CN108090635B - Road performance prediction method based on cluster classification - Google Patents

Road performance prediction method based on cluster classification Download PDF

Info

Publication number
CN108090635B
CN108090635B CN201810116585.1A CN201810116585A CN108090635B CN 108090635 B CN108090635 B CN 108090635B CN 201810116585 A CN201810116585 A CN 201810116585A CN 108090635 B CN108090635 B CN 108090635B
Authority
CN
China
Prior art keywords
time sequence
road
cluster
function
time series
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810116585.1A
Other languages
Chinese (zh)
Other versions
CN108090635A (en
Inventor
赵永利
陈安琪
李靖
刘国强
潘园园
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southeast University
Original Assignee
Southeast University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southeast University filed Critical Southeast University
Priority to CN201810116585.1A priority Critical patent/CN108090635B/en
Publication of CN108090635A publication Critical patent/CN108090635A/en
Application granted granted Critical
Publication of CN108090635B publication Critical patent/CN108090635B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/20—Administration of product repair or maintenance

Abstract

The invention discloses a road performance prediction method based on cluster classification, which comprises the following steps: (1) collecting road detection data, and recording the road detection data as a first historical time sequence; (2) cleaning the first historical time sequence, removing data of the maintained road section, and recording the data as a second historical time sequence; (3) according to the similarity measurement of the road indexes, carrying out cluster classification on the second historical time sequence according to a cluster model; (4) respectively calculating the weight of each category time sequence in the second historical time sequence and the time sequence function of each index in each category time sequence; (5) calculating the time sequence function of each index of all clustering categories according to the weight of each category time sequence and the time sequence function; (6) and determining a road performance comprehensive prediction function. The invention improves the accuracy of road performance prediction.

Description

Road performance prediction method based on cluster classification
Technical Field
The invention relates to a road performance prediction method, in particular to a road performance prediction method based on cluster classification.
Background
And the road maintenance scheme is made according to the maintenance decision result, and the final scheme of the maintenance decision is determined by the road performance prediction result. Therefore, the accurate road performance prediction model can provide an effective and scientific maintenance decision, so that the road maintenance is more targeted and scientific.
At the present stage, the prediction method of the road surface performance is to firstly investigate the road condition and then predict the road surface performance according to the investigation data by using a road surface performance prediction model. However, in a road, there are usually some sections that have not been maintained, some sections that have been maintained, and there is a possibility that the service performance of different sections may be greatly different in the same road. When road surface performance prediction is carried out, if data of a section which is maintained is not removed, noise of detected data is increased, and the error of a prediction result is large. If the road sections with different road performances are studied together and processed uniformly, the information is not concentrated enough, and a lot of key information is lost. The loss of the key information can reduce the accuracy of performance prediction, so that the decision of maintenance measures is influenced, and finally, the pavement diseases cannot be treated in a targeted manner.
Disclosure of Invention
The purpose of the invention is as follows: in order to overcome the defects of the prior art, the invention provides a road performance prediction method based on cluster classification, and the method solves the problem of low accuracy of road performance prediction.
The technical scheme is as follows: the invention relates to a road performance prediction method based on cluster classification, which comprises the following steps:
(1) collecting road detection data, and recording the road detection data as a first historical time sequence;
(2) cleaning the first historical time sequence, removing data of the maintained road section, and recording the data as a second historical time sequence;
(3) according to the similarity measurement of the road indexes, carrying out cluster classification on the second historical time sequence according to a cluster model;
(4) respectively calculating the weight of each category time sequence in the second historical time sequence and the time sequence function of each index in each category time sequence;
(5) calculating the time sequence function of each index of all clustering categories according to the weight of each category time sequence and the time sequence function;
(6) and determining a road performance comprehensive prediction function.
Preferably, in the step (3), the road surface indexes include a road surface running quality index RQI, a road surface rutting condition index RDI, and a road surface damage condition index PCI.
Preferably, in step (3), the clustering model is based on euler distance:
Figure GDA0003240800010000021
wherein x isikIs the k-th component, x, of the feature vector ijkIs the kth component of the feature vector j and m is the total number of components. Preferably, in step (5), the weight calculation formula is:
αi＝xi/X
wherein x isiAnd X is the number of all the unserviced sections in the second time sequence.
Preferably, in step (5), the RQI time series function of each type of time series is calculated as:
Figure GDA0003240800010000022
wherein i is the ith class for classifying the second historical time series, t is the number of years and t is more than or equal to 1, and QiAnd q isiAre respectively RQIi(t) coefficients and indices;
the RDI time series function calculation formula of each type of time series is as follows:
Figure GDA0003240800010000023
wherein i is the ith class for classifying the second historical time series, t is the number of years and t is more than or equal to 1, DiAnd diAre respectively RDIi(t) coefficients and indices;
the PCI time sequence function calculation formula of each type of time sequence is as follows:
Figure GDA0003240800010000024
wherein i is the ith class for classifying the second historical time series, and t isYears and t is more than or equal to 1, CiAnd ciAre respectively PCIiThe coefficient and the index of (t).
Preferably, in step (5), the RQI time-series functions of all cluster categories are:
Figure GDA0003240800010000025
wherein the content of the first and second substances,
Figure GDA0003240800010000026
n is the total number of cluster classifications;
the RDI time series function for all cluster categories is:
Figure GDA0003240800010000031
wherein the content of the first and second substances,
Figure GDA0003240800010000032
n is the total number of cluster classifications;
the PCI time series function for all cluster categories is:
Figure GDA0003240800010000033
wherein the content of the first and second substances,
Figure GDA0003240800010000034
n is the total number of cluster classifications.
Has the advantages that: the method eliminates the data of the maintained road sections, divides the road sections into different categories by using cluster analysis, captures the differences of samples of different categories, respectively studies the road performance of different road sections, and improves the accuracy of road performance prediction.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a graph of experimental data comparing the method of use of the present invention with conventional methods.
Detailed Description
Example 1
In order to accurately predict the road performance, the invention needs to eliminate the data of the maintained road sections and classify the non-maintained road sections, as shown in fig. 1.
Step 1, collecting road detection data, and recording the data as a first historical time sequence.
And 2, cleaning the first historical time sequence, checking data consistency, processing invalid values and missing values, eliminating data of the maintained road section, and recording the processed first historical time sequence as a second historical time sequence.
And 3, dividing the second historical time sequence into a first category time sequence, … … and an Nth category time sequence according to the road running quality index RQI, the road rut condition index RDI and the road damage condition index PCI similarity measurement and a clustering model, wherein N is greater than 1.
Different clustering methods have different classification results, and the Euclidean distance is the most common method and is suitable for the condition that the standards of all vectors are unified, so that the similarity is reflected by selecting the method. The clustering model is based on Euler distance:
Figure GDA0003240800010000035
wherein x isikIs the k-th component, x, of the feature vector ijkIs the kth component of the feature vector j and m is the total number of components.
Step 4, respectively calculating the weight alpha of each category time sequencei。
The weights are calculated as follows: alpha is alphai＝xiand/X. Wherein x isiAll the number of the segments in the ith category time sequence, and X is the number of all the unserved segments in the second category time sequence.
Step 5, respectively calculating RQI, RDI and PCI time sequence function RQI in each category time sequencei(t)、RDIi(t)、PCIi(t) of (d). According to the weight alpha of each classiAnd RQIiAnd (t) calculating the RQI time series functions RQI (t) of all cluster types through weighted average. Similarly, the RDI, the PCI time series function RDI (t), and the PCI (t) of all the cluster types are obtained.
The time series function of the RQI of the ith class and the t (t is more than or equal to 1) year is as follows:
Figure GDA0003240800010000041
the class i year-t (t ≧ 1) RDI time series function is:
Figure GDA0003240800010000042
the ith type and t (t is more than or equal to 1) year PCI time sequence function is as follows:
Figure GDA0003240800010000043
wherein Q isiAnd q isiAre respectively RQIi(t) coefficients and indices; diAnd diAre respectively RDIi(t) coefficients and indices; ciAnd ciAre respectively PCIi(t) coefficient and exponent.
The RQI time series function for all the uncured sections is:
Figure GDA0003240800010000044
the RDI time series function for all the sections not maintained is:
Figure GDA0003240800010000045
the PCI time series function for all the sections not maintained is:
Figure GDA0003240800010000046
wherein the content of the first and second substances,
Figure GDA0003240800010000047
Figure GDA0003240800010000048
and 6, determining a comprehensive road performance prediction function PQI (t) by using the RDI (t), the RDI (t) and the PCI (t).
The road performance prediction model is as follows:
PQI(t)＝[0.4×RQI(t)+0.15×RDI(t)+0.35×PCI(t)]/0.9。
where 0.4, 0.15 and 0.35 are the values given by the specification.
Example 2
The steps of the invention when used on a specific road include:
first, since a road is divided into a plurality of sections, detection data of the plurality of sections in the road is collected and recorded as a first historical time series,
and performing data cleaning on the first historical time sequence, checking data consistency, and processing invalid values and missing values. Then, the sections that have been maintained are removed according to the maintenance data of the past year, as shown in table 1.
TABLE 1 cured sections to be rejected
Figure GDA0003240800010000051
The remaining detection data is recorded as a second historical time series.
And classifying the second historical time sequence by utilizing a clustering model according to the second historical time sequence and according to the similarity measurement of the road surface damage condition index PCI, the road surface rutting condition index RDI and the road surface running quality index RQI. The three segments with similar indexes are classified into one category. All the sections of the uncured section were classified into 9 categories as shown in table 2.
TABLE 2 all classes of section not maintained
Figure GDA0003240800010000052
Figure GDA0003240800010000061
Calculating the weight alpha of each category according to a weight formulaiAs shown in table 3.
TABLE 3 weight of class αi
Cluster classification Weight αi
First kind 0.11
Second class 0.2
Class III 0.05
Class IV 0.15
Fifth class 0.01
Class six 0.25
Class seven 0.01
Class VIII 0.2
Ninth class 0.02
The time series function of RQI, RDI and PCI in each category is calculated as follows.
Figure GDA0003240800010000062
Figure GDA0003240800010000063
Figure GDA0003240800010000071
Wherein, RQIi(t)、RDIi(t) and PCIiAnd (t) is respectively RQI, RDI and PCI corresponding to all road sections of the ith class in the t (t is more than or equal to 1) th year. QiAnd q isiAre respectively RQIi(t) coefficients and indices; diAnd diAre respectively RDIi(t) coefficients and indices; ciAnd ciAre respectively PCIiThe coefficient and the index of (t). These parameters were determined by regression based on the road survey data over the years, as shown in table 4.
TABLE 4 parameters of Performance indicator time series function for each Path
Cluster classification Qi qi Di di Ci ci
First kind 95.331 0.001 91.7142 0.015 97.9669 0.00227
Second class 95.1956 0.000671 96.9709 0.0387 100.3 0.00726
Class III 95.3232 0.00139 96.5424 0.0333 100.3 0.00548
Class IV 95.4014 0.00187 99.3507 0.0564 99.5128 0.00495
Fifth class 96.4148 0.00738 104.1 0.0876 101.7 0.0116
Class six 95.227 0.0009 95.163 0.0307 99.0654 0.00343
Class seven 105 0.053 94.8778 0.0132 100.8 0.0038
Class VIII 95.5087 0.00165 96.922 0.065 100.2 0.00662
Ninth class 95.9858 0.00336 97.0159 0.0368 104.2 0.0327
According to the weight alpha of each cluster categoryiAnd QiWeighted average calculation
Figure GDA0003240800010000072
According to the weight alpha of each cluster categoryiAnd q isiWeighted average calculation
Figure GDA0003240800010000073
Determine RQI time series functions for all cluster classes:
Figure GDA0003240800010000074
similarly, determining the RDI time sequence functions and the PCI time sequence functions of all the sections which are not maintained: rdi (t) 96.32 × e-0.042t、PCI(t)＝99.7×e-0.0057t。
And substituting the time-series functions into the comprehensive road performance prediction model according to the calculated time-series functions of RQI, RDI and PCI to finally determine the comprehensive road performance prediction function of the damaged road: PQI (t) 42.42e-0.0018t+16.05e-0.042t+38.77e-0.0057t。
At present, according to the existing method, when a road performance comprehensive prediction function is calculated, a maintained road section is not removed, so that the PCI, RDI and RQI of an intact road section are all calculated, the PQI index is higher, and the actual situation cannot be truly reflected. In addition, in the existing method, the road sections are not classified, and the arithmetic mean value of PCI, RDI and RQI is directly adopted, so that a lot of effective information is hidden, and the difference between the prediction result and the actual situation is large.
Referring to fig. 2, pqi (t) calculated by the prior art method and the method of the present invention are compared. As a result, the PQI index predicted by the existing method is almost unchanged, and is still about 95 in the 25 th year, which is seriously inconsistent with the actual situation. Therefore, compared with the existing method, the method can accurately and effectively predict the road performance comprehensive function PQI (t) and improve the prediction accuracy.

Claims (3)

1. A road performance prediction method based on cluster classification is characterized by comprising the following steps:
(1) collecting road detection data, and recording the road detection data as a first historical time sequence;
(2) cleaning the first historical time sequence, removing data of the maintained road section, and recording the data as a second historical time sequence;
(3) according to the similarity measurement of the road surface indexes, carrying out cluster classification on the second historical time sequence according to a cluster model, wherein the road surface indexes comprise a road surface running quality index RQI, a road surface rutting condition index RDI and a road surface damage condition index PCI;
(4) respectively calculating the weight of each category time sequence in the second historical time sequence and the time sequence function of each index in each category time sequence;
(5) calculating the time sequence function of each index of all clustering categories according to the weight of each category time sequence and the time sequence function;
the calculation formula of the RQI time series function of each type of time series is as follows:
Figure FDA0003240795000000011
wherein i is the ith class for classifying the second historical time series, t is the number of years and t is more than or equal to 1, and QiAnd q isiAre respectively RQIi(t) coefficients and indices;
the RDI time series function calculation formula of each type of time series is as follows:
Figure FDA0003240795000000012
wherein i is the ith class for classifying the second historical time series, t is the number of years and t is more than or equal to 1, DiAnd diAre respectively RDIi(t) coefficients and indices;
the PCI time sequence function calculation formula of each type of time sequence is as follows:
Figure FDA0003240795000000013
wherein i is the ith class for classifying the second historical time series, t is the number of years and t is more than or equal to 1, CiAnd ciAre respectively PCIi(t) coefficients and indices;
the RQI time series function for all cluster classes is:
Figure FDA0003240795000000014
wherein the content of the first and second substances,
Figure FDA0003240795000000015
n is the total number of cluster classifications, αiA weight for the ith category;
the RDI time series function for all cluster categories is:
Figure FDA0003240795000000021
wherein the content of the first and second substances,
Figure FDA0003240795000000022
n being cluster-classifiedTotal number, αiA weight for the ith category;
the PCI time series function for all cluster categories is:
Figure FDA0003240795000000023
wherein the content of the first and second substances,
Figure FDA0003240795000000024
n is the total number of cluster classifications, αiA weight for the ith category;
(6) and determining a road performance comprehensive prediction function.
2. The road use performance prediction method based on cluster classification as claimed in claim 1, wherein in step (3), the cluster model is a euler distance-based cluster model:
Figure FDA0003240795000000025
wherein d isijEuler distance, x, for eigenvectors i and eigenvectors jikIs the k-th component, x, of the feature vector ijkIs the kth component of the feature vector j and m is the total number of components.
3. The cluster classification-based road use performance prediction method according to claim 1, wherein in the step (5), the weight calculation formula is as follows:
αi＝xi/X
wherein alpha isiIs the weight of the ith class, xiAnd X is the number of all the unserviced sections in the second historical time sequence.
CN201810116585.1A 2018-02-06 2018-02-06 Road performance prediction method based on cluster classification Active CN108090635B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810116585.1A CN108090635B (en) 2018-02-06 2018-02-06 Road performance prediction method based on cluster classification

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810116585.1A CN108090635B (en) 2018-02-06 2018-02-06 Road performance prediction method based on cluster classification

Publications (2)

Publication Number Publication Date
CN108090635A CN108090635A (en) 2018-05-29
CN108090635B true CN108090635B (en) 2021-10-29

Family

ID=62193855

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810116585.1A Active CN108090635B (en) 2018-02-06 2018-02-06 Road performance prediction method based on cluster classification

Country Status (1)

Country Link
CN (1) CN108090635B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109947755B (en) * 2019-03-05 2023-04-14 南京道润交通科技有限公司 Pavement usability detection data quality control method, storage medium and electronic equipment
CN111177895B (en) * 2019-12-13 2023-10-20 中公高科养护科技股份有限公司 Method and system for establishing prediction model of tri-fold line road surface technical condition
CN113822387B (en) * 2021-11-24 2022-04-01 佛山市交通科技有限公司 Road surface damage condition index prediction method, system, equipment and medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103091480A (en) * 2013-01-07 2013-05-08 河北工业大学 Entropy weight-based underground road bituminous pavement service performance evaluation method
CN104268213A (en) * 2014-09-24 2015-01-07 长安大学 Maintenance road segment dividing method based on multisource detection data
CN106249601A (en) * 2016-09-29 2016-12-21 广东华路交通科技有限公司 A kind of road section length division methods based on Ordered Clustering Analysis
CN106251625A (en) * 2016-08-18 2016-12-21 上海交通大学 Three-dimensional urban road network global state Forecasting Methodology under big data environment

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103091480A (en) * 2013-01-07 2013-05-08 河北工业大学 Entropy weight-based underground road bituminous pavement service performance evaluation method
CN104268213A (en) * 2014-09-24 2015-01-07 长安大学 Maintenance road segment dividing method based on multisource detection data
CN106251625A (en) * 2016-08-18 2016-12-21 上海交通大学 Three-dimensional urban road network global state Forecasting Methodology under big data environment
CN106249601A (en) * 2016-09-29 2016-12-21 广东华路交通科技有限公司 A kind of road section length division methods based on Ordered Clustering Analysis

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
应用聚类分析法确定沥青路面预防性养护方案;曾峰、张肖宁、李智;《华南理工大学学报( 自然科学版)》;20080615;第36卷(第6期);全文 *
灰色聚类法在沥青路面性能评价中的应用研究;周育名、李金明、马旺宇;《华东公路》;20110820(第4期);全文 *

Also Published As

Publication number Publication date
CN108090635A (en) 2018-05-29

Similar Documents

Publication Publication Date Title
CN109146705B (en) Method for detecting electricity stealing by using electricity characteristic index dimension reduction and extreme learning machine algorithm
CN110503245B (en) Prediction method for large-area delay risk of airport flight
CN114445387A (en) Fiberboard quality classification method based on machine vision
CN110197205B (en) Image identification method of multi-feature-source residual error network
CN108090635B (en) Road performance prediction method based on cluster classification
CN113838054B (en) Mechanical part surface damage detection method based on artificial intelligence
CN111222458A (en) Rolling bearing fault diagnosis method based on ensemble empirical mode decomposition and convolutional neural network
CN109816031B (en) Transformer state evaluation clustering analysis method based on data imbalance measurement
CN108847022B (en) Abnormal value detection method of microwave traffic data acquisition equipment
CN111161814A (en) DRGs automatic grouping method based on convolutional neural network
CN110377605B (en) Sensitive attribute identification and classification method for structured data
CN108304567B (en) Method and system for identifying working condition mode and classifying data of high-voltage transformer
CN111626821B (en) Product recommendation method and system for realizing customer classification based on integrated feature selection
CN111914090A (en) Method and device for enterprise industry classification identification and characteristic pollutant identification
CN110826785B (en) High-risk road section identification method based on k-medoids clustering and Poisson inverse Gaussian
CN113918642A (en) Data filtering, monitoring and early warning method based on power Internet of things equipment
CN107480441B (en) Modeling method and system for children septic shock prognosis prediction
CN111967535A (en) Fault diagnosis method and device for temperature sensor in grain storage management scene
CN103902798B (en) Data preprocessing method
CN114548272A (en) Centrifugal pump cavitation state identification method
CN117033912B (en) Equipment fault prediction method and device, readable storage medium and electronic equipment
CN114167237A (en) GIS partial discharge fault identification method and system, computer equipment and storage medium
CN113112067A (en) Method for establishing TFRI weight calculation model
CN110083637B (en) Bridge disease rating data-oriented denoising method
CN115454990A (en) Oil paper insulation data cleaning method based on improved KNN

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant