CN114580176A

CN114580176A - Automobile load intelligent reconstruction method based on multi-dimensional feature clustering

Info

Publication number: CN114580176A
Application number: CN202210219409.7A
Authority: CN
Inventors: 王川; 高庆飞; 王统; 李其远; 刘洋
Original assignee: Harbin Institute of Technology
Current assignee: Harbin Institute of Technology
Priority date: 2022-03-08
Filing date: 2022-03-08
Publication date: 2022-06-03
Anticipated expiration: 2042-03-08
Also published as: CN114580176B

Abstract

An automobile load intelligent reconstruction method based on multi-dimensional feature clustering relates to a load reconstruction method. Collecting vehicle data and dividing vehicles with the same number of axles into sets M_n(ii) a The transport property, the axle weight coefficient, the axle distance and the maximum axle weight coefficient are used as characteristic values to form a characteristic vector m ═ omega, alpha_i,d_j,α_max](ii) a Determining the number of vehicle type division, and collecting M_nIs divided into k₀Cluster and randomly select k₀Individual vehicles as cluster centers; clustering according to the structural similarity between the feature vectors; re-determining the cluster center of each cluster after the vehicles are completely divided, and repeating the step four until the final result is not changed any more so as to obtain the final resultThe cluster center is a representative vehicle type; calculating the probability distribution type and parameters of the distance between the vehicle heads and the vehicle weight; and (3) regarding the automobile load of the traffic flow on the bridge as a random process, and adopting a Monte Carlo method to reconstruct the automobile load. The method can reflect the actual condition more fully and improve the simulation precision of the automobile load.

Description

Automobile load intelligent reconstruction method based on multi-dimensional feature clustering

Technical Field

The invention relates to a load reconstruction method, in particular to an automobile load intelligent reconstruction method based on multi-dimensional feature clustering, and belongs to the technical field of bridge structure health monitoring and state evaluation.

Background

Roads and railways are important components in the transportation network, and bridges are scattered on key nodes of the transportation network, and the healthy operation of the bridges is the key of safe and smooth transportation. With the continuous promotion of urbanization construction and the closer and closer connection between areas, the increasing traffic demand promotes the rapid development of bridge construction. A large number of in-service bridges provide convenience for people to live and travel, and meanwhile, great challenges are brought to health monitoring and state evaluation of bridge structures.

Automobile load is an important basis for bridge health monitoring and state evaluation, and plays an important role in each link of bridge research, so that how to establish a reasonable automobile load model and correctly reflect the service condition of a bridge has important engineering value. At present, when an automobile load model is established, vehicle types are generally divided based on-site actual measurement traffic flow data, random traffic flows are generated according to the dividing results, and the purpose of simulating automobile loads is achieved. When the vehicle type is divided, the vehicle is divided into representative vehicle types according to the principle that the total weight of the vehicle, the axle weight of the vehicle, the number of axles of the vehicle, the axle distance or the transportation property of the vehicle are similar, or a new representative vehicle type is constructed by fitting the axle weight and the axle distance of the vehicle. The former determines similar standards from person to person, and the finally divided vehicle type results are different, while the latter is not representative of a vehicle type constructed when the vehicle composition is complicated, and cannot sufficiently reflect the actual vehicle load condition.

In addition, the automobile load has multidimensional characteristic attributes, namely continuous numerical characteristics such as axle weight and maximum axle weight coefficient, and discrete attribute characteristics such as transportation property and wheel base, so that the problem of vehicle type division is not fully considered at present, and the simulation precision of the automobile load is not facilitated.

Disclosure of Invention

In order to overcome the defects in the background art, the invention provides the automobile load intelligent reconstruction method based on the multi-dimensional feature clustering, and an automobile load model established by using the method can more fully reflect the actual automobile load condition and improve the simulation precision of the automobile load.

In order to achieve the purpose, the invention adopts the following technical scheme: an automobile load intelligent reconstruction method based on multi-dimensional feature clustering comprises the following steps:

step one, collecting vehicle data

Collecting vehicle data and dividing vehicles with the same number of axles into sets M_nN is 2,3,4,5,6, indicating the number of axes;

step two, extracting the characteristic vector

For set M_nBy transport properties omega, coefficient of axle weight alpha_iWheelbase d_jAnd the maximum axial weight coefficient alpha_maxFor the characteristic values, the characteristic vector m ═ ω, α, which constitutes the vehicle load_i,d_j,α_max]Wherein: the transport property ω is 0 when carrying passengers and 1 when carrying cargo;

step three, determining the vehicle type division number

Set M_nIs divided into k₀An individual cluster C_n1,C_n2,...,C_nk0And is in M_nIn selecting k at random₀Individual vehicle x_n1,x_n2,...,x_nk0As cluster C_n1,C_n2,...,C_nk0Cluster center of (a);

step four, clustering according to the structural similarity between the characteristic vectors

For set M_nRespectively comparing the feature vector of x with each cluster center x_n1,x_n2,...,x_nk0The structural similarity between the characteristic vectors is calculated, and the vehicle x is divided into clusters with the highest structural similarity of the characteristic vectors;

step five, iterative optimization

And re-determining the cluster center of each cluster after the vehicles are completely divided, wherein the characteristic value of the cluster center is as follows: for the transport properties omega and wheelbase d_jTaking the mode of the features, for the axial weight coefficient alpha_iAnd the maximum axial weight coefficient alpha_maxTaking the average value of the features, then repeating the fourth step based on the re-determined cluster center until the final result is not changed any more, and taking the final cluster center x_n1',x_n2',...,x_nk0' is a representative vehicle type for the cluster;

sixthly, calculating the probability distribution of the distance between the vehicle heads and the vehicle weight

Calculating the probability distribution types and parameters of all vehicle headway obeys to form a cluster C_n1,C_n2,...,C_nk0Taking the vehicle weights of all the vehicles as samples, and fitting the cluster of representative vehicle models x respectively_n1',x_n2',...,x_nk0' and calculating the probability distribution type and parameters of vehicle weight obedience;

seventhly, reconstructing the load of the automobile

And (3) regarding the automobile load of the traffic flow on the bridge as a random process, and adopting a Monte Carlo method to reconstruct the automobile load.

Compared with the prior art, the invention has the beneficial effects that: the invention fully considers the continuous numerical characteristics of the axle weight and the maximum axle weight coefficient and the discrete attribute characteristics of the transportation property and the axle distance, so that the established automobile load model can more fully reflect the actual automobile load condition, the simulation precision of the automobile load is improved, and the invention is beneficial to solving the problems of large error caused by the fact that the typical automobile type division is not objective and the actual automobile load characteristic is difficult to reflect when the existing automobile load model is established.

Drawings

FIG. 1 shows number C in the example of the present invention₂₁The vehicle weight distribution map of (a);

FIG. 2 shows number C in the embodiment of the present invention₃₁The vehicle weight distribution map;

FIG. 3 shows the number C in the embodiment of the present invention₄₁The vehicle weight distribution map of (a);

FIG. 4 is the present inventionNumber C in the examples of the invention₅₁The vehicle weight distribution map;

FIG. 5 shows the number C in the embodiment of the present invention₆₁The vehicle weight distribution map;

FIG. 6 is a plot of headway distance fitted to all vehicles in an embodiment of the present invention;

FIG. 7 is a diagram of a model of a simply supported beam in an embodiment of the invention;

FIG. 8 is an influence line graph of a simple girder model extracted in an embodiment of the present invention;

FIG. 9 is a graph of the loading effect of traffic shifting loading in an embodiment of the present invention;

FIG. 10 is a graph of the maximum loading effect of the vehicle cyclically loaded during the period in the embodiment of the invention.

Detailed Description

The technical solutions in the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the invention, rather than all embodiments, and all other embodiments obtained by those skilled in the art without any creative work based on the embodiments of the present invention belong to the protection scope of the present invention.

An automobile load intelligent reconstruction method based on multi-dimensional feature clustering comprises the following steps:

step one, collecting vehicle data

Collecting vehicle data, dividing vehicles according to the number of axles, and dividing vehicles with the same number of axles into a set M_n(n is 2,3,4,5,6, indicating the number of axes);

step two, extracting the characteristic vector

For set M_nBy transport properties omega, coefficient of axle weight alpha_i(i ═ 1., n), wheelbase d_j(j ═ 1.., n-1) and the maximum axle weight coefficient α_maxFor the characteristic values, the characteristic vector m ═ ω, α, which constitutes the vehicle load_i,d_j,α_max]Wherein: the value of the transport property omega corresponds to two situations of carrying passengers or goods, the value of the transport property omega is 0 when carrying passengers, the value of the transport property omega is 1 when carrying goods, and the axle weight coefficient alpha_iComputingIs given by the formula

f_iRepresenting the axle weight, the maximum axle weight coefficient alpha_max＝max(α₁,...,α_n)；

Step three, determining the number of vehicle type division

Set M_nIs divided into k₀An individual cluster C_n1,C_n2,...,C_nk0And is in M_nIn (1) random selection of k₀Individual vehicle x_n1,x_n2,...,x_nk0As cluster C_n1,C_n2,...,C_nk0Cluster center of (a);

step four, clustering according to structural similarity between the characteristic vectors

For set M_nRespectively comparing the eigenvectors of x with each cluster center x_n1,x_n2,...,x_nk0And dividing the vehicle x into clusters with the highest feature vector structure similarity, specifically, when calculating the structure similarity between the feature vectors of two vehicles:

for the transport properties omega and wheelbase d_jFor such discrete attribute features, if the features are the same, the structural similarity of the feature is marked as 1, otherwise, the structural similarity is marked as 0, such as

Then vehicle x and vehicle x_nk0The structural similarity between the transport properties ω is noted as 1, e.g.

Then vehicle x and vehicle x_nk0The structural similarity between transport properties omega is noted as 0,

for the axial weight coefficient alpha_iAnd maximum axial weight coefficient alpha_maxThe threshold value beta is set for the continuous numerical characteristics, and if the characteristic ratio is between [ 1-beta, 1+ beta ]]If so, the structural similarity of the feature is recorded as 1, otherwise, the structural similarity is recorded as 0, e.g., in the case of

Then vehicle x and vehicle x_nk0First axial weight coefficient alpha₁Structural similarity between them is noted as 1, e.g.

Or

Then vehicle x and vehicle x_nk0First axial weight coefficient alpha₁The structural similarity between them is noted as 0,

the structural similarity between the characteristic vectors of the two final vehicles is the transport property omega and the wheelbase d_jAxial weight coefficient alpha_iAnd maximum axial weight coefficient alpha_maxThe sum of structural similarities of (a);

step five, iterative optimization

And re-determining the cluster center of each cluster after the vehicles are completely divided, wherein the characteristic value of the cluster center is as follows:

for the transport properties omega and wheelbase d_jSuch discrete attribute features, taking the mode of the feature,

for the axial weight coefficient alpha_iAnd maximum axial weight coefficient alpha_maxSuch continuous numerical features, taking the average of the features,

and then repeating the step four based on the redetermined cluster center until the final result is not changed any more, and calculating the final cluster center x_n1',x_n2',...,x_nk0' is a representative vehicle type for the cluster;

step seven, reconstructing the automobile load

For moving traffic over bridgesThe automobile load is regarded as a random process, and the representative automobile type x_n1',x_n2',...,x_nk0' the vehicle weight and the distance between the heads are changed along with time but obey a certain probability distribution, and after the obeyed probability distribution is obtained, the Monte Carlo method can be adopted to reconstruct the vehicle load, and the concrete steps are as follows:

1) setting each representative vehicle type x in the traffic flow_n1',x_n2',...,x_nk0' quantity Q_n1,Q_n2,...,Q_nk0；

2) Generating a vehicle weight random number meeting the vehicle weight probability distribution based on the sixth step;

3) generating a random number of the distance between the vehicle heads, which meets the probability distribution of the distance between the vehicle heads, based on the sixth step;

4) randomly arranging all the vehicle weights according to the distance between the vehicle heads to generate a random traffic flow;

5) carrying out moving loading on random traffic flow on a bridge structure influence line, wherein the moving step length is 1m each time, calculating an automobile load effect value F after the traffic flow moves each time, and calculating an automobile load effect maximum value F in the whole loading process_max＝max(F)；

6) Generating random traffic flow in the period T, circularly loading the bridge structure, and recording the maximum value F of the automobile load effect after each complete loading_maxStatistics of F_maxAnd (5) completing the reconstruction of the automobile load according to the probability distribution type and parameters.

Examples

In this embodiment, 50000 cars are collected, and the specific number is as follows after dividing according to the number of axles:

selecting the characteristic values of the more representative cars in the five-axle cars as follows (the axle weight coefficient retains 2 decimal places, and the axle distance retains 1 decimal place):

the method is adopted to perform multi-dimensional characteristic clustering on the automobiles with the same number of axes, wherein the parameters are set as follows:

number of axes	Categorizing categories	Numbering	Threshold value beta
				2	4	C₂₀/C₂₁/C₂₂/C₂₃	0.3
3	2	C₃₀/C₃₁	0.3
				4	2	C₄₀/C₄₁	0.3
5	3	C₅₀/C₅₁/C₅₂	0.3
				6	2	C₆₀/C₆₁	0.3

The final cluster partitioning results are as follows:

to number C_n1The vehicles (n is 2,3,4,5,6) are subjected to vehicle weight distribution fitting,

C₂₁the vehicle weight compliance parameter is [2.1501,0.4346 ]]The lognormal distribution of (a) is shown with reference to fig. 1;

C₃₁the vehicle weight compliance parameter is [21.7188, 5.8436 ]]Normal distribution of (2), as shown in FIG. 2;

C₄₁the vehicle weight compliance parameter is [29.0278, 7.0336 ]]The normal distribution of (2) is shown with reference to fig. 3;

C₅₁the vehicle weight compliance parameters are [17.2254, 2.1654, 38.0784, 15.1845, 0.4236 ]]The bimodal normal distribution of (a) is shown with reference to FIG. 4;

C₆₁the vehicle weight compliance parameter is [47.9739, 8.77646 ]]The normal distribution of (2) is shown in FIG. 5;

fitting the headway distribution of all vehicles, wherein the results show that the headway distribution obeys the lognormal distribution with the parameter of [2.58,0.33], and the method is shown in the figure 6;

randomly arranging the vehicle weights of all vehicle types according to the distance between the vehicle heads to generate traffic flow, loading a bridge, wherein the bridge type adopts a simply supported beam with the span of 50 meters, the cross section is a concrete box beam cross section, the material strength is C50, and a simply supported beam model is shown in a reference figure 7;

the influence line of the bridge is extracted and is shown in figure 8;

carrying out moving loading on the traffic flow on a bridge structure influence line, wherein the moving step length is 1m each time, and calculating the automobile load effect value after the traffic flow moves each time, which is shown in a figure 9;

generating random traffic flow in a period T of 1000, circularly loading the bridge structure, and recording the maximum value F of the automobile load effect after each complete loading_maxStatistics of F_maxThe method completes the reconstruction of the automobile load according to the obedient probability distribution type and parameters, and the result shows that the maximum obedient parameter of the automobile load effect is [7412.3900, 540.5784 ]]The normal distribution of (2) is shown in FIG. 10.

It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned.

Furthermore, it should be understood that although the present description refers to embodiments, not every embodiment may contain only a single embodiment, and such description is for clarity only, and those skilled in the art should integrate the description, and the embodiments may be combined as appropriate to form other embodiments understood by those skilled in the art.

Claims

1. An automobile load intelligent reconstruction method based on multi-dimensional feature clustering is characterized by comprising the following steps: the method comprises the following steps:

step one, collecting vehicle data

step two, extracting the characteristic vector

For set M_nBy transport properties omega, coefficient of axle weight alpha_iWheelbase d_jAnd the maximum axial weight coefficient alpha_maxFor the characteristic values, the characteristic vector m ═ ω, α, which constitutes the vehicle load_i,d_j,α_max]Wherein: the transport property omega is 0 when carrying passengers and 1 when carrying cargo;

step three, determining the vehicle type division number

Set M_nIs divided into k₀An individual cluster

And at M_nIn (1) random selection of k₀One vehicle

As clusters

Cluster center of (a);

For set M_nRespectively comparing the eigenvectors of x with each cluster center

The structural similarity between the characteristic vectors of (1) and dividing the vehicle x into the clusters with the highest structural similarity of the characteristic vectors thereof;

step five, iterative optimization

And re-determining the cluster center of each cluster after the vehicles are completely divided, wherein the characteristic value of the cluster center is as follows: for the transport properties omega and wheelbase d_jTaking the mode of the features, for the axial weight coefficient alpha_iAnd maxCoefficient of axial weight alpha_maxAveraging the features, and then repeating step four based on the re-determined cluster center until the final result is no longer changed, and taking the final cluster center

Is a representative vehicle type for the cluster;

Calculating the probability distribution types and parameters of all vehicle headway obeys to form clusters

The vehicle weights of all the vehicles are taken as samples, and the representative vehicle models of the cluster are respectively fitted

The vehicle weight distribution, and calculating the probability distribution type and parameters of the vehicle weight obedience;

seventhly, reconstructing the load of the automobile

2. The intelligent automobile load reconstruction method based on multi-dimensional feature clustering as claimed in claim 1, characterized in that: in the fourth step, during the calculation of the structural similarity: for the transport properties omega and wheelbase d_jIf the characteristics are the same, the structural similarity is recorded as 1, otherwise, the structural similarity is recorded as 0, and the axial weight coefficient alpha is obtained_iAnd maximum axial weight coefficient alpha_maxSetting a threshold value beta, if the characteristic ratio is between [ 1-beta, 1+ beta ]]If not, the structural similarity is marked as 1, otherwise, the structural similarity is marked as 0, and finally, the structural similarity between the characteristic vectors of the two vehicles is the transportation property omega and the wheelbase d_jAxial weight coefficient alpha_iAnd maximum axial weight coefficient alpha_maxSum of structural similarity of (a).

3. The intelligent automobile load reconstruction method based on the multidimensional feature clustering according to claim 1, characterized in that: the reconstruction by adopting the Monte Carlo method in the seventh step comprises the following steps:

1) setting each representative vehicle type in the traffic flow

Number of (2)

3) generating a random number of the locomotive interval meeting the probability distribution of the locomotive interval based on the sixth step;

5) carrying out moving loading on random traffic on a bridge structure influence line, wherein the step length of each moving is 1m, and calculating an automobile load effect value F after each time of traffic moving;

6) generating random traffic flow in the period T, circularly loading the bridge structure, and recording the maximum value F of the automobile load effect after each complete loading_maxStatistics of F_maxAnd (5) completing the reconstruction of the automobile load according to the obeyed probability distribution type and parameters.