An urban road collapse disaster risk assessment method and system based on artificial intelligence

Abstract

The present invention provides an urban road collapse disaster risk assessment method and system based on artificial intelligence, comprising: step S1, obtaining collapse-related data in a target area; step S2, using the collapse-related data to construct an integrated collapse data database for the target area; step S3, labeling and classifying the data in the integrated collapse data database, dividing them into collapse occurrence data, collapse non-occurrence data and variable data, and forming a sample data set; step S4, based on the sample data set, using iterative reweighted least squares method to train a logistic regression model for road collapse risk assessment in the target area; step S5, using the trained logistic regression model, combined with relevant information of detected collapse hazards, to detect and assess the collapse risk of collapse hazards. The present invention can systematically, accurately and quickly assess the collapse risk of detected collapse hazards, and well meet the actual urban road collapse disaster risk assessment needs.

Description

Technical Field

[0001] The present invention relates to a road collapse disaster assessment method, in particular to an urban road collapse disaster risk assessment method based on artificial intelligence, and further to a system adopting the urban road collapse disaster risk assessment method based on artificial intelligence. Background Art

[0002] In recent years, cities of all sizes across the country have launched road subsidence detection work, successfully detected and treated a large number of subsidence risks, effectively improved road safety, and achieved remarkable results. However, behind these fruitful results, there are still some technical problems that need to be solved.

[0003] The detection of a large number of road collapse hazards in a short period of time has posed a huge challenge to the management of hidden dangers, and also exposed a series of problems in the management of hidden dangers, such as some high-risk detected hidden dangers cannot be managed in time, and some low-risk detected hidden dangers are over-managed, which leads to waste of resources. This fully reflects the problem of the lack of refined management of existing hidden danger management work. There is obviously a lack of effective risk assessment of hidden dangers, which makes the hidden danger management plan lack of professionalism and pertinence, resulting in some resources and hidden danger management forces being wasted on low-risk hidden dangers, while some high-risk hidden dangers have not been managed in time. In addition, the existing technology adopts extensive hidden danger management, which will obviously increase the cost of hidden danger management. In order to improve this situation, hidden danger management units urgently need a risk assessment method that can systematically, accurately and quickly evaluate the probability that detected collapse hidden dangers can form a collapse disaster, so as to provide a better basis for guiding hidden danger management units to scientifically and rationally formulate refined hidden danger management plans, thereby improving the effect of hidden danger management and controlling the cost of hidden danger management.

[0004] Although, there are a few methods for risk assessment of road collapse hazards in the prior art. However, the existing risk assessment methods all take a single collapse hazard as the assessment object. Based on the comprehensive detection of collapse hazards, combined with the surrounding environmental information, the risk level is determined and corresponding risk control countermeasures are proposed. The calculation process of this existing assessment method is cumbersome. The assessment process not only requires the clarification of the parameter information of the collapse hazard itself, but also requires the acquisition of adjacent facility information and surrounding environmental factor information. In actual applications, it is very difficult to obtain some underground facility information, and the acquisition process is time-consuming, which will lead to the overall low efficiency of risk assessment work.

[0005] However, the detected collapse hazards may cause road collapse disasters at any time. It is impossible to wait for the current inefficient risk assessment method to give the assessment results, and it is impossible to provide a data basis for guiding governance suggestions in a timely manner. In addition, when the number of collapse hazards detected in a short period of time increases sharply, the existing assessment methods are even more difficult to meet the actual needs of hazard risk assessment. Therefore, there is an urgent need for a method that can systematically, accurately and quickly assess the risk of detected collapse hazards to meet the actual needs of urban road collapse disaster risk assessment and provide a better basis for the detection and governance of urban road collapse disasters. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide an urban road collapse disaster risk assessment method based on artificial intelligence, which aims to systematically, accurately and quickly assess the risk of detecting collapse hazards, meet the actual urban road collapse disaster risk assessment needs, and provide a better basis for the detection and management of urban road collapse disasters. On this basis, it further relates to a system that adopts the urban road collapse disaster risk assessment method based on artificial intelligence.

[0007] In this regard, the present invention provides an urban road collapse disaster risk assessment method based on artificial intelligence, comprising the following steps:

[0008] Step S1, obtaining collapse-related data in the target area, wherein the collapse-related data includes any one or more of historical collapse data, underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data, and deep foundation pit construction data;

[0009] Step S2, using the collapse-related data to construct an integrated collapse information database of the target area;

[0010] Step S3, marking and classifying the data in the integrated collapse data database into collapse occurrence data, collapse non-occurrence data and variable data, and forming a sample data set;

[0011] Step S4, based on the sample data set, using iterative reweighted least squares method to train a logistic regression model for road collapse risk assessment in the target area;

[0012] Step S5, using the trained logistic regression model and combining the relevant information of the detected potential collapse hazard to detect and evaluate the collapse risk of the potential collapse hazard.

[0013] A further improvement of the present invention is that step S1 includes the following sub-steps:

[0014] Step S101, determining the range of the target area, and when the range of the target area is smaller than a preset range threshold, expanding the range of the target area;

[0015] Step S102, obtaining the collapse-related data, wherein the historical collapse data includes historical collapse disaster data and detected collapse hazard data.

[0016] A further improvement of the present invention is that step S2 includes the following sub-steps:

[0017] Step S201, cleaning the acquired collapse-related data;

[0018] Step S202, performing data conversion on the data cleaned in step S201;

[0019] Step S203, classifying the data converted in step S202 and integrating them into thematic layers;

[0020] Step S204, spatially aligning the thematic layers integrated in step S203 according to the spatial correlation between the layers, that is, the geographic coordinates, and superimposing them to form a subsidence map, and storing the subsidence map in the integrated subsidence data database.

[0021] A further improvement of the present invention is that step S201 includes the following sub-steps:

[0022] Step S2011, sorting the acquired collapse-related data, performing duplication detection according to a predefined duplication identification rule, and deleting data identified as duplicates;

[0023] Step S2012, grading the importance of data in different structure fields in the data table, formulating integrity indicators according to the missing ratio and field importance, and deleting data that does not meet the integrity indicators;

[0024] Step S2013, counting and identifying the error values ​​or abnormal values ​​in the data, and judging whether the error values ​​or abnormal values ​​can be corrected according to the preset correction rules, if so, correcting the error values ​​or abnormal values; if not, deleting the error values ​​or abnormal values;

[0025] Step S2014, verify the correlation of the collapse-related data from multiple sources, trace the source of the data when conflicting data is found, and adjust or delete the conflicting data.

[0026] A further improvement of the present invention is that step S202 includes the following sub-steps:

[0027] Step S2021, converting the non-numeric data in the collapse-related data into discrete numeric data according to a preset data conversion rule;

[0028] Step S2022, normalizing the converted discrete numerical data.

[0029] A further improvement of the present invention is that, in the preset data conversion rules, according to the data indicator type, indicator name and assignment basis, the non-numerical data in the collapse-related data is converted, the indicator type includes collapse intrinsic information, underground pipeline information, geological survey, meteorological data, underground construction and traffic load; the indicator names corresponding to the collapse intrinsic information include whether collapse, area, clearance and top depth, the indicator names corresponding to the underground pipeline information include pipeline type, first distance collapse distance and service life, the indicator names corresponding to the geological survey include soil type, the indicator names corresponding to the meteorological data include whether it is in the rainy season and the number of days after rain, the indicator names corresponding to the underground construction include second distance collapse distance, the indicator names corresponding to the traffic load include traffic flow data; the assignment basis of each indicator name in the indicator type has a value range of 0-1.

[0030] A further improvement of the present invention is that step S3 includes the following sub-steps:

[0031] Step S301, constructing a collapse data structure body with historical collapse data as the core, dividing the historical collapse data into collapse occurred data and collapse non-occurrence data, taking the collapse occurred data as positive samples and taking the collapse non-occurrence data as negative samples;

[0032] Step S302, classify underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data and deep foundation pit construction data as associated variable data, integrate the associated variable data into the collapse data structure, and in each of the collapse data structures, include at least one associated data other than historical collapse data, so as to form the sample data set.

[0033] A further improvement of the present invention is that step S4 includes the following sub-steps:

[0034] Step S401, according to the K-fold cross-validation method, the data set is divided into K mutually exclusive subsets of the same size, and all subsets are traversed in sequence, each time the current subset is used as a validation set, and all other subsets are used as training sets for training, where K is a natural number;

[0035] Step S402, based on the K training sets, using iterative reweighted least squares method to train K logistic regression models, and using the corresponding validation set to evaluate each trained model;

[0036] Step S403, after selecting the best logistic regression model according to the evaluation result, retain the parameters of the best logistic regression model, and retrain the best logistic regression model using the complete data set to obtain the final logistic regression model.

[0037] A further improvement of the present invention is that step S5 includes the following sub-steps:

[0038] Step S501, inputting collapse related information of detected collapse hazards, wherein the collapse related information includes collapse intrinsic information, and the collapse intrinsic information includes collapse location, area, top depth and clearance;

[0039] Step S502, based on the location information of the detected collapse hazard, calling the relevant information in the integrated collapse data database constructed in step S2, and improving the relevant information of the detected collapse hazard;

[0040] Step S503, inputting the improved relevant information of the collapse hazard into the logistic regression model trained in step S4, calculating the collapse risk coefficient, and combining the preset logic formula to obtain the collapse disaster probability corresponding to the hazard;

[0041] Step S504: outputting the collapse-related information improved by the integrated collapse data database and the collapse disaster probability calculated by the logistic regression model in the form of a text file.

[0042] The present invention also provides an urban road collapse disaster risk assessment system based on artificial intelligence, which adopts the urban road collapse disaster risk assessment method based on artificial intelligence as described above, and comprises:

[0043] A data acquisition subsystem is used to acquire collapse-related data in the target area, wherein the collapse-related data includes any one or more of historical collapse data, underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data, and deep foundation pit construction data;

[0044] A collapse database subsystem, using the collapse related data to construct an integrated collapse information database for the target area;

[0045] A sample annotation subsystem is used to annotate and classify the data in the integrated collapse data database into collapse occurrence data, collapse non-occurrence data and variable data, and form a sample data set;

[0046] A model training subsystem, based on the sample data set, uses iterative reweighted least squares method to train a logistic regression model for road collapse risk assessment in a target area;

[0047] The risk assessment subsystem uses the trained logistic regression model and the relevant information of the detected collapse hazards to detect and assess the collapse risk of the collapse hazards.

[0048] Compared with the prior art, the beneficial effects of the present invention are as follows: by utilizing various types of collapse-related data in the target area, an integrated collapse data database of the target area with historical collapses as the core data is constructed, so as to train a logistic regression model for road collapse risk assessment in the target area, thereby effectively ensuring the pertinence and accuracy of the logistic regression model; then, based on the relevant information of the detected collapse hazards, the collapse risk of the detected collapse hazards can be systematically, accurately and quickly assessed, and a better basis can be provided for timely guiding the disposal of the detected collapse hazards, thereby meeting the actual urban road collapse disaster risk assessment needs, reducing the waste of some resources and hazard management forces caused by the lack of assessment work, and avoiding collapse disasters caused by untimely disposal. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the working process of an embodiment of the present invention;

[0050] Figure 2 It is a system structure block diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In the description of the present invention, if it involves a description of orientation, for example, the orientation or position relationship indicated by "upper", "lower", "front", "back", "left", "right", etc. is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. If a certain technical feature is referred to as "set", "fixed", "connected", or "installed" on another technical feature, it can be directly set, fixed, or connected to the other technical feature, or it can be indirectly set, fixed, connected, or installed on the other technical feature.

[0052] In the description of the present invention, if "several" is involved, it means more than one; if "multiple" is involved, it means more than two; if "greater than", "less than", or "exceed" is involved, it should be understood as not including the number itself; if "above", "below", or "within" is involved, it should be understood as including the number itself. If "first", "second", etc. are involved, they should be understood as being used only to distinguish the names of the same or similar technical features, and should not be understood as implying / indicating the relative importance of the technical features, the number of the technical features, or the order of the technical features.

[0053] The preferred embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings:

[0054] like Figure 1 As shown, this embodiment provides an urban road collapse disaster risk assessment method based on artificial intelligence, comprising the following steps:

[0055] Step S1, obtaining various types of collapse-related data in the target area, wherein the collapse-related data includes any one or more of historical collapse data, underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data, and deep foundation pit construction data;

[0056] Step S2, after the original collapse-related data is cleaned, format converted, classified and integrated, an integrated collapse data database of the target area is constructed;

[0057] Step S3, marking and classifying the data in the integrated collapse data database into collapse occurrence data, collapse non-occurrence data and variable data, and forming a sample data set;

[0058] Step S4, based on the sample data set, using iterative reweighted least squares method to train a logistic regression model for road collapse risk assessment in the target area;

[0059] Step S5, using the trained logistic regression model and combining the relevant information of the detected potential collapse hazard to detect and evaluate the collapse risk of the potential collapse hazard.

[0060] Step S1 in this embodiment includes the following sub-steps:

[0061] Step S101, determining the range of the target area. When the range of the target area is less than the preset range threshold, it indicates that the data volume is too small. In order to ensure the accuracy and reliability of the collapse-related data, the range of the target area should be expanded. The preset range threshold refers to the preset range threshold of the target area, which can be customized and adjusted according to actual conditions and needs. The range of the target area can be expanded directly to the preset range threshold.

[0062] Step S102, obtaining the collapse-related data, wherein the historical collapse data includes historical collapse disaster data and detected collapse hazard data.

[0063] In the process of obtaining the collapse-related data, this embodiment can obtain collapse-related public information from libraries, archives, meteorological stations, geological institutes, etc. through offline queries and online crawlers. At the same time, after obtaining permission from relevant government departments, the collapse-related information recorded by relevant departments can also be synchronized.

[0064] The step S2 in this embodiment is used to realize the construction of an integrated collapse data database, and the process includes the following sub-steps:

[0065] Step S201, cleaning the acquired collapse-related data;

[0066] Step S202, performing data conversion on the data cleaned in step S201;

[0067] Step S203, classifying the data converted in step S202 and integrating them into thematic layers;

[0068] Step S204, spatially align the thematic layers integrated in step S203 according to the spatial correlation between the layers, i.e., the geographic coordinates, and superimpose them to form a collapse map, and store the collapse map in the integrated collapse data database. Specifically, the thematic layer refers to a data set that is geo-referenced by longitude and latitude coordinates; the implementation process of constructing an integrated collapse data database according to the spatial correlation between the layers is to spatially align the thematic layers based on longitude and latitude coordinates, such as spatially aligning the Shenzhen road network layer and the Shenzhen administrative planning layer, and superimposing the Shenzhen road network distribution map to obtain the corresponding collapse map.

[0069] More specifically, step S201 in this embodiment is used to implement data cleaning, and preferably includes the following sub-steps:

[0070] Step S2011, sorting the acquired collapse-related data, performing duplicate detection according to a predefined duplicate identification rule, and deleting the data identified as duplicate; the duplicate identification rule is used to define the repeatability index of the collapse-related data, avoiding using duplicate data to affect the training effect, and also avoiding increasing the data computing power requirement;

[0071] Step S2012, the importance of the data in different structural fields in the data table is graded, and the grading rules can be customized and adjusted according to the actual situation. The integrity index is formulated according to the missing ratio and the importance of the field, and the data that does not meet the integrity index is deleted; the data that does not meet the integrity index refers to the data whose integrity index does not reach the preset integrity threshold, and the preset integrity threshold can be set and adjusted according to the actual situation, so as to screen out the complete data and sort the importance level to improve the accuracy of the training; the integrity index is formulated according to the missing ratio and the importance of the field, and this process can be adjusted according to the actual situation and needs. For example, the importance level of the data field shown in the following table, the data meets the integrity requirement that the field with level 2 must not be missing, and at least one field with level 1 is not missing. Taking the collapse intrinsic information data as an example, when the collapse information is missing, even if other information exists, it is determined that the data does not meet the integrity index; when the collapse information exists, but the area, clearance and top depth fields do not exist, it is determined that the data does not meet the integrity index; when the collapse information exists, at least one of the area, clearance and top depth fields exists, it is determined that the data meets the integrity index.

[0072]

[0073] Step S2013, count and identify the error values ​​or abnormal values ​​in the data, such as identifying the error values ​​or abnormal values ​​in the data through statistical analysis, rules based on common sense, etc., and then determine whether the error values ​​or abnormal values ​​can be corrected according to the preset correction rules. If so, the error values ​​or abnormal values ​​are corrected; if not, the error values ​​or abnormal values ​​are deleted; the preset correction rules can also be set and adjusted according to actual needs. For example, a historical collapse area is displayed as 0, which is obviously an error value. If it is consistent with the data source display, it is deleted, otherwise it can be corrected; meteorological data from a certain place in Gansu Province shows that it was in the rainy season in January. According to historical data, the data is obviously an error value and can be modified based on historical data.

[0074] Step S2014, performs correlation verification on collapse-related data from multiple sources, traces the data when conflicting data is found, and adjusts or deletes the conflicting data. For example, when the pipeline archive data in a certain area shows that the pipeline type is a concrete pipe, but the actual exploration data shows that it is a reinforced concrete pipe, the erroneous data is adjusted; when the pipeline data near a historical collapse accident shows that the distance between the pipeline location and the collapse location is significantly different from the distance shown by the current exploration information, the erroneous data is deleted because the historical data can no longer be verified.

[0075] Step S202 in this embodiment is used to implement data conversion, and preferably includes the following sub-steps:

[0076] Step S2021, converting the non-numeric data in the collapse-related data into discrete numeric data according to a preset data conversion rule;

[0077] Step S2022, normalizing the converted discrete numerical data.

[0078] As shown in the following table, in the preset data conversion rules described in this embodiment, it is preferred to convert the non-numerical data in the collapse-related data according to the data indicator type, indicator name and assignment basis. The indicator type includes collapse intrinsic information, underground pipeline information, geological survey, meteorological data, underground construction and traffic load; the indicator names corresponding to the collapse intrinsic information include whether it is collapsed, area, clearance and top depth, the indicator names corresponding to the underground pipeline information include pipeline type, first distance collapse distance and service life, the indicator name corresponding to the geological survey includes soil type, the indicator name corresponding to the meteorological data includes whether it is in the rainy season and the number of days after rain, the indicator name corresponding to the underground construction includes the second distance collapse distance, and the indicator name corresponding to the traffic load includes traffic flow data; the assignment basis of each indicator name in the indicator type has a value range of 0-1. Of course, in actual applications, the data indicator type, indicator name and assignment basis can be adjusted according to actual conditions and needs.

[0079]

[0080] More specifically, step S203 in this embodiment is used to implement data classification and integration, and preferably includes the following sub-steps:

[0081] Step S2031, dividing the converted collapse-related data into six categories according to the above table: collapse intrinsic data, underground pipeline data, geological survey data, meteorological data, underground construction data and traffic load data;

[0082] Step S2032, associate the six categories of data with their related geographic location information, that is, associate them with the longitude and latitude coordinates, and build a GIS information thematic layer based on the coordinate information.

[0083] Step S3 in this embodiment is used to form a large-scale sample data set, and its implementation process includes the following sub-steps:

[0084] Step S301, constructing a collapse data structure body with historical collapse data as the core, dividing the historical collapse data into collapse occurred data and collapse non-occurrence data, taking the collapse occurred data as positive samples and taking the collapse non-occurrence data as negative samples;

[0085] Step S302, classifying underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data and deep foundation pit construction data as associated variable data, integrating the associated variable data into the collapse data structure based on geographic coordinates, that is, putting related data at the same location together to form the collapse data structure, and in each of the collapse data structures, at least one associated data other than historical collapse data is included, thereby forming the sample data set.

[0086] Step S4 in this embodiment is used to train the final logistic regression model, and the process preferably includes the following sub-steps:

[0087] Step S401, according to the K-fold cross-validation method, the data set is divided into K mutually exclusive subsets of the same size, and all subsets are traversed in sequence, each time the current subset is used as a validation set, and all other subsets are used as training sets for training, where K is a natural number;

[0088] Step S402, based on the K training sets, using iterative reweighted least squares method to train K logistic regression models, and using the corresponding validation set to evaluate each trained model;

[0089] Step S403, after selecting the best logistic regression model according to the evaluation result, retain the parameters of the best logistic regression model, and retrain the best logistic regression model using the complete data set to obtain the final logistic regression model.

[0090] Step S5 in this embodiment is used to perform risk assessment, and the process preferably includes the following sub-steps:

[0091] Step S501, inputting collapse-related information of detected collapse hazards into the urban road collapse disaster risk assessment system based on artificial intelligence, wherein the collapse-related information must include collapse intrinsic information, and other relevant information is optional; the collapse intrinsic information must include all information except the indicator of "whether it has collapsed", including collapse location, area, top depth and clearance;

[0092] Step S502, based on the location information of the detected collapse hazard, call the relevant information in the integrated collapse data database constructed in step S2 to improve the relevant information of the detected collapse hazard; for example, if the input information is only the intrinsic information of the detected collapse hazard, the output information will be improved to all the relevant information stored in the collapse database at the location of the collapse hazard, including historical collapse data, underground pipeline data, geological survey data, meteorological data, underground construction data and traffic load data.

[0093] Step S503, inputting the improved relevant information of the collapse hazard into the logistic regression model trained in step S4, calculating the collapse risk coefficient, and combining the preset logic formula to obtain the collapse disaster probability corresponding to the hazard;

[0094] Step S504, outputting the collapse-related information improved by the integrated collapse data database in step S502 and the collapse disaster probability calculated by the logistic regression model in step S503 in the form of a text file. Preferably, when outputting, the relationship between the collapse-related information and the collapse disaster probability is also displayed synchronously, so that the user can quickly obtain the data in the integrated collapse data database, providing a better basis for guiding the treatment of collapse hazards.

[0095] Step S503 in this embodiment is used to calculate the probability of collapse disaster, and preferably includes the following sub-steps:

[0096] Step S5031, the collapse hazard related information improved in step S502 is weighted averaged according to the importance level, and recorded as collapse intrinsic coefficient M1, underground pipeline coefficient M2, geological coefficient M3, meteorological coefficient M4, underground construction coefficient M5 and traffic load coefficient M6 respectively. The formula for weighted average calculation according to the importance level is coefficient = (type importance level * type value + distance to collapse importance level * distance to collapse) / (type importance level + distance to collapse importance level). For example, the underground pipeline information of a collapse hazard has a pipeline type of "other non-metallic material pipe", a distance to collapse of 2 meters, and an unknown service life. According to the data conversion table, the collapse hazard information has a pipeline type value of 0.75, a distance to collapse of 0.67, and a service life ignored. According to the importance level table, the pipeline type importance level is level 1, and the distance to collapse importance level is level 2. Therefore, the underground pipeline coefficient corresponding to the collapse hazard ;

[0097] Step S5032, based on the trained logistic regression model, combined with the M1-M6 coefficients calculated in the previous step, calculate the collapse risk coefficient y. For example, the logistic regression model trained based on a certain area is , where B is the correction factor, which defaults to 3.477 in this example and is affected by the missing coefficients in M2-M6, that is, the value of the missing coefficient is subtracted from the default value, as shown in the following table. When M4 and M5 data are missing, , substitute M1-M6 into it to calculate the collapse risk coefficient y;

[0098]

[0099] Step S5033: Based on the collapse risk coefficient y calculated in the previous step and the probability calculation formula, the collapse disaster probability can be obtained. .like Figure 2 As shown, this embodiment also provides an urban road collapse disaster risk assessment system based on artificial intelligence, which adopts the urban road collapse disaster risk assessment method based on artificial intelligence as described above, and includes:

[0100] A data acquisition subsystem is used to acquire collapse-related data in the target area, wherein the collapse-related data includes any one or more of historical collapse data, underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data, and deep foundation pit construction data;

[0101] A collapse database subsystem, using the collapse related data to construct an integrated collapse information database for the target area;

[0102] A sample annotation subsystem is used to annotate and classify the data in the integrated collapse data database into collapse occurrence data, collapse non-occurrence data and variable data, and form a sample data set;

[0103] A model training subsystem, based on the sample data set, uses iterative reweighted least squares method to train a logistic regression model for road collapse risk assessment in a target area;

[0104] The risk assessment subsystem uses the trained logistic regression model and the relevant information of the detected collapse hazards to detect and assess the collapse risk of the collapse hazards.

[0105] To sum up, this embodiment can utilize various types of collapse-related data in the target area to construct an integrated collapse data database of the target area with historical collapse as the core data, so as to train a logistic regression model for road collapse risk assessment in the target area, effectively ensuring the pertinence and accuracy of the logistic regression model; then, based on the relevant information of the detected collapse hazards, the collapse risk of the detected collapse hazards can be systematically, accurately and quickly evaluated, and a better basis can be provided for timely guiding the disposal of the detected collapse hazards, meeting the actual urban road collapse disaster risk assessment needs, reducing the waste of some resources and hazard management forces caused by the lack of assessment work, and avoiding collapse disasters caused by untimely disposal.

[0106] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. An urban road collapse disaster risk assessment method based on artificial intelligence, characterized in that: The following steps are involved: Step S1, obtaining collapse-related data in the target area, wherein the collapse-related data includes any one or more of historical collapse data, underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data, and deep foundation pit construction data; Step S2, using the collapse-related data to construct an integrated collapse information database of the target area; Step S3, marking and classifying the data in the integrated collapse data database into collapse occurrence data, collapse non-occurrence data and variable data, and forming a sample data set; Step S4, based on the sample data set, using iterative reweighted least squares method to train a logistic regression model for road collapse risk assessment in the target area; Step S5, using the trained logistic regression model and the relevant information of the detected collapse hazard to detect and evaluate the collapse risk of the collapse hazard; The step S2 comprises the following sub-steps: Step S201, cleaning the acquired collapse-related data; Step S202, performing data conversion on the data cleaned in step S201; Step S203, classifying the data converted in step S202 and integrating them into thematic layers; Step S204, spatially registering the thematic layers integrated in step S203 according to the spatial correlation between the layers, and superimposing them to form a subsidence map, and storing the subsidence map in an integrated subsidence data database; The step S202 includes the following sub-steps: Step S2021, converting the non-numeric data in the collapse-related data into discrete numeric data according to a preset data conversion rule; Step S2022, normalizing the converted discrete numerical data; In the preset data conversion rule, according to the data indicator type, indicator name and assignment basis, the non-numerical data in the collapse-related data are converted, and the indicator type includes collapse intrinsic information, underground pipeline information, geological survey, meteorological data, underground construction and traffic load; the indicator names corresponding to the collapse intrinsic information include whether it is collapsed, area, clearance and top depth, the indicator names corresponding to the underground pipeline information include pipeline type, first distance collapse distance and service life, the indicator names corresponding to the geological survey include soil type, the indicator names corresponding to the meteorological data include whether it is in the rainy season and the number of days after rain, the indicator names corresponding to the underground construction include the second distance collapse distance, and the indicator names corresponding to the traffic load include vehicle flow data; the assignment basis of each indicator name in the indicator type has a value range of 0-1; The step S3 comprises the following sub-steps: Step S301, constructing a collapse data structure body with historical collapse data as the core, dividing the historical collapse data into collapse occurred data and collapse non-occurrence data, taking the collapse occurred data as positive samples and taking the collapse non-occurrence data as negative samples; Step S302, classifying underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data and deep foundation pit construction data as associated variable data, integrating the associated variable data into the collapse data structure, and including at least one associated data other than the historical collapse data in each collapse data structure, so as to form the sample data set; The step S5 comprises the following sub-steps: Step S501, inputting collapse related information of detected collapse hazards, wherein the collapse related information includes collapse intrinsic information, and the collapse intrinsic information includes collapse location, area, top depth and clearance; Step S502, based on the location information of the detected collapse hazard, calling the relevant information in the integrated collapse data database constructed in step S2, and improving the relevant information of the detected collapse hazard; Step S503, inputting the improved relevant information of the collapse hazard into the logistic regression model trained in step S4, calculating the collapse risk coefficient, and combining the preset logic formula to obtain the collapse disaster probability corresponding to the hazard; Step S504: outputting the collapse-related information improved by the integrated collapse data database and the collapse disaster probability calculated by the logistic regression model in the form of a text file.

2. The urban road collapse disaster risk assessment method based on artificial intelligence according to claim 1 is characterized in that: The step S1 comprises the following sub-steps: Step S101, determining the range of the target area, and when the range of the target area is smaller than a preset range threshold, expanding the range of the target area; Step S102, obtaining the collapse-related data, wherein the historical collapse data includes historical collapse disaster data and detected collapse hazard data.

3. The urban road collapse disaster risk assessment method based on artificial intelligence according to claim 1 or 2 is characterized in that: The step S201 includes the following sub-steps: Step S2011, sorting the acquired collapse-related data, performing duplication detection according to a predefined duplication identification rule, and deleting data identified as duplicates; Step S2012, grading the importance of data in different structure fields in the data table, formulating integrity indicators according to the missing ratio and field importance, and deleting data that does not meet the integrity indicators; Step S2013, counting and identifying the error values ​​or abnormal values ​​in the data, and judging whether the error values ​​or abnormal values ​​can be corrected according to the preset correction rules, if so, correcting the error values ​​or abnormal values; if not, deleting the error values ​​or abnormal values; Step S2014, verify the correlation of the collapse-related data from multiple sources, trace the source of the data when conflicting data is found, and adjust or delete the conflicting data.

4. The urban road collapse disaster risk assessment method based on artificial intelligence according to claim 1 or 2 is characterized in that: The step S4 comprises the following sub-steps: Step S401, according to the K-fold cross-validation method, the data set is divided into K mutually exclusive subsets of the same size, and all subsets are traversed in sequence, each time the current subset is used as a validation set, and all other subsets are used as training sets for training, where K is a natural number; Step S402, based on the K training sets, using iterative reweighted least squares method to train K logistic regression models, and using the corresponding validation set to evaluate each trained model; Step S403, after selecting the best logistic regression model according to the evaluation result, retain the parameters of the best logistic regression model, and retrain the best logistic regression model using the complete data set to obtain the final logistic regression model.

5. An urban road collapse disaster risk assessment system based on artificial intelligence, characterized in that: The method for assessing urban road collapse disaster risk based on artificial intelligence as claimed in any one of claims 1 to 4 is adopted, and comprises: A data acquisition subsystem is used to acquire collapse-related data in the target area, wherein the collapse-related data includes any one or more of historical collapse data, underground pipeline data, geological survey data, meteorological observation data, traffic load data, subway construction data, and deep foundation pit construction data; A collapse database subsystem, using the collapse related data to construct an integrated collapse information database for the target area; A sample annotation subsystem is used to annotate and classify the data in the integrated collapse data database into collapse occurrence data, collapse non-occurrence data and variable data, and form a sample data set; A model training subsystem, based on the sample data set, uses iterative reweighted least squares method to train a logistic regression model for road collapse risk assessment in a target area; The risk assessment subsystem uses the trained logistic regression model and the relevant information of the detected collapse hazards to detect and assess the collapse risk of the collapse hazards.

Images