and Bayesian optimization is used to optimize the model hyperparameters. S42: Evaluate the trained classification model on the test set and calculate performance metrics based on the confusion matrix. Performance metrics include accuracy, precision, recall, and AUC value under the ROC curve.

9. The intelligent diagnostic method for urban road collapse safety risks according to claim 1, characterized in that, S5 specifically includes the following steps: S51: Use the probability value output by the classification model as the probability value of road collapse, and evaluate the collapse probability of all evaluation units in the target area. S52: The natural breakpoint method is used to classify the collapse probability of all evaluation units into Level I, Level II, Level III, Level IV and Level V from high to low; S53: Based on the geographic information system, a road collapse risk map is created, and each evaluation unit is assigned a different color for visualization according to the collapse probability level. Among them, Level I, Level II, Level III, Level IV and Level V correspond to red, orange, yellow, green and blue, respectively.

10. The intelligent diagnostic method for urban road collapse safety risks according to claim 1, characterized in that, S6 specifically includes the following steps: S61: Calculate the gain value contributed by each disaster-causing factor at the split node of the extreme gradient boosting algorithm model, and quantify its importance in road collapse prediction. The gain importance of the disaster-causing factors is... The calculation formula is expressed as: ； in, Disaster-causing factors The total number of times it is used as a split node in the model; Disaster-causing factors In the Information gain from each split node; S62: Plot the partial dependency curves of the disaster-causing factors to reveal the influence of these factors on the probability of road collapse. Partial dependency The calculation formula is expressed as: ； in, The total number of samples; For the first j Values ​​of each disaster-causing factor for each sample; Disaster-causing factors Specific values ​​that can be taken.

Images