CN107909822B - SCATS coil checker automatic diagnosis method based on flow and saturation analysis - Google Patents
SCATS coil checker automatic diagnosis method based on flow and saturation analysis Download PDFInfo
- Publication number
- CN107909822B CN107909822B CN201711223421.0A CN201711223421A CN107909822B CN 107909822 B CN107909822 B CN 107909822B CN 201711223421 A CN201711223421 A CN 201711223421A CN 107909822 B CN107909822 B CN 107909822B
- Authority
- CN
- China
- Prior art keywords
- data
- flow
- coil
- distribution
- saturation degree
- 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
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/042—Detecting movement of traffic to be counted or controlled using inductive or magnetic detectors
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
A kind of the step of SCATS coil checker automatic diagnosis method based on flow and saturation analysis, construction training set, is as follows: 1.1: to fixing the date, inquiring all data and be grouped by coil;1.2: drawing the time distribution map of each coil flow, saturation degree, the relational graph and flow of flow saturation degree and the histogram of saturation degree ratio;1.3: counting statistics measure feature;1.4: regarding statistic and annotation results output as training set;The step of coil based on machine learning diagnoses is as follows: 2.1: training set being passed to decision tree classifier, is trained to model;2.2: the selected date for needing to carry out loop data diagnosis, searched targets data;2.3: calculating the statistical nature of target data, target data is described;2.4: classifying to the corresponding coil operating condition of target data and the quality of data.The working condition and the quality of data of the effective detection coil detector of the present invention.
Description
Technical field
The present invention relates to the diagnostic method of the adaptive traffic control system SCATS in Sydney a kind of, especially a kind of SCATS line
Enclose detector automatic diagnosis method.
Background technique
The adaptive traffic control system in Sydney (Sydney Coordinated Adaptive Traffic System, letter
Claim SCATS or abbreviation SCATS system), it is researched and developed by New South Wales,Australia road traffic office (RTA), is current
One of rare several advanced city signal traffic control systems in the world.
Vehicle ring coil detector is data acquisition facility important in SCATS system.When vehicle is by being embedded in road
It can cause the variation of coil magnetic field when loop coil (hereinafter referred to as coil) under face, detector calculates the stream of vehicle accordingly
The traffic parameters such as amount, saturation degree, Period Start Time and length, and it is uploaded to central control system, to meet traffic control system
The needs of system.
Coil checker is usually embedded under the road surface before the Way in stop line of intersection.
SCATS believes that control system can be according to the vehicle flowrate and saturation degree that coil checker detects, dynamic adjustment control ginseng
Number, the flow and saturation degree that coil checker detects.
Saturation degree is that the master data further processing that SCATS system is acquired according to coils such as flow and time headways calculates
Obtained data theoretically have positively related relationship with flow, therefore draw two using saturation degree and flow as reference axis respectively
The binary crelation figure of person, should be presented approximate linear relationship under normal circumstances.
Detection about coil operating condition is all being all the time to pass through people as regular works of equipment O&M link
Work executes, and action, which is also only to look at detector, has no signal to pass back and registered, and can not judge whether coil is passed back
Normally, the workable data of actual traffic situation are able to reflect.
Loop data quality is seldom studied as a problem in science or technical problem, and existing research is mostly to use
The relationship boundary of saturation degree and flow is arranged to judge exceptional data point in the method linearly returned to.However such method can only needle
Dynamic data exchange manual operations to each coil, it is difficult to which automation executes in batches, and it is even more impossible to utilize machine learning method to more
The data problem of subdivision is diagnosed.
Summary of the invention
In order to overcome the shortcomings of prior art can not detection coil detector working condition and the quality of data, the present invention mentions
A kind of working condition of effective detection coil detector and the SCATS line based on flow and saturation analysis of the quality of data are supplied
Enclose detector automatic diagnosis method.
The technical solution adopted by the present invention to solve the technical problems is:
A kind of SCATS coil checker automatic diagnosis method based on flow and saturation analysis, the method includes
A part is the sample data of construction mark, provides training set for machine learning;Second part is based on machine learning method pair
Coil operating condition and the quality of data are judged;
In first part, construct training set the step of it is as follows:
Step 1.1: to fixing the date, inquiring all data and be grouped by coil;
Step 1.2: drawing the time distribution map V-t of each coil flow, saturation degree, DS-t, the relationship of flow saturation degree
Scheme the histogram Freq (DS/V) of DS-V and flow and saturation degree ratio;
Step 1.3: calculating a series of statistics measure features of V-t, DS-t, DS-V, Freq (DS/V), including maximum/minimum
Value/mean value/kurtosis/the degree of bias/5%/10%/50%/90%/95% cumulative probability quantile, and to coil operating condition and data matter
Amount is demarcated;
Step 1.4: the statistic calculated above and annotation results are exported to the training set as subsequent processes;
The step of second part, the coil based on machine learning diagnoses, is as follows:
Step 2.1: above-mentioned training set being passed to decision tree classifier, model is trained;
Step 2.2: the selected date for needing to carry out loop data diagnosis, searched targets data;
Step 2.3: calculating the statistical nature of target data, target data is described;
Step 2.4: classifying to the corresponding coil operating condition of target data and the quality of data.
Further, the saturation degree refers to SCATS internal system, is based on historical traffic and present flow rate, calculated to work as
Ratio of the vehicle number that preceding phase is mutually passed through relative to traffic volume maximum under saturation state, unit %;The phase refers to friendship
The time span of green light signals of prong opening and closing, the green light Qi Liangwei phase time started, the corresponding several lanes of the phase
Vehicle is let pass, and phase finish lamp is green to redden, and system enters next phase, another group of lane of letting pass.
Further, in the step 2.4, using Time Series Method, the continuity of flow V is checked, i.e., in vehicle flowrate
In the biggish period, flow V whether multiple periods are significantly lower than normal value, to judge coil communication failure.
Further, in the step 2.4, using the approximate line style relationship of saturation degree DS and flow V, to by saturation degree-
The data point distribution form that two dimensions of flow are constituted is judged, judges the quality of data of coil.
In the step 2.4, using the probability distribution shape of saturation degree and the ratio DS/V of flow, the data of coil are judged
Quality.
In the step 1.2, the process for drawing V-t, DS-t, DS-V and Freq (DS/V) distribution map is as follows:
In distribution from flow about the time, it is clear that the flow at a crossing under normal circumstances in one day
Undulation situation, also calculating saturation data DS, obtain DS-t figure with V-t relational graph;Using saturation degree as ordinate, stream
Amount is abscissa, obtains DS-V relational graph, and under normal circumstances, the relationship of approximately linear is presented with flow for saturation degree, and data are basic
On pool approximate oblique line;If it is but very high to satisfy when flow very little, the linear approximate relationship of flow and saturation degree is significantly not
It sets up, it is not possible that this thing happens in reality, therefore is speculated as data exception;
The histogram Freq (DS/V) of saturation degree and flow ratio, abscissa indicate the value of DS/V, and ordinate indicates ratio
The frequency of appearance observes the frequency distribution of saturation degree and flow ratio: under normal circumstances, the distribution of DS/V has a peak
Value, matched curve are distributed in right avertence;DS/V under coil operating condition is abnormal is distributed then without obvious peak value, illustrates the loop data
It is abnormal.
In the step 1.3, the process for calculating the statistic of loop data is as follows:
Volume of data feature is extracted to describe 4 kinds of distribution maps, the data are statistical nature, the system including examining distribution
Metering, the statistic for describing distribution shape, and embody the statistic of distribution accumulated probability;
Ks examines a kind of test of fitness of fot, and whether inspection data meets certain distribution, the Limit Distribution shape of statistic
Formula are as follows:
Kurtosis reflects the sharp degree of figure peak potion, and the degree of bias is the measurement of statistical data distribution skew direction and degree, the two
The shape of distribution can be described, formula is as follows:
Cumulative distribution function describes the probability distribution of a real number stochastic variable, indicates all values less than or equal to flow x
The sum of probability of appearance is the integral of probability density function, and what cumulative probability quantile indicated is for any flow x, institute
The ratio for having the number smaller than x shared in data set.
It is as follows according to statistical nature artificial judgment coil operating condition and the quality of data, process:
Judgement based on V and DS:
Apparent coil can be directly based upon independent V extremely or DS is judged, if V and DS is zero;Or V or
The median of DS has been more than given threshold value, i.e., directly judges that coil damages;
Judged by the product of flow and time span at " cliff of displacement ", when product is more than that a threshold value A is sentenced
It is disconnected to occur communication abnormality at this time, if product is more than bigger threshold value B, illustrate " the flow cliff of displacement " at this to entire traffic condition
Analyzing influence is larger, is judged as that loss of data is serious;
Judgement based on Freq (DS/V): the statistical tool of Python can help to calculate the accumulation of data 95%, 99%
Probability quantile illustrates in this section of section if the corresponding numerical value of data at 95% is greater than the maximum value of artificial observation
There are many points, their data far from normal condition, these are all noises;If the corresponding numerical value of data at 99% is greater than artificial
The maximum value of observation then illustrates that data have a small amount of noise;
If data have passed through above inspection, then it is assumed that coil working is normal.
The method also includes Part III: the coil through artificial determination, recalibration determines as a result, determining as machine learning
The study collection of plan tree method learns the statistical nature threshold value in judgement coil fault and the quality of data based on study collection
And optimization, coil operating condition and the quality of data with realization based on machine learning method determine automatically.
Technical concept of the invention are as follows: the SCATS coil checker automatic diagnosis method based on flow and saturation analysis,
First, it is determined that whether current coil normally passes flow and saturation data back, if traffic loss or super full for a long time occur
The significant data exception situation with degree etc.;Secondly, judge to whether there is normal linear approximate relationship between saturation degree and flow,
If the linear approximate relationship is significantly invalid, the working condition of coil is abnormal;Finally, judging data with the presence or absence of few
The glitch such as amount noise, temporary communications interruption also restore just after a short period of time although these failures do not influence the use of data
Often, but frequently occurring for glitch similarly means that the working life of coil may soon terminate.
Based on analyzing SCATS system coil flow and saturation degree, judge whether coil working state is normal, with
And existing for tentative data the problem of specific defect.Specifically, the main problem of coil includes:
1) data are seriously lost, and the V and DS that whole day or daytime most of period are passed back are lasting 0 value.
2) data are seriously supersaturated, and whole day or most of period DS value in one day all remain above 120 value.
3) communicating interrupt the long period or relatively frequently occurs at flow on daytime biggish period (such as 7:00-22:00)
Flow falls to 0 value from the larger value suddenly, restores normal phenomenon again later.
4) the linear relationship disorder of data exception, DS and V, then meaning DS or V, there are logic errors, and such data cannot
Characterize true traffic flow character.
The coil operating condition and subdivided data quality problems that this method can judge are as follows:
A. coil is normal: coil working is normal;
B. coil damages: no data;
C. coil damages: Traffic Anomaly;
D. coil damages: saturation anomaly;
E. coil damages: flow and saturation anomaly;
F. data exception: mass data noise;
G. data exception: loss of data is serious;
H. data noise: low volume data noise;
I. communication abnormality: there are Communications.
Beneficial effects of the present invention are mainly manifested in: the working condition and the quality of data of effective detection coil detector, base
It is analyzed in SCATS system coil flow and saturation degree, judges whether coil working state normal and tentative data
Existing specific defect.
Detailed description of the invention
Fig. 1 is four kinds of form schematic diagrames of flow-time distribution map (V-t).
Fig. 2 is two kinds of distribution maps of flow Yu saturation degree relationship.
Fig. 3 is two kinds of distribution maps of saturation degree Yu flow ratio histogram.
Fig. 4 is cumulative probability quantile schematic diagram.
Fig. 5 is the schematic diagram of coil damage (no data).
Fig. 6 is the schematic diagram of coil damage (flow saturation anomaly).
Fig. 7 is the schematic diagram of communication abnormality.
Fig. 8 is the serious schematic diagram of loss of data.
Fig. 9 is the schematic diagram of mass data noise.
Figure 10 is the schematic diagram of low volume data noise.
Figure 11 is the normal schematic diagram of coil working.
Specific embodiment
The invention will be further described below in conjunction with the accompanying drawings.
Referring to Fig.1~Figure 11, a kind of SCATS coil checker automatic diagnosis method based on flow and saturation analysis,
First part is the sample data of construction mark, provides training set for machine learning;Second part is based on machine learning method
Coil operating condition and the quality of data are judged.
The step of first part, construction training set, is as follows:
Step 1.1: to fixing the date, inquiring all data and be grouped by coil;
Step 1.2: drawing the time distribution map (V-t, DS-t) of each coil flow, saturation degree, the pass of flow saturation degree
The histogram (Freq (DS/V)) of system figure (DS-V) and flow and saturation degree ratio;
Step 1.3: calculate a series of statistics measure features of V-t, DS-t, DS-V, Freq (DS/V), it is main include it is maximum/
Minimum value/mean value/kurtosis/the degree of bias/5%/10%/50%/90%/95% cumulative probability quantile etc., and to coil operating condition and
The quality of data is demarcated;
Step 1.4: the statistic calculated above and annotation results are exported to the training set as subsequent processes;
The step of second part, the coil based on machine learning diagnoses, is as follows:
Step 2.1: above-mentioned training set being passed to decision tree classifier, model is trained;
Step 2.2: the selected date for needing to carry out loop data diagnosis, searched targets data;
Step 2.3: calculating the statistical nature of target data, target data is described;
Step 2.4: classifying to the corresponding coil operating condition of target data and the quality of data.
In the step 1.2, V-t, DS-t, DS-V and Freq (DS/V) distribution map is drawn
By observation V-t figure, several frequently seen coil operating condition can be significantly differentiated, as shown in Fig. 1.
Shown in Fig. 1 (a), in the distribution from flow about the time, it is clear that the positive reason of the flow at a crossing
In intraday undulation situation under condition: flow is lower between 0:00 to 6:00;After 6:00 enter morning peak, flow gradually on
It rises;Flow on daytime maintains higher level, until flow slowly declines after 18:00.
Fig. 1 (b) shows that a kind of typical coil damaged condition, data on flows are all zero, i.e., the coil damages completely,
Flow and saturation data cannot normally be passed back.
It is shown in Fig. 1 (c) " the flow cliff of displacement ", but flow, outside the period for the cliff of displacement occur, the rule of fluctuation is overall
Normally, it is inferred that being provisional communication failure occurred.
In Fig. 1 (d), the fluctuation pattern of generally flow is normal, but has a data point that flow mutation has occurred, therefore sentence
Break as system noise, such noise can be rejected in the later period.
In addition to flow, SCATS system also calculates saturation data DS.DS is data relevant to V, additionally by vehicle
The influence of type, but the relationship of near-linear dependency should generally be presented therebetween, therefore DS-t figure and V-t relational graph
Shape is very close, and it is also similar abnormal mistake occur.Using saturation degree as ordinate, flow is abscissa, and available DS-V is closed
System's figure, is shown in Fig. 2.Under normal circumstances, the relationship of approximately linear is presented in saturation degree and flow, and data substantially pool approximate oblique
Line, as shown in Fig. 2 (a).
The situation as shown in Fig. 2 (b), full but very high when flow very little, the linear approximate relationship of flow and saturation degree is significant
It is invalid, it is not possible that this thing happens in reality, therefore it is speculated as data exception.Such data cannot be directly used to science
Analysis, therefore operation maintenance personnel need to be reminded to check coil in time.
Fig. 3 shows the histogram (Freq (DS/V)) of saturation degree and flow ratio, is observed that saturation degree in figure
With the frequency distribution of flow ratio: abscissa indicates the value of DS/V, and ordinate indicates the frequency that ratio occurs.Fig. 3 (a) is shown in
Under normal circumstances, the distribution of DS/V has a peak value, and matched curve is distributed in right avertence.DS/V points under coil operating condition is abnormal
Cloth illustrates loop data exception, situation shown by Fig. 3 (b) is a kind of typical data exception then without obvious peak value.
In the step 1.3, the process for calculating the statistic of loop data is as follows:
It, can be with convenient for distinguishing the working condition and the quality of data of coil in order to quantify the information contained in different graphic
Volume of data feature is extracted to describe above-mentioned 4 kinds of distribution maps.These data characteristicses pass through after the preliminary observation that is distributed to figure
Selection, predominantly statistical nature, it is tired including the statistic examined the statistic being distributed, describe distribution shape, and embodiment distribution
The statistic of probability is counted, the distribution of data representation figure can be used in detail.
Whether it is a kind of test of fitness of fot that ks, which examines (Kolmogorov-Smirnov test), can be accorded with inspection data
Close certain distribution.The Limit Distribution concrete form of its statistic are as follows:
Kurtosis (Kurtosis) reflects the sharp degree of figure peak potion, and the degree of bias (Skewness) is statistical data distribution deflection side
To the measurement with degree, the two can describe the shape of distribution, specific formula is as follows:
Cumulative distribution function (cdf) describes the probability distribution of a real number stochastic variable, indicates all and is less than or equal to stream
Measure the sum of the probability that the value of x occurs, be the integral of probability density function (pdf), and cumulative probability quantile indicates be for
Any flow x, all numbers smaller than x ratio shared in data set.
As shown in figure 4, dotted line represents different cumulative probability quantiles from the point of cdf curve intersection meaning, such as see
Examine flow value representated by 95% quantile, it can be seen that flow value largely all represents under flow value in 95% quantile.
Multiple quantiles are taken, representative flow value is observed, that is, can determine whether data distribution has exception.
By these data characteristicses, distribution described in step 1.2 is described, table 1 is the statistics for Coil Detector
Feature (V-t figure part).
Table 1
According to statistical nature artificial judgment coil operating condition and the quality of data
By the way that some criterion can be summed up to determine line in conjunction with artificial experience according to several figures and statistical nature
Operating condition and the quality of data are enclosed, two classes: the flow that 95% quantile represents generally are divided into
Judgement based on V and DS:
Apparent coil can be directly based upon independent V extremely or DS is judged, if V and DS is zero such as Fig. 5 institute
Show;Or the median of V or DS has been more than given threshold value, as shown in fig. 6, directly judging that coil damages.
Communication abnormality will cause data on flows and " cliff of displacement " occurs.If occurring " the flow cliff of displacement ", carrying out judgement is to flow greatly
It measures period appearance " cliff of displacement " or " cliff of displacement " occurs in small flow-time section.There is " cliff of displacement " at big flow to whole flow number
It is very big according to influencing, but caused by " cliff of displacement " not necessarily communication abnormality at small flow, but this period is really without vehicle
Pass through, communication abnormality can be just judged as by being only chronically at " cliff of displacement ".Therefore, two kinds of situations need to pass through the flow at " cliff of displacement "
Judged with the product of time span.When product is more than that a threshold value A judges occur communication abnormality at this time, if product is super
Bigger threshold value B is crossed, illustrates that " the flow cliff of displacement " at this is larger to the analyzing influence of entire traffic condition, is judged as loss of data
Seriously.For influence degree of the prominent flow in product, flow value can be carried out to a square calculating, area is in two kinds of situation.
Judgement based on Freq (DS/V): the data exception of another kind of coil is needed through analysis DS's and V than more covert
Proportionate relationship could judge that a large amount of saturation degrees are big and flow is small as there is a situation where in data.This feature table in histogram
It is now very big for the maximum abscissa zone of histogram, and frequency is excessively high.The statistical tool of Python can help to calculate data
95%, 99% cumulative probability quantile is said if the corresponding numerical value of data at 95% is greater than the maximum value of artificial observation
Bright to have many points in this section of section, their data far from normal condition, these are all noises.If the data pair at 99%
The numerical value answered is greater than the maximum value of artificial observation, then illustrates that data have a small amount of noise.
If data have passed through above inspection, then it is assumed that coil working is normal.
Based on machine learning algorithm detection coil operating condition and the quality of data, process is as follows:
Training set will be formed by the characteristic quantity of manual sort and tag along sort, incoming decision tree classifier carries out classification instruction
Practice.Wherein last column of training set are target classification variables, remaining is all characteristic quantity.
Decision tree classifier used in the present invention is the packet scikit-learn that machine learning is used in Python.
Another group of not incoming classifier progress class test of the data at same date and crossing, classifier are chosen from database
After output winding status number, number and state are matched again in program, the results are shown in Table 2 for final output.
Table 2
As shown in Table 2, decision tree can distinguish multiple classifications.
It can clearly judge whether current coil normally passes flow and saturation data back from result, whether there is or not show
The data exception situation of work whether there is the glitch such as a small amount of noise, temporary communications interruption, realizes and calculates with machine learning
Method diagnoses SCATS coil checker by analysis flow and saturation degree automatically.
The result of result and manual sort Jing Guo decision tree classification is compared, it can be seen that decision tree from table 3
The accuracy rate of classification is also higher.
Table 3
Method used in the present embodiment has the characteristics that big data quantity operation, main to be realized by Python, key point
Mainly in the following.
Python programming: the program write in the present invention has used multiple packets in Python, such as uses cx_
Oracle connection database carries out data processing using pandas and numpy, and scipy.stats is for statistical analysis and makes
It is drawn etc. with matplotlib.
Statistic selection: there are many statistical methods in the statistics packet of Python, select suitable statistical method also very heavy
It wants.Such as the quantile for the degree of bias, kurtosis and several cumulative probabilities used in this method.
Coil state judgement: from distinguishing from different state needs manually first to carry out and classify in loop data, because
This needs to cooperate the canalization information for understanding crossing, and has the observation experience of long period, ability to the wagon flow rule of corresponding intersection
Judge whether data are reasonable.
Claims (9)
1. a kind of SCATS coil checker automatic diagnosis method, it is characterised in that: the method includes first part be construction mark
The sample data of note provides training set for machine learning;Second part is based on machine learning method to coil operating condition and data
Quality is judged;
In first part, construct training set the step of it is as follows:
Step 1.1: to fixing the date, inquiring all data and be grouped by coil;
Step 1.2: drawing the time distribution map V-t, DS-t, the relational graph DS- of flow saturation degree of each coil flow, saturation degree
The histogram Freq (DS/V) of V and flow and saturation degree ratio;
Step 1.3: a series of statistics measure features of calculating V-t, DS-t, DS-V, Freq (DS/V), including maximum/minimum value/
Value/kurtosis/degree of bias and/or/5%/10%/50%/90%/95% cumulative probability quantile, and to coil operating condition and data matter
Amount is demarcated;
Step 1.4: the statistic calculated above and annotation results are exported to the training set as subsequent processes;
The step of second part, the coil based on machine learning diagnoses, is as follows:
Step 2.1: above-mentioned training set being passed to decision tree classifier, model is trained;
Step 2.2: the selected date for needing to carry out loop data diagnosis, searched targets data;
Step 2.3: calculating the statistical nature of target data, target data is described;
Step 2.4: classifying to the corresponding coil operating condition of target data and the quality of data.
2. SCATS coil checker automatic diagnosis method as described in claim 1, it is characterised in that: the saturation degree refers to
SCATS internal system is based on historical traffic and present flow rate, and the vehicle number that calculated current phase is mutually passed through is relative to saturation
The ratio of maximum traffic volume, unit % under state;The phase refers to the time span of the green light signals in intersection opening and closing,
The vehicle of green light Qi Liangwei phase time started, the corresponding several lanes of the phase are let pass, and phase finish lamp is green to redden, and is
System enters next phase, another group of lane of letting pass.
3. SCATS coil checker automatic diagnosis method as claimed in claim 1 or 2, it is characterised in that: the step 2.4
In, using Time Series Method, check the continuity of flow V, i.e., in the vehicle flowrate biggish period, flow V whether multiple weeks
Phase is significantly lower than normal value, to judge coil communication failure.
4. SCATS coil checker automatic diagnosis method as claimed in claim 1 or 2, it is characterised in that: the step 2.4
In, using the approximate line style relationship of saturation degree DS and flow V, to the data point distribution being made of two dimensions of saturation degree-flow
Form is judged, judges the quality of data of coil.
5. SCATS coil checker automatic diagnosis method as claimed in claim 1 or 2, it is characterised in that: the step 2.4
In, using the probability distribution shape of saturation degree and the ratio DS/V of flow, judge the quality of data of coil.
6. SCATS coil checker automatic diagnosis method as claimed in claim 1 or 2, it is characterised in that: the step 1.2
In, the process for drawing V-t, DS-t, DS-V and Freq (DS/V) distribution map is as follows:
In distribution from flow about the time, it is clear that the flow at a crossing under normal circumstances in intraday wave
Dynamic fluctuating situation, also calculating saturation data DS, obtain DS-t figure and V-t relational graph;Using saturation degree as ordinate, flow is cross
Coordinate obtains DS-V relational graph, and under normal circumstances, the relationship of saturation degree and flow presentation approximately linear, data substantially converge
At approximate oblique line;If flow very little and saturation degree is very high, the linear approximate relationship of flow and saturation degree is significantly invalid, In
It is not possible that this thing happens in reality, therefore it is speculated as data exception;
The histogram Freq (DS/V) of saturation degree and flow ratio, abscissa indicate the value of DS/V, and ordinate indicates that ratio occurs
Frequency, observe the frequency distribution of saturation degree and flow ratio: under normal circumstances, the distribution of DS/V has a peak value,
Matched curve is distributed in right avertence;DS/V under coil operating condition is abnormal is distributed then without obvious peak value, illustrates that the loop data is different
Often.
7. SCATS coil checker automatic diagnosis method as claimed in claim 6, it is characterised in that: in the step 1.3,
The process for calculating the statistic of loop data is as follows:
Volume of data feature is extracted to describe 4 kinds of distribution maps, the data are statistical nature, the statistics including examining distribution
The statistic of amount, description distribution shape, and embody the statistic of distribution accumulated probability;
Ks examines a kind of test of fitness of fot, and whether inspection data meets certain distribution, the Limit Distribution form of statistic are as follows:
Kurtosis reflects the sharp degree of figure peak potion, and the degree of bias is the measurement of statistical data distribution skew direction and degree, and the two can
The shape of distribution is described, formula is as follows:
Cumulative distribution function describes the probability distribution of a real number stochastic variable, indicates that all values less than or equal to flow x occur
The sum of probability, be the integral of probability density function, and cumulative probability quantile indicates be for any flow x, it is all to compare x
Small number ratio shared in data set.
8. SCATS coil checker automatic diagnosis method as claimed in claim 7, it is characterised in that: according to statistical nature people
Work judges that coil operating condition and the quality of data, process are as follows:
Judgement based on V and DS:
Apparent coil can be directly based upon independent V extremely or DS is judged, if V and DS is zero;Or in V or DS
Digit has been more than given threshold value, i.e., directly judges that coil damages;
Judged by the product of flow and time span at " cliff of displacement ", when product is more than that a threshold value A judges this
When there is communication abnormality, if product is more than bigger threshold value B, illustrate the analysis of " the flow cliff of displacement " to entire traffic condition at this
It is affected, is judged as that loss of data is serious;
Judgement based on Freq (DS/V): the statistical tool of Python can help to calculate the cumulative probability of data 95%, 99%
Quantile illustrates have very in this section of section if the corresponding numerical value of data at 95% is greater than the maximum value of artificial observation
Multiple spot, their data far from normal condition, these are all noises;If the corresponding numerical value of data at 99% is greater than artificial observation
Maximum value, then illustrate that data have a small amount of noise;
If data have passed through above inspection, then it is assumed that coil working is normal.
9. SCATS coil checker automatic diagnosis method as described in claim 1, it is characterised in that: the method also includes
Part III: the coil through artificial determination, recalibration determines the study collection as a result, as machine learning traditional decision-tree, is based on
Study collection is based on machine to determining that the statistical nature threshold value in coil fault and the quality of data is learnt and optimized to realize
The coil operating condition and the quality of data of learning method determine automatically.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711223421.0A CN107909822B (en) | 2017-11-29 | 2017-11-29 | SCATS coil checker automatic diagnosis method based on flow and saturation analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711223421.0A CN107909822B (en) | 2017-11-29 | 2017-11-29 | SCATS coil checker automatic diagnosis method based on flow and saturation analysis |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107909822A CN107909822A (en) | 2018-04-13 |
CN107909822B true CN107909822B (en) | 2019-11-22 |
Family
ID=61849159
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711223421.0A Active CN107909822B (en) | 2017-11-29 | 2017-11-29 | SCATS coil checker automatic diagnosis method based on flow and saturation analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107909822B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110610607B (en) * | 2019-07-26 | 2020-12-25 | 浙江工业大学 | Traffic-based SCATS control scheme operation cycle automatic diagnosis method and system |
CN111554371A (en) * | 2020-03-30 | 2020-08-18 | 平安医疗健康管理股份有限公司 | Internet data processing method and device, computer equipment and storage medium |
CN113781802B (en) * | 2021-09-17 | 2023-02-28 | 清华大学 | Road intersection traffic state measuring method and device and electronic equipment |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5621661A (en) * | 1993-08-04 | 1997-04-15 | Electromatic (Proprietary) Limited | Inductive loop diagnostic interface |
KR100719856B1 (en) * | 2006-02-06 | 2007-05-18 | 한국건설기술연구원 | Method for ascertaining and correcting velocity error of vehicle detector using auto-calibration algorithm |
CN101739814A (en) * | 2009-11-06 | 2010-06-16 | 吉林大学 | SCATS coil data-based traffic state online quantitative evaluation and prediction method |
CN102169630A (en) * | 2011-03-31 | 2011-08-31 | 上海电科智能***股份有限公司 | Quality control method of road continuous traffic flow data |
CN102708378A (en) * | 2012-04-28 | 2012-10-03 | 浙江工业大学 | Method for diagnosing fault of intelligent traffic capturing equipment based on image abnormal characteristic |
CN103593973A (en) * | 2012-12-18 | 2014-02-19 | 北京科技大学 | Urban road traffic situation assessment system |
CN105931458A (en) * | 2016-05-05 | 2016-09-07 | 杭州智诚惠通科技有限公司 | Method for evaluating credibility of road traffic flow detection device |
-
2017
- 2017-11-29 CN CN201711223421.0A patent/CN107909822B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5621661A (en) * | 1993-08-04 | 1997-04-15 | Electromatic (Proprietary) Limited | Inductive loop diagnostic interface |
KR100719856B1 (en) * | 2006-02-06 | 2007-05-18 | 한국건설기술연구원 | Method for ascertaining and correcting velocity error of vehicle detector using auto-calibration algorithm |
CN101739814A (en) * | 2009-11-06 | 2010-06-16 | 吉林大学 | SCATS coil data-based traffic state online quantitative evaluation and prediction method |
CN102169630A (en) * | 2011-03-31 | 2011-08-31 | 上海电科智能***股份有限公司 | Quality control method of road continuous traffic flow data |
CN102708378A (en) * | 2012-04-28 | 2012-10-03 | 浙江工业大学 | Method for diagnosing fault of intelligent traffic capturing equipment based on image abnormal characteristic |
CN103593973A (en) * | 2012-12-18 | 2014-02-19 | 北京科技大学 | Urban road traffic situation assessment system |
CN105931458A (en) * | 2016-05-05 | 2016-09-07 | 杭州智诚惠通科技有限公司 | Method for evaluating credibility of road traffic flow detection device |
Non-Patent Citations (1)
Title |
---|
基于SCATS的动态交通数据预处理方法研究;魏国容;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20140515(第5期);第47-52页 * |
Also Published As
Publication number | Publication date |
---|---|
CN107909822A (en) | 2018-04-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107909822B (en) | SCATS coil checker automatic diagnosis method based on flow and saturation analysis | |
CN104091035B (en) | Health monitoring method for effective loads of space station based on data-driven algorithm | |
CN109583680B (en) | Power stealing identification method based on support vector machine | |
CN101718634B (en) | Equipment state comprehensive dynamic alarming method based on multivariate probability model | |
CN104712542B (en) | A kind of reciprocating compressor sensitive features based on Internet of Things are extracted and method for diagnosing faults | |
CN106206346A (en) | Measurement Sampling Method with Sampling Rate Determination Mechanism | |
CN100454354C (en) | Road traffic accident multi-happening section identifying method | |
JP6778132B2 (en) | Abnormality diagnosis system for equipment | |
CN106296435A (en) | A kind of mine gas monitoring disorder data recognition method | |
CN106326913A (en) | Money laundering account determination method and device | |
CN110515781B (en) | Complex system state monitoring and fault diagnosis method | |
CN101738998B (en) | System and method for monitoring industrial process based on local discriminatory analysis | |
CN108090657A (en) | Oil & Gas Storage facility risk assessment based on Xiu Hate control theories and probabilistic neural network manages system and method with on-line early warning | |
CN106482967A (en) | A kind of Cost Sensitive Support Vector Machines locomotive wheel detecting system and method | |
CN109389325B (en) | Method for evaluating state of electronic transformer of transformer substation based on wavelet neural network | |
CN112766301B (en) | Oil extraction machine indicator diagram similarity judging method | |
CN109543897A (en) | Local government debt risk dynamic evaluation and prediction technique | |
CN103488561A (en) | kNN (k-nearest neighbor) fault detection method for online upgrading master sample model | |
CN108108890A (en) | A kind of defective workmanship risk analysis method based on process incidence relation | |
CN109240276B (en) | Multi-block PCA fault monitoring method based on fault sensitive principal component selection | |
CN106875613A (en) | A kind of fire alarm Situation analysis method | |
CN109871654A (en) | A kind of water bursting source recognition methods based on PCA and Method of Gray | |
CN112036087A (en) | Multi-strategy fused nuclear power key equipment fault diagnosis method and system | |
CN108510180A (en) | The computational methods of performance interval residing for a kind of production equipment | |
CN109856494A (en) | A kind of Diagnosis Method of Transformer Faults based on support vector machines |
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 | ||
CP01 | Change in the name or title of a patent holder | ||
CP01 | Change in the name or title of a patent holder |
Address after: 310012 1st floor, building 1, 223 Yile Road, Hangzhou City, Zhejiang Province Patentee after: Yinjiang Technology Co.,Ltd. Address before: 310012 1st floor, building 1, 223 Yile Road, Hangzhou City, Zhejiang Province Patentee before: ENJOYOR Co.,Ltd. |