CN113128789B - Urban pavement collapse prevention method, system and storage medium based on probability prediction - Google Patents

Abstract

A urban pavement collapse prevention method based on probability prediction comprises the steps of 1, constructing a pavement collapse related index system, and performing dimension reduction treatment by adopting a principal component analysis method; 2. the classification of different road segment samples is realized by adopting an unsupervised clustering method, and a classification method with adjustable parameters based on density and distance is provided, so that the known road surface collapse accidents are classified; 3. defining the occurrence probability of common pavement collapse of one category, and providing the membership degree of each collected road section sample point in the same category to the pavement collapse probability of the category based on the maximum similarity; 4. and (3) reclassifying the categories according to similarity calculation of geological conditions and road construction quality, and constructing a machine learning prediction model aiming at sample points of each category of road sections to realize prediction of the road surface collapse probability under the running condition of different roads of the same category of road sections.

Description

Urban pavement collapse prevention method, system and storage medium based on probability prediction

Technical Field

The invention relates to the technical field of road collapse prediction, in particular to a method, a system and a storage medium for preventing urban road collapse based on probability prediction.

Background

The occurrence times of road collapse accidents are increased in recent years, and the road collapse accidents have burst performance, so that great hidden danger is brought to life and property safety of people and social public transportation safety. The prediction of the pavement collapse probability is beneficial to people to actively do the prevention work of the emergency, and has very important practical significance. Currently, research in this regard is very rare. The method for preventing the collapse of the urban road surface based on the probability prediction can predict the collapse probability of the road surface by a machine learning method, and further reduce the collapse probability of the road surface to be within a safety range by regulating and controlling the running condition of the road, so that the early prevention work of the collapse of the road surface is realized, and the collapse risk of the road surface is reduced.

Disclosure of Invention

The invention aims to provide a method, a system and a storage medium for preventing urban pavement collapse based on probability prediction.

The invention aims at realizing the technical scheme, and the method comprises the following specific steps:

1) Constructing a pavement index information set, wherein the pavement index set comprises road construction quality information, geological condition information, external environment information and running road condition information;

2) Processing and clustering index data of the pavement index information set, and dividing n road section samples into Q categories;

3) Collecting a road surface index information data set in the historical road surface subsidence accident, constructing a classification sample space, and classifying the road surface index information set in the historical road surface subsidence accident in the classification sample space according to the Q categories by adopting a distance and density-based method;

4) Calculating the pavement collapse probability of each category and the probability membership degree of each sample point;

5) Aiming at each category, respectively constructing a machine learning prediction model for predicting the pavement collapse probability;

6) Collecting road surface index information sets of road sections to be evaluated, determining road category j (j=1, 2, the..A) of the target road section by adopting a distance and density-based method in the step 3) according to geological conditions and road construction quality, setting probability membership degree, and adopting a machine learning prediction model LM obtained in the step 6) _j Predicting the road surface collapse probability of the road section;

7) Setting a safety probability threshold value of road surface subsidence, taking road running road conditions as regulation objects, and obtaining a running road condition range lower than the set safety probability threshold value through an intelligent optimization algorithm.

Further, the specific contents of the road construction quality information, the geological condition information, the external environment information and the running road condition information in the step 1) are as follows:

1-1) the road construction quality information includes:

road repair rateS _r S is the road repair area _a Is the total area of the road;

road condition indexThe weight representing the ith road damage type is mainly divided into crack type, deformation type and other types; omega represents the severity of road damage, the value range is 0 to 1, and the larger the value is, the more serious the damage is; sigma is the damage density;

road surface strength indexa ₀ ,a ₁ Is a constant;

pavement structural strength indexl _d ,l _s Respectively representing pavement design deflection, actually measured deflection, and units: millimeter;

road service life ind_5;

1-2) the geological condition information comprises:

the value range of the geological stability ind_6 is 0 to 1;

foundation bearing capacity ind_7=f _k +ε _b γ(b-3)+ε _d γ ₀ (d-0.5)，f _k Standard value (kN/m) representing bearing capacity of weak soil layer at bottom surface of cushion layer ² )，ε _b ,ε _d Bearing capacity correction factors representing foundation width and burial depth, respectively, b represents foundation width (m), d foundation burial depth (m), gamma substrate bottom weight (kN/m ³ )，γ ₀ -average bottom-up-weight (kN/m) of the substrate ³ )；

1-3) the external environment information comprises:

month average precipitation amount ind_8, month precipitation amount maximum variance ind_9, month average air temperature ind_10, month air temperature maximum variance ind_11, month average humidity ind_12 and month humidity maximum variance ind_13;

1-4) the running road condition information comprises:

average traffic flow ind_14, average traffic flow ind_15, and daily average load ind_16;

further, the specific step of processing and clustering the index data of the pavement index information set in the step 2) is as follows:

2-1) performing standardized processing on data ori_data= [ ind_1, ind_2, & gt, ind_16] of the road index information set, and performing dimension reduction by a principal component analysis method:

2-2) performing unsupervised cluster analysis on the reconstruction data by adopting a K-means algorithm, and classifying n road section samples into Q categories.

Further, the specific steps of the dimension reduction by the principal component analysis method in step 2-1) are as follows:

2-1-1) calculating a correlation coefficient matrix corr _16×16 ；

2-1-2) calculating the eigenvalue lambda of the correlation coefficient matrix _i (i=1, 2,., 16) and feature vector η _i (i＝1,2,...,16)；

2-1-3) arranging the characteristic values in order from large to small, and calculating the cumulative contribution degree of the main component corresponding to the t (t=1, 2,., 16) characteristic values before calculation

2-1-4) when the cut _t When the value is just equal to or greater than 0.95, the characteristic vectors corresponding to the first t principal components are taken to form a transformation matrix tran= [ eta ] ₁ ,η ₂ ,...,η _t ]Wherein eta _t ＝[e ₁ ,e ₂ ,...,e ₁₆ ] ^T ，data＝ori_data×tran＝[ind_1,ind_2,...,ind_t]The reconstructed data after dimension reduction; then the t-th principal component data is ind_t= [ a ] _1t ,a _2t ,...,a _nt ] ^T ，a _nt Nth sample data representing the nth principal component.

Further, the specific steps of performing unsupervised cluster analysis on the reconstructed data by using the K-means algorithm in the step 2-2) are as follows:

2-2-1) setting the number Q and the minimum threshold value of clustering target categories, and randomly selecting Q sample points from the data set data to serve as the center of the category clusters;

2-2-2) assigning each sample point to the cluster center closest to the cluster;

2-2-3) calculating the average value of all samples of each cluster, taking the average value as a new cluster center, and calculating the distances between the centers of mass of the clusters before and after updating;

2-2-4) if the center distance between the front cluster and the rear cluster is greater than or equal to a set threshold value, returning to the step 2-2-2); and if the center distance between the front cluster and the rear cluster is smaller than the set threshold, stopping calculation to obtain a clustering result.

Further, in the step 3), the specific steps of clustering the road surface index information set in the historical road surface subsidence accident according to the Q categories in the classification sample space by adopting the distance and density-based method are as follows:

3-1) collecting road surface index information data event_data in historical road surface subsidence accidents _s×16 Reconstructing all road surface collapse accident index information through a transformation matrix tran to obtain e_data=event_data×tran= [ y ] ₁ ,y ₂ ,...,y _i ,...,y _t ]Wherein y is _i ＝[y _1i ,y _2i ,...,y _si ] ^T S represents the total number of pavement collapse accident samples;

3-2) constructing a t-dimensional coordinate system by using t indexes, and constructing a classified sample point space;

3-3) calculating cluster center coordinates of each cluster of the sample point space, and then the k (k=1, 2,.,. Q) th cluster center coordinates are:

wherein ind_i _kj Data representing an ith index of a jth sample in a kth class cluster, m representing the total number of sample points of the kth class cluster;

3-4) calculating various pavement collapse sample points in the historical pavement collapse accident to various clustersThe distance between the centers is then the (y) th (h=1, 2,..s) pavement collapse sample point _h1 ,y _h2 ,...,y _hi ,...,y _ht ) Then the sample point is to the Center of the cluster _k The distance of (2) can be noted as:

then the h (h=1, 2,., s) pavement collapse sample points are spaced D from the center of all clusters of the class _h ＝[d _h1 ,d _h2 ,...,d _hk ,...,d _hQ ]；

The initial radius r is used for counting how many sample points exist in each type of clusters in the radius r by taking the h (h=1, 2,.. S) pavement collapse sample points as the circle center, and recording the number of the k type of sample points as num _k If the number of the sample points adjacent to the h pavement collapse sample point within the radius r is Num _h ＝[num _h1 ,num _h2 ,...num _hk ,...,num _hQ ]；

The effective value of the radius r should be not greater than the maximum value of the linear distance between the h (h=1, 2..s) th pavement collapse sample point and the sample point of each sampling road section, and not less than the minimum value thereof;

in order to make the number of samples of a certain class nearest to the center of the class cluster and within a set radius be the most, num is calculated _h And D _h Normalized to num_ gui _h And D_ gui _h The degree of attribution is calculated according to the following formula:

Attribution _h ＝α×Num_gui _h -β×D_gui _h ＝[at _h1 ,at _h2 ,...,at _hi ,...,at _hQ ]

wherein, alpha, beta epsilon [0,1] is a weight coefficient, and satisfies alpha+beta=1, and the specific weight value needs to be determined according to the actual data distribution; the more concentrated the data class cluster distribution and the larger the class spacing, the larger the beta value; if the data clusters are distributed more dispersedly and the class spacing is smaller, the alpha value is larger;

selection of Attribute _h At of the medium maximum value _hx The subscript x is the hClass number to which the pavement collapse sample points belong.

Further, the specific steps of calculating the road surface collapse probability of each category and the probability membership degree of each sample point in the step 4) are as follows:

4-1) record k (k=1, 2., Q) presence of P in categories _k Accident of road collapse is counted as N _k ，ε _k For a proportionality constant less than 1, there is a general probability that the road segment of that class collapses:

if P _k Taking P if =0 _k =0.01, i.e. there is a probability of collapse for any road segment, guarantee prob _k Is not 0;

4-2) calculating the similarity between each non-collapse accident sample point and each collapse accident sample point in the kth category, and obtaining a similarity matrix as follows:

b in matrix _ij Representing the similarity of the ith collapse accident sample point and the jth non-collapse accident sample point, wherein m represents the total number of non-collapse accident sample points in the kth category;

4-3) pair sim matrix _k Maximum value is found for each column of (2) to obtain mu_k _max ＝[b _max1 ,b _max2 ,...,b _maxj ,b _maxm ]，b _maxj The occurrence probability of the sample point representing the j non-collapse accident in the k category is prob _k Is a membership of (1).

Further, the specific steps of constructing the machine learning prediction model for predicting the pavement collapse probability in the step 5) are as follows:

5-1) calculating the similarity of all sample points according to geological conditions and road construction quality, namely index ind_1-ind_7 information, classifying the sample points with similarity higher than 0.9 into one class, and recording to obtain A numbersCategory, and i (i=1, 2,., a.) the total number of category samples is count _i ；

5-2) if count _i (i=1, 2.,. A) is not less than set_N1, directly taking index information ind_8-ind_16 of the sample points and probability membership degree as inputs, taking corresponding road surface collapse probability as output, and constructing a machine learning model LM for predicting the road surface collapse probability of the ith road _i ；

5-3) if count _i (i＝1,2,...,A)<Set_n1 and count _i (i=1, 2.,. The A) is more than or equal to set_N2, the data Set size can be enlarged through a data enhancement method, then index information ind_8-ind_16 of the expanded sample points and probability membership are used as inputs, corresponding road surface collapse probability is used as output, and a machine learning model LM for predicting the road surface collapse probability of the ith class of road is constructed _i ；

5-4) if count _i (i＝1,2,...,A)<Set_n2, the sample data is too little to perform modeling analysis, and is regarded as noise data.

Further, a urban road surface collapse prevention system based on probability prediction, comprising:

the road surface index information set construction module is used for constructing a road surface index information set;

the clustering module is used for processing and clustering the index data of the pavement index information set and classifying n road section samples into Q categories;

the historical data processing module is used for collecting road surface index information data sets in historical road surface subsidence accidents and classifying the road surface index information data sets according to the Q categories;

the class collapse probability calculation module is used for calculating the pavement collapse probability of each class and the probability membership degree of each sample point;

the prediction model learning module is used for calculating the pavement collapse probability of each category and the probability membership degree of each sample point;

the target prediction module is used for collecting road surface index information sets of road sections to be evaluated and predicting the road surface collapse probability of the road sections;

and the collapse prevention module is used for setting the running road condition range of the safety probability threshold value.

Further, a storage medium stores instructions adapted to be loaded by a processor to perform the urban road surface collapse prevention method.

Due to the adoption of the technical scheme, the invention has the following advantages:

1. constructing a pavement collapse related index system, and adopting a principal component analysis method to perform dimension reduction treatment on the pavement collapse related index system;

2. the classification of different road segment samples is realized by adopting an unsupervised clustering method, and a classification method with adjustable parameters based on density and distance is provided, so that the known road surface collapse accidents are classified;

3. defining the occurrence probability of common pavement collapse of one category, and providing the membership degree of each collected road section sample point in the same category to the pavement collapse probability of the category based on the maximum similarity;

4. reclassifying the categories according to similarity calculation of geological conditions and road construction quality, constructing a machine learning prediction model aiming at sample points of each category of road sections, and realizing prediction of road surface collapse probability under the running condition of different roads of the same category of road sections;

5. setting the collapse probability of the safety road surface, solving the running state of the safety road through an intelligent optimization algorithm, and preventing the collapse accident of the road surface.

Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof.

Drawings

The drawings of the present invention are described below.

FIG. 1 is a schematic flow chart of the present invention;

fig. 2 is a schematic diagram of a classification flow of a road surface subsidence accident.

Detailed Description

The invention is further described below with reference to the drawings and examples.

Embodiment one: the sample data set collected is assumed to be as shown in the following table (unified for each index unit):

table one: collected road segment sample set

And (II) table: pavement collapse accident sample set

Step one: unsupervised cluster analysis of sample sets:

1) Principal component analysis was performed on the sample set of table one by the matlab platform, and the first 10 principal components (cumulative contribution 96.87%) were extracted as shown in the following table:

2) Carrying out unsupervised clustering on the sample set by adopting a K-means tool box in matlab, and obtaining the following clustering result under the assumption that the clustering class is 4:

step two: classifying and processing the known road surface collapse accidents by a classification method based on distance and density:

1) The K-means tool box in the matlab platform is adopted to obtain the cluster center coordinates of the four types:

center ₁ ＝(3.48 0.63 -0.25 -0.41 -0.72 0.17 -0.09 0.14 -0.26 -0.25)

center ₂ ＝(0.46 -1.24 -0.27 0.36 0.39 -0.49 -0.03 0.04 0.17 0.13)

center ₃ ＝(-0.85 0.10 1.62 -0.67 0.13 0.56 0.02 -0.65 0.19 0.05)

center ₄ ＝(-2.30 1.36 -0.78 0.29 -0.22 0.16 0.08 0.39 -0.23 -0.08)

2) The main component data obtained by dimension reduction of each pavement collapse accident sample point is as follows:

3) The distances between each pavement collapse accident sample point and the centers of the clusters are as follows:

	center 1	center 2	center 3	center 4
					B1	51.88	19.78	21.11	10.39
B2	18.78	22.64	11.44	20.18
					B3	19.00	5.31	26.04	23.56
B4	23.72	7.38	20.69	27.30
					B5	6.54	33.91	34.73	49.21
B6	37.94	19.30	8.02	23.90
					B7	43.64	27.25	21.39	4.77
B8	32.48	6.76	10.18	14.76
					B9	11.67	4.05	12.58	24.01

4) The distance matrix of each road surface subsidence accident sample point from each collected road section sample point is calculated as follows, wherein one column of the matrix B_Dis represents the linear distance of one subsidence accident sample point from 24 sampled road section sample points:

solving the maximum value and the minimum value of each column of the matrix, namely an effective value range of an initial radius r of the corresponding collapse accident sample point, namely when r is smaller than the minimum value of the column, no other sampling section sample points exist in the radius r, and when r is larger than the maximum value of the column, all sampling section sample points are contained in the radius r;

for the collapse accident sample point B1, as shown in the first column of the matrix B_Dis, 2.78< r is less than or equal to 16.26, assuming r=12, the number of samples in each category is shown in the following table III in a range with the dimension-reduced B1 coordinate as the center and 12 as the radius:

category(s)	1	2	3	4
					Number of	0	3	2	3

Then there is D ₁ ＝B_Dis(1，:)＝[51.88 19.78 21.11 10.39]，Num ₁ ＝[0 3 2 3]；

5) Will D ₁ ，N ₁ After normalization, the method comprises the following steps:

D_gui ₁ ＝[1 0.23 0.26 0]，Num_gui ₁ ＝[0 1 0.67 1]；

assuming α=0.6 and β=0.4, there are:

Attribution ₁ ＝α×Num_gui ₁ -β×D_gui ₁ ＝[-0.4 0.51 0.3 0.6]

obviously attributon ₁ The fourth element value of (1) is the largest, so the collapse accident sample point B1 is classified into the fourth category.

As shown in fig. 1, the present invention mainly comprises seven parts: constructing a pavement collapse index system, performing unsupervised cluster analysis on a sample set, classifying and processing known pavement collapse accidents by a classification method based on distance and density, calculating the pavement collapse probability of each category and the probability membership degree of each sample point, building a pavement collapse probability prediction model, predicting the pavement collapse probability of a target road section sample point, and regulating and controlling the road running condition of the target road section sample point by an intelligent optimization algorithm;

the first part is to construct a pavement index information set, wherein the pavement index set comprises road construction quality information, geological condition information, external environment information and running road condition information, and the pavement index information set specifically comprises the following steps as shown in a table one in an embodiment one:

1) The road construction quality information includes:

road repair rateS _r S is the road repair area _a Is the total area of the road;

road condition indexThe weight representing the ith road damage type is mainly divided into crack type, deformation type and other types; omega represents the severity of road damage, the value range is 0 to 1, and the larger the value is, the more serious the damage is; sigma is the damage density;

road surface strength indexa ₀ ,a ₁ Is a constant;

pavement structural strength indexl _d ,l _s Respectively representing pavement design deflection, actually measured deflection, and units: millimeter;

road service life ind_5;

2) The geological condition information comprises:

the value range of the geological stability ind_6 is 0 to 1;

foundation bearing capacity ind_7=f _k +ε _b γ(b-3)+ε _d γ ₀ (d-0.5)，f _k Standard value (kN/m) representing bearing capacity of weak soil layer at bottom surface of cushion layer ² )，ε _b ,ε _d Bearing capacity correction factors representing foundation width and burial depth, respectively, b represents foundation width (m), d foundation burial depth (m), gamma substrate bottom weight (kN/m ³ )，γ ₀ -average bottom-up-weight (kN/m) of the substrate ³ )；

3) The external environment information comprises:

daily average precipitation amount ind_8, daily precipitation amount variance ind_9, daily average air temperature ind_10, daily air Wen Fangcha ind_11, daily average humidity ind_12 and daily humidity variance ind_13;

4) The running road condition information comprises:

average traffic flow ind_14, average traffic flow ind_15, and daily average load ind_16;

in the second part, clustering is performed on the index data of the road index information set, and n=24 road segment samples are classified into q=4 categories, and as shown in step 1) in the embodiment one, the specific steps are as follows:

1) Data ori_data= [ ind_1, ind_2, & gt, ind_16] of the road surface index information set are subjected to standardized processing, and dimension reduction is performed by a principal component analysis method:

s1, calculating a correlation coefficient matrix corr _16×16 ；

S2, calculating the characteristic value lambda of the correlation coefficient matrix _i (i=1, 2,., 16) and feature vector η _i (i＝1,2,...,16)；

S3, arranging the characteristic values in order from large to small, and calculating the accumulated contribution degree of the main components corresponding to the previous t (t=1, 2,.,. 16) characteristic values/>

S4, when the cut _t When the value is just equal to or greater than 0.95, the characteristic vectors corresponding to the first t=10 principal components are taken to form a transformation matrix tran= [ eta ] ₁ ,η ₂ ,...,η _t ]，η _t ＝[e ₁ ,e ₂ ,...,e ₁₆ ] ^T data=ori_data×tran=main_Agred is the dimension reductionThe reconstructed data; wherein main_agred= [ ind_1, ind_2, ind_t]Then the t-th principal component data is ind_t= [ a _1t ,a _2t ,...,a _nt ] ^T ，a _nt Nth sample data representing a nth principal component;

2) As shown in step 2) in the first embodiment, the K-means algorithm is adopted to perform unsupervised clustering analysis on the reconstructed data, and n=24 road segment samples are classified into q=4 categories:

s1: setting the number Q and the minimum threshold value of clustering target categories, and randomly selecting Q sample points from the data set data to serve as the center of the category clusters;

s2: each sample point is distributed to the center of the class cluster closest to the sample point;

s3: calculating the average value of all samples of each cluster, taking the average value as a new cluster center, and calculating the distances between the centers of mass centers of the clusters before and after updating;

s4: if the center distance between the front cluster and the rear cluster is greater than or equal to a set threshold value, repeating the steps S2-S3; if the center distance between the front cluster and the rear cluster is smaller than a set threshold, stopping calculation to obtain a clustering result;

the third part, as shown in fig. 2, collects the road surface index information data set in the historical road surface subsidence accident, constructs the classification sample space, and adopts the distance and density-based method to classify the road surface index information set in the historical road surface subsidence accident in the classification sample space, and the specific steps are as follows:

1) Collecting road surface index information data event_data in historical road surface subsidence accidents _s×16 (see table two in embodiment one), all road surface collapse accident index information is reconstructed by the transformation matrix tran to obtain e_data=event_data×tran=main_bgred= [ y ₁ ,y ₂ ,...,y _i ,...,y _t ]Wherein y is _i ＝[y _1i ,y _2i ,...,y _si ] ^T S=9 represents the total number of pavement collapse accident samples;

2) Constructing a t-dimensional coordinate system by using t=10 indexes, and constructing a classified sample point space;

3) Calculating cluster center coordinates of various clusters of the sample point space, there is a k (k=1, 2.,. Q.) class cluster center coordinate:

wherein ind_i _kj Data representing an ith index of a jth sample in a kth class cluster, m representing the total number of sample points of the kth class cluster;

then there are four cluster center coordinates in embodiment one as shown in step 1) in embodiment one;

4) Calculating the distance from each pavement collapse sample point to the center of each cluster in the historical pavement collapse accident, wherein the h (h=1, 2,.. S) th pavement collapse sample point is (y) _h1 ,y _h2 ,...,y _hi ,...,y _ht ) Then the sample point is to the Center of the cluster _k The distance of (2) can be noted as:

then the h (h=1, 2,., s) pavement collapse sample points are spaced D from the center of all clusters of the class _h ＝[d _h1 ,d _h2 ,...,d _hk ,...,d _hQ ]As shown in step 2) in example one;

5) As shown in step 4) in step two of the first embodiment, taking a radius r, counting how many sample points exist in each kind of clusters in the radius r by taking h (h=1, 2,., s) th pavement collapse sample point as a circle center, and counting the number of k kinds of sample points as num _k If the number of the sample points adjacent to the h pavement collapse sample point within the radius r is Num _h ＝[num _h1 ,num _h2 ,...num _hk ,...,num _hQ ]；

The effective value of the radius r should be not greater than the maximum value of the linear distance between the h (h=1, 2..s) th pavement collapse sample point and the sample point of each sampling road section, and not less than the minimum value thereof;

6) As shown in step 5) of the first embodiment, to the nearest cluster-like center and at a set radiusThe number of samples in a certain category is the most principle, num _h And D _h Normalized to num_ gui _h And D_ gui _h The degree of attribution is calculated according to the following formula:

Attribution _h ＝α×Num_gui _h -β×D_gui _h ＝[at _h1 ,at _h2 ,...,at _hi ,...,at _hQ ]

selection of Attribute _h At of the medium maximum value _hx The subscript x is the class serial number of the h pavement collapse sample point;

wherein, alpha, beta epsilon [0,1] is a weight coefficient, and satisfies alpha+beta=1, and the specific weight value needs to be determined according to the actual data distribution; the more concentrated the data class cluster distribution and the larger the class spacing, the larger the beta value; if the data clusters are distributed more dispersedly and the class spacing is smaller, the alpha value is larger;

the fourth part, calculating the pavement collapse probability of each category and the probability membership degree of each sample point, wherein the specific steps are as follows:

1) Note that P is present in category k (k=1, 2.,. Q) _k Accident of road collapse is counted as N _k ，ε _k For a proportionality constant less than 1, there is a general probability that the road segment of that class collapses:

if P _k Taking P if =0 _k =0.01, i.e. there is a probability of collapse for any road segment, guarantee prob _k Is not 0;

2) In the kth category, calculating the similarity (the value range is 0 to 1) between each non-collapse accident sample point and each collapse accident sample point, and obtaining a similarity matrix as follows:

b in matrix _ij Representing the ith collapse accident sample point and the jth non-collapse accident sample pointThe similarity of collapse accident sample points, m represents the total number of non-collapse accident sample points in the kth category;

3) For sim matrix _k Maximum value is found for each column of (2) to obtain mu_k _max ＝[b _max1 ,b _max2 ,...,b _maxj ,b _maxm ]，b _maxj The occurrence probability of the sample point representing the j non-collapse accident in the k category is prob _k Membership degree of (3);

the fifth part, for each category, respectively constructing a machine learning prediction model for predicting the pavement collapse probability, and specifically comprises the following steps:

1) Calculating the similarity of all sample points according to geological conditions and road construction quality, namely index ind_1-ind_7 information, classifying sample points with similarity higher than 0.9 into one category, and recording to obtain A categories, wherein the total number of samples in the i (i=1, 2.,) th category is count _i ；

2) If count _i (i=1, 2.,. A) is not less than set_N1, directly taking index information ind_8-ind_16 of the sample points and probability membership degree as inputs, taking corresponding road surface collapse probability as output, and constructing a machine learning model LM for predicting the road surface collapse probability of the ith road _i ；

3) If count _i (i＝1,2,...,A)<Set_n1 and count _i (i=1, 2.,. The A) is more than or equal to set_N2, the data Set size can be enlarged through a data enhancement method, then index information ind_8-ind_16 of the expanded sample points and probability membership are used as inputs, corresponding road surface collapse probability is used as output, and a machine learning model LM for predicting the road surface collapse probability of the ith class of road is constructed _i ；

4) If count _i (i＝1,2,...,A)<Set_n2, the sample data is too little to perform modeling analysis, and can be regarded as noise data;

a sixth step of collecting road index information sets of road segments to be evaluated, determining road category j (j=1, 2, a.) of the target road segment by the distance and density-based method in step 3) according to geological conditions and road construction quality, i.e., index ind_1 to ind_7 information, setting probability membership, and obtaining in step 6)Machine learning prediction model LM _j Predicting the road surface collapse probability of the road section;

and a seventh step of setting a safety probability threshold value of road surface subsidence, taking road running road conditions as regulation objects, and obtaining a running road condition range lower than the set safety probability threshold value through an intelligent optimization algorithm.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

Finally, it should be noted that: the above embodiments are only for illustrating the technical aspects of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those of ordinary skill in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.

Claims (9)

1. A urban pavement collapse prevention method based on probability prediction is characterized by comprising the following specific steps:

1) Constructing a pavement index information set, wherein the pavement index set comprises road construction quality information, geological condition information, external environment information and running road condition information;

2) Processing and clustering index data of the pavement index information set, and dividing n road section samples into Q categories;

3) Collecting a road surface index information data set in the historical road surface subsidence accident, constructing a classification sample space, adopting a distance and density-based method to collect the road surface index information set in the historical road surface subsidence accident in the classification sample space, and classifying according to the Q categories;

4) Calculating the pavement collapse probability of each category and the probability membership degree of each sample point;

5) Aiming at each category, respectively constructing a machine learning prediction model for predicting the pavement collapse probability;

6) Collecting road surface index information sets of road sections to be evaluated, and determining road surface types of the road sections to be evaluated by adopting the distance and density-based method in the step 3) according to geological conditions and road construction qualityOther j (j=1, 2,.,. A), setting probability membership, and using the machine learning prediction model LM obtained in step 6) _j Predicting the road surface collapse probability of a road section to be evaluated;

7) Setting a safety probability threshold value of road surface subsidence, taking road running road conditions as regulation objects, and obtaining a running road condition range lower than the set safety probability threshold value through an intelligent optimization algorithm;

the specific contents of the road construction quality information, the geological condition information, the external environment information and the running road condition information in the step 1) are as follows:

1-1) the road construction quality information includes:

road repair rateS _r S is the road repair area _a Is the total area of the road;

road condition indexp _i (i=1, 2, 3) represents the weight of the i-th road damage type, which is mainly classified into a crack type, a deformation type, and other types; omega represents the severity of road damage, the value range is 0 to 1, and the larger the value is, the more serious the damage is; sigma is the damage density;

road surface strength indexa ₀ ,a ₁ Is a constant;

pavement structural strength indexl _d ,l _s Respectively representing pavement design deflection, actually measured deflection, and units: millimeter;

road service life ind_5;

1-2) the geological condition information comprises:

the geological stability ind_6 is in a value range of 0 to 1, and the larger the value is, the higher the stability is;

foundation bearing capacity ind_7=f _k +ε _b γ(b-3)+ε _d γ ₀ (d-0.5)，f _k Standard value kN/m representing bearing capacity of weak soil layer at bottom surface of cushion layer ² ，ε _b ,ε _d The bearing capacity correction coefficients respectively representing the foundation width and the embedded depth, b represents the foundation width m, d the foundation embedded depth m and the bottom weight kN/m of the gamma substrate ³ ，γ ₀ -average bottom-up weight kN/m of the substrate ³ ；

1-3) the external environment information comprises:

month average precipitation amount ind_8, month precipitation amount maximum variance ind_9, month average air temperature ind_10, month air temperature maximum variance ind_11, month average humidity ind_12 and month humidity maximum variance ind_13;

1-4) the running road condition information comprises:

average traffic flow ind_14, average traffic flow ind_15, and daily all bear load ind_16.

2. The urban road surface subsidence prevention method based on probability prediction as set forth in claim 1, wherein the specific step of processing and clustering the index data of the road surface index information set in step 2) is as follows:

2-1) performing standardized processing on data ori_data= [ ind_1, ind_2, & gt, ind_16] of the road index information set, and performing dimension reduction by a principal component analysis method:

2-2) performing unsupervised cluster analysis on the reconstruction data by adopting a K-means algorithm, and classifying n road section samples into Q categories.

3. The urban road surface collapse prevention method based on probability prediction according to claim 2, wherein the specific steps of dimension reduction by the principal component analysis method in step 2-1) are as follows:

2-1-1) calculating a correlation coefficient matrix corr _16×16 ；

2-1-2) calculating the eigenvalue lambda of the correlation coefficient matrix _i (i=1, 2,., 16) and feature vector η _i (i＝1,2,...,16)；

2-1-3) arranging the characteristic values in order from large to small, and calculating the cumulative contribution degree of the main component corresponding to the t (t=1, 2,., 16) characteristic values before calculation

2-1-4) when the cut _t When the value is just equal to or greater than 0.95, the characteristic vectors corresponding to the first t principal components are taken to form a transformation matrix tran= [ eta ] ₁ ,η ₂ ,...,η _t ]Wherein eta _t ＝[e ₁ ,e ₂ ,...,e ₁₆ ] ^T ，data＝ori_data×tran＝[ind_1,ind_2,...,ind_t]The reconstructed data after dimension reduction; then the t-th principal component data is ind_t= [ a ] _1t ,a _2t ,...,a _nt ] ^T ，a _nt Nth sample data representing the nth principal component.

4. The urban road surface collapse prevention method based on probability prediction as set forth in claim 3, wherein the specific steps of performing unsupervised cluster analysis on the reconstructed data using the K-means algorithm in step 2-2) are as follows:

2-2-1) setting the number Q and the minimum threshold value of clustering target categories, and randomly selecting Q sample points from the data set data to serve as the center of the category clusters;

2-2-2) assigning each sample point to the cluster center closest to the cluster;

2-2-3) calculating the average value of all samples of each cluster, taking the average value as a new cluster center, and calculating the distances between the centers of mass of the clusters before and after updating;

2-2-4) if the center distance between the front cluster and the rear cluster is greater than or equal to a set threshold value, returning to the step 2-2-2); and if the center distance between the front cluster and the rear cluster is smaller than the set threshold, stopping calculation to obtain a clustering result.

5. The urban road surface subsidence prevention method based on probability prediction as set forth in claim 4, wherein the specific steps of clustering the road surface index information set in the historical road surface subsidence accident by Q categories in the classification sample space using the distance and density based method in step 3) are as follows:

3-1) collecting road surface index information data event_data in historical road surface subsidence accidents _s×16 Reconstructing all road surface collapse accident index information through a transformation matrix tran to obtain e_data=event_data×tran= [ y ] ₁ ,y ₂ ,...,y _i ,...,y _t ]Wherein y is _i ＝[y _1i ,y _2i ,...,y _si ] ^T S represents the total number of pavement collapse accident samples;

3-2) constructing a t-dimensional coordinate system by using t indexes, and constructing a classified sample point space;

3-3) calculating cluster center coordinates of each cluster of the sample point space, and then the k (k=1, 2,.,. Q) th cluster center coordinates are:

wherein ind_i _kj Data representing an ith index of a jth sample in a kth class cluster, m representing the total number of sample points of the kth class cluster;

3-4) calculating the distances from each pavement collapse sample point to the centers of the clusters in the historical pavement collapse accident, wherein the h (h=1, 2, the..s) pavement collapse sample points are (y) _h1 ,y _h2 ,...,y _hi ,...,y _ht ) Then the sample point is to the Center of the cluster _k The distance of (2) can be noted as:

then the h (h=1, 2,., s) pavement collapse sample points are spaced D from the center of all clusters of the class _h ＝[d _h1 ,d _h2 ,...,d _hk ,...,d _hQ ]；

The method comprises the steps of (1) counting an initial radius r, counting how many sample points exist in each type of clusters in the radius r by taking an h (h=1, 2,.. S) pavement collapse sample point as a circle center, and counting the number of the k type of sample pointsnum _k If the number of the sample points adjacent to the h pavement collapse sample point within the radius r is Num _h ＝[num _h1 ,num _h2 ,...num _hk ,...,num _hQ ]；

The effective value of the radius r is not greater than the maximum value of the linear distance between the h (h=1, 2, s) road surface collapse sample point and each sampling road section sample point, and is not less than the minimum value thereof;

in order to make the number of samples of a certain class nearest to the center of the class cluster and within a set radius be the most, num is calculated _h And D _h Normalized to num_ gui _h And D_ gui _h The degree of attribution is calculated according to the following formula:

Attribution _h ＝α×Num_gui _h -β×D_gui _h ＝[at _h1 ,at _h2 ,...,at _hi ,...,at _hQ ]

wherein, alpha, beta epsilon [0,1] is a weight coefficient, and satisfies alpha+beta=1, and the specific weight value needs to be determined according to the actual data distribution; the more concentrated the data class cluster distribution and the larger the class spacing, the larger the beta value; if the data clusters are distributed more dispersedly and the class spacing is smaller, the alpha value is larger;

selection of Attribute _h At of the medium maximum value _hx The subscript x is the class number to which the h pavement collapse sample point belongs.

6. The urban road surface subsidence prevention method based on probability prediction according to claim 5, wherein the specific steps of calculating the road surface subsidence probability of each class and the probability membership degree of each sample point in step 4) are as follows:

4-1) record k (k=1, 2., Q) presence of P in categories _k Accident of road collapse is counted as N _k ，ε _k For a proportionality constant less than 1, there is a general probability that the road segment of that class collapses:

if P _k Taking P if =0 _k =0.01, i.e. there is a probability of collapse for any road segment, guarantee prob _k Is not 0;

4-2) calculating the similarity between each non-collapse accident sample point and each collapse accident sample point in the kth category, and obtaining a similarity matrix as follows:

b in matrix _ij Representing the similarity of the ith collapse accident sample point and the jth non-collapse accident sample point, wherein m represents the total number of non-collapse accident sample points in the kth category;

4-3) pair sim matrix _k Maximum value is found for each column of (2) to obtain mu_k _max ＝[b _max1 ,b _max2 ,...,b _maxj ,b _maxm ]，b _maxj The occurrence probability of the sample point representing the j non-collapse accident in the k category is prob _k Is a membership of (1).

7. The urban road surface collapse prevention method based on probability prediction according to claim 6, wherein the specific steps of constructing the machine learning prediction model for predicting the road surface collapse probability in step 5) are as follows:

5-1) calculating the similarity of all sample points according to geological conditions and road construction quality, namely index ind_1-ind_7 information, classifying sample points with similarity higher than 0.9 into one category, and recording to obtain A categories in total, wherein the total number of samples in the i (i=1, 2., A) category is count _i ；

5-2) if count _i (i=1, 2.,. A) is not less than set_N1, directly taking index information ind_8-ind_16 of the sample points and probability membership degree as inputs, taking corresponding road surface collapse probability as output, and constructing a machine learning model LM for predicting the road surface collapse probability of the ith road _i ；

5-3) if count _i (i＝1,2,...,A)<Set_n1 and count _i (i=1, 2,.,. A) is equal to or greater than set_n2, then can be passedExpanding the data set scale by a data enhancement method, taking index information ind_8-ind_16 of the expanded sample points and probability membership as inputs, taking corresponding road surface collapse probability as output, and constructing a machine learning model LM for predicting the i-th road surface collapse probability _i ；

5-4) if count _i (i＝1,2,...,A)<Set_n2, the sample data is too small to perform modeling analysis, and is regarded as noise data.

8. A urban road surface collapse prevention system based on probabilistic prediction, the system comprising:

the road surface index information set construction module is used for constructing a road surface index information set;

the clustering module is used for processing and clustering the index data of the pavement index information set and classifying n road section samples into Q categories;

the historical data processing module is used for collecting road surface index information data sets in historical road surface subsidence accidents and classifying the road surface index information data sets according to the Q categories;

the class collapse probability calculation module is used for calculating the pavement collapse probability of each class and the probability membership degree of each sample point;

the prediction model learning module is used for calculating the pavement collapse probability of each category and the probability membership degree of each sample point;

the target prediction module is used for collecting road surface index information sets of road sections to be evaluated and predicting the road surface collapse probability of the road sections;

the collapse prevention module is used for setting the running road condition range of the safety probability threshold value;

the specific contents of the road construction quality information, the geological condition information, the external environment information and the running road condition information are as follows:

the road construction quality information includes:

road repair rateS _r S is the road repair area _a Is the total area of the road;

road condition indexp _i (i=1, 2, 3) represents the weight of the i-th road damage type, which is mainly classified into a crack type, a deformation type, and other types; omega represents the severity of road damage, the value range is 0 to 1, and the larger the value is, the more serious the damage is; sigma is the damage density;

road surface strength indexa ₀ ,a ₁ Is a constant;

pavement structural strength indexl _d ,l _s Respectively representing pavement design deflection, actually measured deflection, and units: millimeter;

road service life ind_5;

the geological condition information comprises:

the geological stability ind_6 is in a value range of 0 to 1, and the larger the value is, the higher the stability is;

foundation bearing capacity ind_7=f _k +ε _b γ(b-3)+ε _d γ ₀ (d-0.5)，f _k Standard value kN/m representing bearing capacity of weak soil layer at bottom surface of cushion layer ² ，ε _b ,ε _d The bearing capacity correction coefficients respectively representing the foundation width and the embedded depth, b represents the foundation width m, d the foundation embedded depth m and the bottom weight kN/m of the gamma substrate ³ ，γ ₀ -average bottom-up weight kN/m of the substrate ³ ；

The external environment information comprises:

month average precipitation amount ind_8, month precipitation amount maximum variance ind_9, month average air temperature ind_10, month air temperature maximum variance ind_11, month average humidity ind_12 and month humidity maximum variance ind_13;

the running road condition information comprises:

average traffic flow ind_14, average traffic flow ind_15, and daily all bear load ind_16.

9. A storage medium storing instructions adapted to be loaded by a processor to perform the urban road surface collapse prevention method according to any one of claims 1 to 7.