A DNN-based urban lifeline settlement hazard assessment method and system

Abstract

The present invention discloses a method and system for assessing the hidden dangers of settlement of urban lifelines based on DNN, which relate to the technical field of urban settlement monitoring. The assessment method comprises the following steps: monitoring settlement of several areas in a city, collecting urban information of any area, and performing data processing to obtain a set of characteristic parameter sets; screening any characteristic parameters to obtain a set of urban deformation environment characteristics; obtaining settlement monitoring results of any area, building a deep neural network model for training, generating a settlement hidden danger assessment model, and assessing the settlement degree of any area; periodically assessing the settlement of any area, extracting influencing features that affect the settlement degree assessment results, and obtaining the impact amplitude generated by any influencing features; assessing the settlement degree of each target area in any city to be assessed; drawing an urban deformation monitoring susceptibility map, and identifying areas with abnormal hidden dangers.

Description

Technical Field

[0001] The present invention relates to the technical field of urban subsidence monitoring, and in particular to a method and system for assessing urban lifeline subsidence hazards based on DNN. Background Art

[0002] With rapid economic development, urbanization is advancing rapidly, urban areas are expanding rapidly, and the urban population is increasing exponentially. However, with various urban activities, urban subsidence is inevitable. The hidden danger of urban lifeline subsidence refers to the safety risks in urban infrastructure caused by ground subsidence or other geological changes. These hidden dangers may have a serious impact on the normal operation of the city and the safety of residents.

[0003] Therefore, urban subsidence monitoring is extremely important, and the increasingly complex urban environment makes the prediction of ground subsidence more difficult and cumbersome; traditional methods use machine learning methods such as ELM model and SOM model to fit the model, but the robustness and accuracy of the model are low, and it is not possible to accurately assess the city's lifeline and guarantee the safety of urban infrastructure. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for assessing the hidden dangers of urban lifeline subsidence based on DNN to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a DNN-based urban lifeline settlement hazard assessment method, the assessment method comprising the following steps:

[0006] Step S100: Subsidence monitoring is performed on several areas of the city. Urban information is collected from any area using information collection equipment. The collected urban information is processed to obtain a set of characteristic parameters. Correlation analysis is performed on any characteristic parameters to obtain a set of urban deformation environment characteristics.

[0007] Step S200: Obtain settlement monitoring results for any area, build and train a deep neural network model based on a set of urban deformation environment characteristics, and generate a settlement risk assessment model; and assess the settlement degree of the area according to the settlement risk assessment model;

[0008] Step S300: Periodically assessing the settlement of any area, analyzing the differences in urban deformation environmental characteristics of the area in different periods, and extracting influencing features that affect the settlement assessment results; based on the differences in the settlement assessment results of the area in different periods, the impact magnitude of any influencing feature is obtained;

[0009] Step S400: Extracting a set of urban deformation environment features of any city to be evaluated, and evaluating the settlement degree of each target area of ​​the city to be evaluated; based on the settlement degree evaluation results of each target area, drawing an urban deformation monitoring susceptibility map, and identifying areas with abnormal hidden dangers.

[0010] Furthermore, step S100 includes the following steps:

[0011] Step S101: Using the BeiDou satellite navigation system and interferometric synthetic aperture radar to monitor the city, and integrating the monitored data with the city map to generate a city deformation monitoring susceptibility distribution grid map; setting each pixel grid in the city deformation monitoring susceptibility distribution grid map as a region of the city; reading the deformation amount in any region and setting it as label data to obtain a label data set Y = {y(1), ..., y(c), ..., y(M)}, where y(c) is the cth label data and M is the total number of label data in the label data set;

[0012] Step S102: arbitrarily select an area, collect the city digital elevation model information, geological information and rail transit distribution information of the area respectively, quantify all the collected information to obtain a number of feature data, and generate a feature data set X = {x(1), ..., x(a), ..., x(N)}, wherein x(a) is the ath feature data and N is the total number of feature data in the feature data set; the city digital elevation model information includes elevation value, resolution, slope and slope direction; the geological information includes the distribution, thickness, lithologic physical and mechanical properties and hydrogeological parameters of the stratum; the rail transit distribution information includes road information and station information;

[0013] Step S103: Randomly extract the feature data set, and set the extracted random feature data set containing n1 feature data as X1={x ’ (1),…,x ’ (r),…,x ’ (n1)}, where if there is r-th feature data x in the random feature data set ’ (r) If data is missing, then randomly select n2 feature data from the feature data set and set the value of the bth feature data to z b , according to the formula:

[0014]

[0015] Calculate the average value z of random n2 feature data ave ; The average value z ave For the rth feature data x ’ (r) Fill in missing values;

[0016] Step S104: After all missing values ​​in the feature data set are supplemented, according to the formula:

[0017]

[0018] Among them, z a is the value of the ath feature data in the feature data set; the average value x of the feature data set is calculated ave and standard deviation σ; arbitrarily obtain the value z of the a-th feature data from the feature data set a , if |z a -x ave |>3σ, then the ath feature data is set as abnormal data and removed from the feature data set;

[0019] Step S105: Use the z-score normalization method to non-dimensionalize the processed feature data set, converting each feature data into a dimensionless pure value; eliminate the unit restriction in the data set and convert it into a dimensionless pure value to enhance the reliability of model training; set any feature data as a random variable X(Y) and any label data as a random variable Y(X), according to the formula:

[0020]

[0021] Among them, p(x,y) is the probability of random variables X(Y) and Y(X) occurring simultaneously, p(x) is the probability of random variable X(Y) occurring alone, and p(y) is the probability of random variable Y(X) occurring alone; calculate the mutual information I(x;y) and I(y;x) between any feature data and label data; set a mutual information threshold I max , if I(x;y)>I max , then the characteristic data is set as the urban deformation environment characteristics; after comparing all the characteristic data, an urban deformation environment characteristic set is obtained.

[0022] Furthermore, step S200 includes the following steps:

[0023] Step S201: Obtain a label dataset Y = {y(1), ..., y(c), ..., y(M)} and a feature dataset X = {x(1), ..., x(a), ..., x(N)}, and use the two datasets to build a deep neural network model; extract a set of urban deformation environment features, use any urban deformation environment features as input data of the deep neural network model, divide the deep neural network model into several layers, and obtain the output of each layer by calculating the weight, bias, and activation function, according to the formula:

[0024] α j =δ(Wj α j-1 +β j );

[0025] Among them, W j is the output weight of the jth layer, β j is the output bias of the jth layer, α j-1 is the output value of the j-1th layer, δ() is the model output function; the output value α of the jth layer is calculated j ;

[0026] Step S202: Select mean square error As the loss function, where α ’ is the predicted output result of the deep neural network model, y is any label data in the label data set, and the loss value E between any label data and the predicted output result is calculated. Back propagation is performed through the gradient descent method, and the weight W and bias β in the output formula of each layer are output to minimize the obtained loss value, thus generating a settlement hazard assessment model;

[0027] The deep neural network model is optimized by using the improved beetle whisker search algorithm. It is mainly divided into the deep neural network part and the beetle whisker search algorithm optimization part. The first part is the data preparation of the deep neural network module and the determination of the parameters of the deep neural network. Then the beetle whisker search algorithm optimization part is performed. According to the error value of the deep neural network training, it is repeatedly iterated to find the optimal initial weight. Then the deep neural network is retrained with the optimal initial weight to establish an urban lifeline settlement hazard assessment model. The use of the improved beetle whisker search algorithm to optimize the deep neural network can effectively prevent falling into the local optimal solution, accelerate the convergence speed of the model, and improve the computational efficiency.

[0028] Step S203: arbitrarily select an area, extract the set of urban deformation environment characteristics in the area, perform dimensionless processing on each urban deformation environment characteristic and input it into the settlement hazard assessment model, and output the settlement evaluation value score of the area; the settlement evaluation value is a predicted value after standardization, ranging from 0 to 1, and the score value reflects the possibility of settlement in the area.

[0029] Furthermore, step S300 includes the following steps:

[0030] Step S301: extracting an urban deformation environment feature set for any area every unit period, and obtaining a settlement evaluation value for the area in each unit period; setting the settlement degree evaluation performed in any unit period as one evaluation record, arbitrarily selecting two adjacent evaluation records and extracting the urban deformation environment feature sets for each of the two evaluation records;

[0031] Step S302: Compare any urban deformation environmental feature in the previous assessment record with the set of urban deformation environmental features in the next assessment record. If the urban deformation environmental feature does not exist in the set of urban deformation environmental features in the next assessment record, set the urban deformation environmental feature as an influencing feature to obtain an influencing feature set. Environmental features included in the previous assessment record but not included in the next assessment record indicate that the environmental features were only temporarily identified, resulting in a deviation in the settlement assessment, but are not actual environmental features. Therefore, such environmental features need to be excluded to obtain a more accurate assessment result.

[0032] Step S303: Obtain the set evaluation rules from the settlement hazard assessment model to obtain the numerical proportion of any urban deformation environment feature; obtain the numerical difference between the remaining urban deformation environment features excluding the influencing feature set in the two evaluation records, and set the numerical difference of the kth urban deformation environment feature as Δdata k , according to the formula:

[0033]

[0034] Calculate the difference ratio η of the k-th city deformation environment characteristics k ;

[0035] Step S304: Obtain the settlement assessment values ​​of the two assessment records to obtain the assessment difference Δscore, and calculate the assessment difference ratio τ = Δscore / score be , where score be is the settlement assessment value recorded in the previous assessment; arbitrarily select the sth influencing feature from the influencing feature set, and set the difference ratio of the sth influencing feature to η s , according to the formula:

[0036] θ s =η s ×τ;

[0037] Calculate the influence amplitude θ of the sth influencing feature s .

[0038] Furthermore, step S400 includes the following steps:

[0039] Step S401: Divide the city to be evaluated into several target areas, extract the urban deformation environment characteristics in any e-th target area, compare the urban deformation environment characteristics with the impact characteristics, and obtain several impact characteristics contained in the e-th target area;

[0040] Step S402: Input all the urban deformation environment characteristics in the e-th target area into the settlement hazard assessment model, and output the settlement evaluation value score of the e-th target area. e ; Set the influence amplitude of the s1th influencing feature in the eth target area to θ s1 , according to the formula:

[0041]

[0042] Among them, num e is the number of influencing features contained in the e-th target area; the actual settlement evaluation value score of the e-th target area is calculated ’ e ;

[0043] Step S403: Obtain the actual settlement evaluation value of any target area in the city to be evaluated, draw a city deformation monitoring susceptibility map, and set a settlement evaluation threshold score max , if score ’ e >score max , the e-th target area is judged to be an abnormal hidden danger area and is marked and presented in the urban deformation monitoring susceptibility map.

[0044] In order to better implement the above method, an urban lifeline settlement hidden danger assessment system is also proposed. The assessment system includes an urban information analysis module, a hidden danger model analysis module, a model deviation analysis module and an abnormal hidden danger judgment module.

[0045] The urban information analysis module is used to monitor the settlement of several areas in the city. It collects urban information from any area through information collection equipment, processes the collected urban information to obtain a set of characteristic parameters, and performs correlation analysis on any characteristic parameters to obtain a set of urban deformation environment characteristics.

[0046] The hidden danger model analysis module is used to obtain settlement monitoring results of any area, build and train a deep neural network model based on the urban deformation environment feature set, and generate a settlement hidden danger assessment model; and evaluate the settlement degree of the area according to the settlement hidden danger assessment model;

[0047] The model deviation analysis module is used to periodically evaluate the settlement of any area, analyze the differences in the urban deformation environment characteristics of the area in different periods, and extract the influencing features that affect the settlement assessment results; based on the differences in the settlement assessment results of the area in different periods, the impact magnitude of any influencing feature is obtained;

[0048] The abnormal hidden danger judgment module is used to extract the urban deformation environment feature set of any city to be evaluated and evaluate the settlement degree of each target area in the city to be evaluated; based on the settlement degree assessment results of each target area, it draws the urban deformation monitoring susceptibility map and identifies areas with abnormal hidden dangers.

[0049] Furthermore, the city information analysis module includes a city data collection unit and a characteristic parameter screening unit;

[0050] The urban data acquisition unit is used to monitor the settlement of several areas in the city, collect urban information from any area through information acquisition equipment, and process the collected urban information to obtain a set of characteristic parameter sets; the characteristic parameter screening unit is used to perform correlation analysis on any characteristic parameters and screen out a set of urban deformation environment characteristics.

[0051] Furthermore, the hidden danger model analysis module includes a hidden danger model construction unit and an urban hidden danger assessment unit;

[0052] The hidden danger model construction unit is used to obtain the settlement monitoring results of any area, build a deep neural network model and perform model training based on the urban deformation environment feature set, and generate a settlement hidden danger assessment model; the urban hidden danger assessment unit is used to assess the settlement degree of the area according to the settlement hidden danger assessment model.

[0053] Furthermore, the model deviation analysis module includes an impact feature extraction unit and an impact amplitude calculation unit;

[0054] The impact feature extraction unit is used to perform periodic evaluation of the settlement of any area, analyze the differences in urban deformation environment characteristics of the area in different periods, and extract the impact features that affect the settlement degree evaluation results; the impact amplitude calculation unit is used to obtain the impact amplitude generated by any impact feature based on the differences in settlement degree evaluation of the area in different periods.

[0055] Furthermore, the abnormal hidden danger judgment module includes an assessment result analysis unit and an abnormal hidden danger identification unit;

[0056] The assessment result analysis unit is used to extract the urban deformation environment feature set of any city to be assessed and to assess the degree of settlement of each target area in the city to be assessed; the abnormal hidden danger identification unit is used to draw an urban deformation monitoring susceptibility map based on the settlement degree assessment results of each target area and to identify areas with abnormal hidden dangers.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] 1. This invention uses the mutual information method to screen features that are highly correlated with the target variable and construct a training data set. It also uses an improved BAS-optimized deep neural network model, combined with the characteristics of the urban deformation environment, for training and optimization. The constructed urban settlement monitoring and evaluation model can accurately assess the settlement degree of the area to be evaluated and output the settlement probability value, providing a scientific and reliable basis for hidden danger assessment for urban planning and construction. It is stable, reliable, and highly accurate.

[0059] 2. The present invention optimizes the conventional deep neural network model and uses the optimized settlement hazard assessment model to accurately assess the hidden danger situation in the city. At the same time, taking into account the short-term impact of environmental factors on the assessment process, the assessment results affected by environmental factors are corrected, and the accuracy of the assessment results is further improved, which is conducive to the reliability of subsequent judgments. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram of the steps of a DNN-based urban lifeline settlement hazard assessment method;

[0061] Figure 2 This is a structural diagram of a DNN-based urban lifeline settlement hazard assessment system;

[0062] Figure 3 This is a flowchart of a DNN-based urban lifeline settlement hazard assessment method;

[0063] Figure 4 A flow chart for constructing a model for assessing the potential subsidence hazards of urban lifelines;

[0064] Figure 5 This is the susceptibility map for urban deformation monitoring. DETAILED DESCRIPTION

[0065] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0066] Example: Figures 1 to 5 As shown, the present invention provides a method for assessing the hidden dangers of urban lifeline settlement based on DNN, and the assessment method includes the following steps:

[0067] Step S100: Subsidence monitoring is performed on several areas of the city. Urban information is collected from any area using information collection equipment. The collected urban information is processed to obtain a set of characteristic parameters. Correlation analysis is performed on any characteristic parameters to obtain a set of urban deformation environment characteristics.

[0068] Wherein, step S100 includes the following steps:

[0069] Step S101: Using the BeiDou satellite navigation system and interferometric synthetic aperture radar to monitor the city, and integrating the monitored data with the city map to generate a city deformation monitoring susceptibility distribution grid map; setting each pixel grid in the city deformation monitoring susceptibility distribution grid map as a region of the city; reading the deformation amount in any region and setting it as label data to obtain a label data set Y = {y(1), ..., y(c), ..., y(M)}, where y(c) is the cth label data and M is the total number of label data in the label data set;

[0070] Step S102: arbitrarily select an area, collect urban digital elevation model information, geological information, and rail transit distribution information of the area, quantify all the collected information to obtain a number of feature data, and generate a feature data set X = {x(1), ..., x(a), ..., x(N)}, where x(a) is the ath feature data and N is the total number of feature data in the feature data set;

[0071] Example 1: Acquire key data within the subsidence monitoring area, including urban digital elevation model information, geological information, and rail transit distribution information. The urban digital elevation model information specifically includes key parameters such as elevation H, resolution R, slope θ, and slope aspect α, which can reflect the basic characteristics of the urban topography. The geological information includes the distribution D, thickness T, and lithology of strata, such as sandstone S, shale Sh, and limestone Ls, as well as physical and mechanical properties. Physical and mechanical properties include bulk density γ, elastic modulus E, Poisson's ratio v, cohesion c, and internal friction angle φ. Furthermore, hydrogeological parameters such as permeability K, water supply μ, water release coefficient S, and backflow coefficient ω are also included. Rail transit distribution information includes road information such as road width W and road type T, and station information such as station location P and station size S.

[0072] Step S103: Randomly extract the feature data set, and set the extracted random feature data set containing n1 feature data as X1={x ’ (1),…,x ’ (r),…,x ’ (n1)}, where if there is r-th feature data x in the random feature data set ’(r) If data is missing, then randomly select n2 feature data from the feature data set and set the value of the bth feature data to z b , according to the formula:

[0073]

[0074] Calculate the average value z of random n2 feature data ave ; The average value z ave For the rth feature data x ’ (r) Fill in missing values;

[0075] Step S104: After all missing values ​​in the feature data set are supplemented, according to the formula:

[0076]

[0077] Among them, z a is the value of the ath feature data in the feature data set; the average value x of the feature data set is calculated ave and standard deviation σ; arbitrarily obtain the value z of the a-th feature data from the feature data set a , if |z a -x ave |>3σ, then the ath feature data is set as abnormal data and removed from the feature data set;

[0078] Step S105: Use the z-score normalization method to non-dimensionalize the processed feature data set, and convert each feature data into a dimensionless pure value; set any feature data as a random variable X(Y) and any label data as a random variable Y(X), according to the formula:

[0079]

[0080] Among them, p(x,y) is the probability of random variables X(Y) and Y(X) occurring simultaneously, p(x) is the probability of random variable X(Y) occurring alone, and p(y) is the probability of random variable Y(X) occurring alone; calculate the mutual information I(x;y) and I(y;x) between any feature data and label data; set a mutual information threshold I max , if I(x;y)>I max , then the characteristic data is set as the urban deformation environment characteristics; after comparing all the characteristic data, an urban deformation environment characteristic set is obtained;

[0081] Example 2: The modeling process of the urban subsidence monitoring and evaluation model mainly includes the steps of model input and output, evaluation model parameter selection, selection of evaluation model activation function, model training, etc.; however, in the modeling process, there are many model parameters, and many parameters need to be set manually. Different parameter values ​​will have a great impact on the accuracy and applicability of the model, and will cause great difficulty in model prediction; therefore, in this embodiment, the improved longhorn beetle whisker search algorithm IBAS is used to optimize the initial weights of the DNN neural network, which can greatly improve the convergence speed of the DNN neural network and enhance the learning ability; at the same time, the Monte Carlo criterion in the simulated annealing algorithm SA is used to optimize and improve IBAS, jump out of the local optimum, and greatly improve the stability, and the improved IBAS is applied to urban lifeline hidden danger monitoring.

[0082] Step S200: Obtain settlement monitoring results for any area, build and train a deep neural network model based on a set of urban deformation environment characteristics, and generate a settlement risk assessment model; and assess the settlement degree of the area according to the settlement risk assessment model;

[0083] Wherein, step S200 includes the following steps:

[0084] Step S201: Obtain a label dataset Y = {y(1), ..., y(c), ..., y(M)} and a feature dataset X = {x(1), ..., x(a), ..., x(N)}, and use the two datasets to build a deep neural network model; extract a set of urban deformation environment features, use any urban deformation environment features as input data of the deep neural network model, divide the deep neural network model into several layers, and obtain the output of each layer by calculating the weight, bias, and activation function, according to the formula:

[0085] α j =δ(W j α j-1 +β j );

[0086] Among them, W j is the output weight of the jth layer, β j is the output bias of the jth layer, α j-1 is the output value of the j-1th layer, δ() is the model output function; the output value α of the jth layer is calculated j ;

[0087] Step S202: Select mean square error As the loss function, where α ’is the predicted output result of the deep neural network model, y is any label data in the label data set, and the loss value E between any label data and the predicted output result is calculated. Back propagation is performed through the gradient descent method, and the weight W and bias β in the output formula of each layer are output to minimize the obtained loss value, thus generating a settlement hazard assessment model;

[0088] Step S203: arbitrarily select an area, extract the urban deformation environment feature set in the area, perform dimensionless processing on each urban deformation environment feature and input it into the settlement hazard assessment model, and output the settlement evaluation value score of the area.

[0089] Step S300: Periodically assessing the settlement of any area, analyzing the differences in urban deformation environmental characteristics of the area in different periods, and extracting influencing features that affect the settlement assessment results; based on the differences in the settlement assessment results of the area in different periods, the impact magnitude of any influencing feature is obtained;

[0090] Wherein, step S300 includes the following steps:

[0091] Step S301: extracting an urban deformation environment feature set for any area every unit period, and obtaining a settlement evaluation value for the area in each unit period; setting the settlement degree evaluation performed in any unit period as one evaluation record, arbitrarily selecting two adjacent evaluation records and extracting the urban deformation environment feature sets for each of the two evaluation records;

[0092] Step S302: Compare any urban deformation environment feature in the previous evaluation record with the urban deformation environment feature set in the next evaluation record. If the urban deformation environment feature does not exist in the urban deformation environment feature set in the next evaluation record, set the urban deformation environment feature as an influencing feature to obtain an influencing feature set.

[0093] Step S303: Obtain the set evaluation rules from the settlement hazard assessment model to obtain the numerical proportion of any urban deformation environment feature; obtain the numerical difference between the remaining urban deformation environment features excluding the influencing feature set in the two evaluation records, and set the numerical difference of the kth urban deformation environment feature as Δdata k , according to the formula:

[0094]

[0095] Calculate the difference ratio η of the k-th city deformation environment characteristics k ;

[0096] Step S304: Obtain the settlement assessment values ​​of the two assessment records to obtain the assessment difference Δscore, and calculate the assessment difference ratio τ = Δscore / score be , where score be is the settlement assessment value recorded in the previous assessment; arbitrarily select the sth influencing feature from the influencing feature set, and set the difference ratio of the sth influencing feature to η s , according to the formula:

[0097] θ s =η s ×τ;

[0098] Calculate the influence amplitude θ of the sth influencing feature s .

[0099] Step S400: extracting a set of urban deformation environment features of any city to be assessed, and assessing the degree of settlement of each target area in the city to be assessed; based on the settlement degree assessment results of each target area, drawing an urban deformation monitoring susceptibility map to identify areas with abnormal hidden dangers;

[0100] Step S400 includes the following steps:

[0101] Step S401: Divide the city to be evaluated into several target areas, extract the urban deformation environment characteristics in any e-th target area, compare the urban deformation environment characteristics with the impact characteristics, and obtain several impact characteristics contained in the e-th target area;

[0102] Step S402: Input all the urban deformation environment characteristics in the e-th target area into the settlement hazard assessment model, and output the settlement evaluation value score of the e-th target area. e ; Set the influence amplitude of the s1th influencing feature in the eth target area to θ s1 , according to the formula:

[0103]

[0104] Among them, num e is the number of influencing features contained in the e-th target area; the actual settlement evaluation value score of the e-th target area is calculated ’ e ;

[0105] Step S403: Obtain the actual settlement evaluation value of any target area in the city to be evaluated, draw a city deformation monitoring susceptibility map, and set a settlement evaluation threshold score max , if score ’e >score max , the e-th target area is judged to be an abnormal hidden danger area and is marked and presented in the urban deformation monitoring susceptibility map.

[0106] An urban lifeline settlement hazard assessment system, comprising an urban information analysis module, a hazard model analysis module, a model deviation analysis module, and an abnormal hazard judgment module;

[0107] The urban information analysis module is used to monitor the settlement of several areas in the city. It collects urban information from any area through information collection equipment, processes the collected urban information to obtain a set of characteristic parameters, and performs correlation analysis on any characteristic parameters to obtain a set of urban deformation environment characteristics.

[0108] The hidden danger model analysis module is used to obtain settlement monitoring results of any area, build and train a deep neural network model based on the urban deformation environment feature set, and generate a settlement hidden danger assessment model; and evaluate the settlement degree of the area according to the settlement hidden danger assessment model;

[0109] The model deviation analysis module is used to periodically evaluate the settlement of any area, analyze the differences in the urban deformation environment characteristics of the area in different periods, and extract the influencing features that affect the settlement assessment results; based on the differences in the settlement assessment results of the area in different periods, the impact magnitude of any influencing feature is obtained;

[0110] The abnormal hidden danger judgment module is used to extract the urban deformation environment feature set of any city to be evaluated and evaluate the settlement degree of each target area in the city to be evaluated; based on the settlement degree assessment results of each target area, it draws the urban deformation monitoring susceptibility map and identifies areas with abnormal hidden dangers.

[0111] Among them, the city information analysis module includes a city data collection unit and a characteristic parameter screening unit;

[0112] The urban data acquisition unit is used to monitor the settlement of several areas in the city, collect urban information from any area through information acquisition equipment, and process the collected urban information to obtain a set of characteristic parameter sets; the characteristic parameter screening unit is used to perform correlation analysis on any characteristic parameters and screen out a set of urban deformation environment characteristics.

[0113] Among them, the hidden danger model analysis module includes the hidden danger model construction unit and the urban hidden danger assessment unit;

[0114] The hidden danger model construction unit is used to obtain the settlement monitoring results of any area, build a deep neural network model and perform model training based on the urban deformation environment feature set, and generate a settlement hidden danger assessment model; the urban hidden danger assessment unit is used to assess the settlement degree of the area according to the settlement hidden danger assessment model.

[0115] Among them, the model deviation analysis module includes an impact feature extraction unit and an impact amplitude calculation unit;

[0116] The impact feature extraction unit is used to perform periodic evaluation of the settlement of any area, analyze the differences in urban deformation environment characteristics of the area in different periods, and extract the impact features that affect the settlement degree evaluation results; the impact amplitude calculation unit is used to obtain the impact amplitude generated by any impact feature based on the differences in settlement degree evaluation of the area in different periods.

[0117] Among them, the abnormal hidden danger judgment module includes an assessment result analysis unit and an abnormal hidden danger identification unit;

[0118] The assessment result analysis unit is used to extract the urban deformation environment feature set of any city to be assessed and to assess the degree of settlement of each target area in the city to be assessed; the abnormal hidden danger identification unit is used to draw an urban deformation monitoring susceptibility map based on the settlement degree assessment results of each target area and to identify areas with abnormal hidden dangers.

[0119] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A DNN-based urban lifeline settlement hazard assessment method, characterized by: The evaluation method comprises the following steps: Step S100: Subsidence monitoring is performed on several areas of the city. Urban information is collected from any area using information collection equipment. The collected urban information is processed to obtain a set of characteristic parameters. Correlation analysis is performed on any characteristic parameters to obtain a set of urban deformation environment characteristics. Step S200: Obtain settlement monitoring results for any area, build and train a deep neural network model based on a set of urban deformation environment characteristics, and generate a settlement risk assessment model; and assess the settlement degree of the area according to the settlement risk assessment model; Step S300: Periodically assessing the settlement of any area, analyzing the differences in urban deformation environmental characteristics of the area in different periods, and extracting influencing features that affect the settlement assessment results; based on the differences in the settlement assessment results of the area in different periods, the impact magnitude of any influencing feature is obtained; Step S400: extracting a set of urban deformation environment features of any city to be assessed, and assessing the degree of settlement of each target area in the city to be assessed; based on the settlement degree assessment results of each target area, drawing an urban deformation monitoring susceptibility map to identify areas with abnormal hidden dangers; The step S400 includes the following steps: Step S401: Divide the city to be evaluated into several target areas, extract the urban deformation environment characteristics in any e-th target area, compare the urban deformation environment characteristics with the impact characteristics, and obtain several impact characteristics contained in the e-th target area; Step S402: Input all the urban deformation environment characteristics in the e-th target area into the settlement hazard assessment model, and output the settlement evaluation value score of the e-th target area. e ; Set the influence amplitude of the s1th influencing feature in the eth target area to θ s1 , according to the formula: Among them, num e is the number of influencing features contained in the e-th target area; the actual settlement evaluation value score' of the e-th target area is calculated e ; Step S403: Obtain the actual settlement evaluation value of any target area in the city to be evaluated, draw a city deformation monitoring susceptibility map, and set a settlement evaluation threshold score max , if score' e >score max , the e-th target area is judged to be an abnormal hidden danger area and is marked and presented in the urban deformation monitoring susceptibility map.

2. The method for assessing urban lifeline subsidence hazards based on DNN according to claim 1 is characterized by: The step S100 includes the following steps: Step S101: Using the BeiDou satellite navigation system and interferometric synthetic aperture radar to monitor the city, and integrating the monitored data with the city map to generate a city deformation monitoring susceptibility distribution grid map; setting each pixel grid in the city deformation monitoring susceptibility distribution grid map as a region of the city; reading the deformation amount in any region and setting it as label data to obtain a label data set Y = {y(1), ..., y(c), ..., y(M)}, where y(c) is the cth label data and M is the total number of label data in the label data set; Step S102: arbitrarily select an area, collect urban digital elevation model information, geological information, and rail transit distribution information of the area, quantify all the collected information to obtain a number of feature data, and generate a feature data set X = {x(1), ..., x(a), ..., x(N)}, where x(a) is the ath feature data and N is the total number of feature data in the feature data set; Step S103: Randomly extract the feature data set, and set the random feature data set containing n1 feature data to be X1 = {x'(1), ..., x'(r), ..., x'(n1)}, wherein, if the rth feature data x'(r) is missing in the random feature data set, then randomly extract n2 feature data from the feature data set, and set the value of the bth feature data to z b , according to the formula: Calculate the average value z of random n2 feature data ave ; The average value z ave Fill in the missing values ​​of the rth feature data x'(r); Step S104: After all missing values ​​in the feature data set are supplemented, according to the formula: Among them, z a is the value of the ath feature data in the feature data set; the average value x of the feature data set is calculated ave and standard deviation σ; arbitrarily obtain the value z of the a-th feature data from the feature data set a , if |z a -x ave |>3σ, then the ath feature data is set as abnormal data and removed from the feature data set; Step S105: Use the z-score normalization method to non-dimensionalize the processed feature data set, and convert each feature data into a dimensionless pure value; set any feature data as a random variable X(Y) and any label data as a random variable Y(X), according to the formula: Among them, p(x,y) is the probability of random variables X(Y) and Y(X) occurring simultaneously, p(x) is the probability of random variable X(Y) occurring alone, and p(y) is the probability of random variable Y(X) occurring alone; calculate the mutual information I(x;y) and I(y;x) between any feature data and label data; set a mutual information threshold I max , if I(x;y)>I max , then the characteristic data is set as the urban deformation environment characteristics; after comparing all the characteristic data, an urban deformation environment characteristic set is obtained.

3. The method for assessing urban lifeline subsidence hazards based on DNN according to claim 2 is characterized by: The step S200 includes the following steps: Step S201: Obtain a label dataset Y = {y(1), ..., y(c), ..., y(M)} and a feature dataset X = {x(1), ..., x(a), ..., x(N)}, and use the two datasets to build a deep neural network model; extract a set of urban deformation environment features, use any urban deformation environment features as input data of the deep neural network model, divide the deep neural network model into several layers, and obtain the output of each layer by calculating the weight, bias, and activation function, according to the formula: a j =δ(W j a j -1+b j ); Among them, W j is the output weight of the jth layer, β j is the output bias of the jth layer, α j-1 is the output value of the j-1th layer, δ() is the model output function; the output value α of the jth layer is calculated j ; Step S202: Select mean square error As the loss function, where α ’ is the predicted output result of the deep neural network model, y is any label data in the label data set, and the loss value E between any label data and the predicted output result is calculated. Back propagation is performed through the gradient descent method, and the weight W and bias β in the output formula of each layer are output to minimize the obtained loss value, thus generating a settlement hazard assessment model; Step S203: arbitrarily select an area, extract the urban deformation environment feature set in the area, perform dimensionless processing on each urban deformation environment feature and input it into the settlement hazard assessment model, and output the settlement evaluation value score of the area.

4. The method for assessing urban lifeline subsidence hazards based on DNN according to claim 3 is characterized by: The step S300 includes the following steps: Step S301: extracting an urban deformation environment feature set for any area every unit period, and obtaining a settlement evaluation value for the area in each unit period; setting the settlement degree evaluation performed in any unit period as one evaluation record, arbitrarily selecting two adjacent evaluation records and extracting the urban deformation environment feature sets for each of the two evaluation records; Step S302: Compare any urban deformation environment feature in the previous evaluation record with the urban deformation environment feature set in the next evaluation record. If the urban deformation environment feature does not exist in the urban deformation environment feature set in the next evaluation record, set the urban deformation environment feature as an influencing feature to obtain an influencing feature set. Step S303: Obtain the set evaluation rules from the settlement hazard assessment model to obtain the numerical proportion of any urban deformation environment feature; obtain the numerical difference between the remaining urban deformation environment features excluding the influencing feature set in the two evaluation records, and set the numerical difference of the kth urban deformation environment feature as Δdata k , according to the formula: Calculate the difference ratio η of the k-th city deformation environment characteristics k ; Step S304: Obtain the settlement assessment values ​​of the two assessment records to obtain the assessment difference Δscore, and calculate the assessment difference ratio τ = Δscore / score be , where score be is the settlement assessment value recorded in the previous assessment; arbitrarily select the sth influencing feature from the influencing feature set, and set the difference ratio of the sth influencing feature to η s , according to the formula: i s =the s ×t; Calculate the influence amplitude θ of the sth influencing feature s .

5. A city lifeline settlement hazard assessment system, configured to execute the DNN-based city lifeline settlement hazard assessment method according to any one of claims 1 to 4, characterized in that: The evaluation system includes a city information analysis module, a hidden danger model analysis module, a model deviation analysis module and an abnormal hidden danger judgment module; The urban information analysis module is used to monitor the settlement of several areas in the city. It collects urban information from any area through information collection equipment, processes the collected urban information to obtain a set of characteristic parameters, and performs correlation analysis on any characteristic parameters to screen and obtain a set of urban deformation environment characteristics. The hidden danger model analysis module is used to obtain settlement monitoring results of any area, build a deep neural network model based on the urban deformation environment feature set, perform model training, and generate a settlement hidden danger assessment model; and assess the settlement degree of the area according to the settlement hidden danger assessment model; The model deviation analysis module is used to periodically evaluate the settlement of any area, analyze the differences in urban deformation environment characteristics of the area in different periods, and extract the influencing features that affect the settlement assessment results; based on the differences in the settlement assessment results of the area in different periods, the impact amplitude of any influencing feature is obtained; The abnormal hidden danger judgment module is used to extract the urban deformation environment feature set of any city to be evaluated, and evaluate the settlement degree of each target area of ​​the city to be evaluated; based on the settlement degree assessment results of each target area, draw an urban deformation monitoring susceptibility map to identify areas with abnormal hidden dangers.

6. The urban lifeline subsidence hazard assessment system according to claim 5 is characterized by: The city information analysis module includes a city data collection unit and a characteristic parameter screening unit; The urban data acquisition unit is used to monitor the subsidence of several areas in the city, collect urban information from any area through information acquisition equipment, and process the collected urban information to obtain a set of characteristic parameters; The characteristic parameter screening unit is used to perform correlation analysis on any characteristic parameters to screen out a set of urban deformation environment characteristics.

7. The urban lifeline subsidence hazard assessment system according to claim 5 is characterized by: The hidden danger model analysis module includes a hidden danger model construction unit and an urban hidden danger assessment unit; The hidden danger model construction unit is used to obtain the settlement monitoring results of any area, build a deep neural network model based on the urban deformation environment feature set, and perform model training to generate a settlement hidden danger assessment model; the urban hidden danger assessment unit is used to assess the settlement degree of the area according to the settlement hidden danger assessment model.

8. The urban lifeline subsidence hazard assessment system according to claim 5 is characterized by: The model deviation analysis module includes an impact feature extraction unit and an impact amplitude calculation unit; The impact feature extraction unit is used to periodically evaluate the settlement of any area, analyze the differences in urban deformation environment characteristics of the area in different periods, and extract the impact features that affect the settlement degree assessment results; The impact amplitude calculation unit is used to obtain the impact amplitude generated by any impact feature based on the difference in the settlement degree assessment of the area in different periods.

9. The urban lifeline subsidence hazard assessment system according to claim 5 is characterized by: The abnormal hidden danger judgment module includes an assessment result analysis unit and an abnormal hidden danger identification unit; The evaluation result analysis unit is used to extract the urban deformation environment feature set of any city to be evaluated and evaluate the settlement degree of each target area of ​​the city to be evaluated; The abnormal hidden danger identification unit is used to draw an urban deformation monitoring susceptibility map based on the settlement degree assessment results of each target area and identify areas with abnormal hidden dangers.

Images