The specific embodiment
Below in conjunction with accompanying drawing, technical scheme of the present invention is carried out concrete elaboration.
As shown in Figure 1, the detection method of a kind of flexible pavement data of the present invention may further comprise the steps:
Step 1: the acquisition testing data form original data sequence: flexible pavement detected index collection detects data, and all corresponding pile No. information of each detection data, sort from small to large according to pile No. information, form the original data sequence that detects data.These detect data is the data of directly gathering, and does not pass through any processing.
In this step 1, detect index and comprise pavement distress, ride quality, pavement strength, pavement skid resistance condition and the rut degree of depth.According to " asphalt highway maintenance technology standard " requirement (JTJ073.2-2001) aforementioned detection index is detected data acquisition.The acquisition testing data demand is evenly gathered according to certain intervals, and the size in sampling interval will be chosen with reference to " asphalt highway maintenance technology standard " requirement (JTJ073.2-2001).Every kind is detected index and forms one group of original data sequence respectively.For guarantee detecting the representativeness of data sample, require every group of pending detection data, promptly the data number in every group of original data sequence is greater than 100.Detection data after gathering by the pile No. size, are sorted from small to large, form original data sequence.
Step 2: check the detection data of collection in the step 1 and misdata is mended survey, form data sequence d.
In this step 2, the inspection that detects data is comprised improper value inspection and uniformity inspection:
A. improper value inspection: choose a data sample that detects index, calculate with reference to following formula:
Formula (1)
In the formula (1):
x
iI data in the expression data sample;
The average of
expression data sample;
S representes the standard deviation of data sample.
If x
iSatisfy formula (1), then this point is mended survey.Still do not satisfy above-mentioned condition if mend the survey result, then this value is regarded as normal value and carries out follow-up data.
B. uniformity inspection: inspection is checked through improper value as follows, and through mending the uniformity of the data of surveying:
Uniformity=| the poor-sampling interval of adjacent 2 pile No. |/sampling interval formula (2)
Uniformity between adjacent 2 should satisfy formula (3):
Uniformity<=1 formula (3)
If do not satisfy above-mentioned condition, then should mend survey, until satisfying above-mentioned condition in the relevant position.
Every kind detect the detection data of gathering in the index through inspection and mend survey after, form data sequence d, the data bulk that wherein comprises is n, promptly d is by d
1, d
2, d
3, d
4, d
5... d
I-1, d
i, d
I+1... d
N-1, d
nForm.
Step 3: smoothing processing: the detection data to step 2 obtains are carried out smoothing processing, obtain the detection data after the smoothing processing.
In this step 3, data smoothing is handled and is referred to eliminate the influence of the peak-to-valley value of one group of data to overall distribution, and it is smooth-going smooth to make data and curves try one's best.
Concrete smoothing processing process is carried out according to formula (4):
ds
1=d
1;
ds
2=(d
1+d
2+d
3)/3;
ds
3=(d
1+d
2+d
3+d
4+d
5)/5;
.......
Ds
i=(d
I-(span-1)/2+ ...+d
I-1+ d
i+ d
I+1+ ...+d
I+ (span-1)/2)/span; Formula (4)
.......
ds
n-2=(d
n-4+d
n-3+d
n-2+d
n-1+d
n)/5;
ds
n-1=(d
n-2+d
n-1+d
n)/3;
ds
n=d
n;
In the formula (4):
In this formula: by ds
1, ds
2, ds
3... ds
i... ds
N-2, ds
N-1, ds
nThe data sequence ds that forms, the data sequence after the expression smoothing processing, total amount is n, wherein, ds
1Data in the expression data sequence after first smoothing processing, ds
2Data in the expression data sequence after second smoothing processing, ds
3Data in the expression data sequence after the 3rd smoothing processing ..., ds
iData in the expression data sequence after i smoothing processing, ds
N-2Data in the expression data sequence after n-2 smoothing processing, ds
N-1Data in the expression data sequence after n-1 smoothing processing, ds
nData in the expression data sequence after n smoothing processing.By d
1, d
2, d
3, d
4, d
5... d
I (span-1)/2... d
I-1, d
i, d
I+1... d
I+ (span-1)/2... d
N-4, d
N-3, d
N-2, d
N-1, d
nThe data sequence d that forms is the data sequence that step 2 obtains, and total amount is n, and is consistent with the total amount of data sequence ds.Span is level and smooth scope, and value is less than or equal to n, and span is the odd number in 5 to 20.The span value is crossed conference and is caused section length excessive, loses the segmentation meaning, and the span value is too small then can to cause segmentation too meticulous, is unfavorable for schematic design.
Step 4: to detecting the data section of presorting: be the control index with the extreme difference; The detection data section of presorting after the smoothing processing that step 3 is obtained; Extreme difference in each segmentation is less than or equal to maximum and allows extreme difference, and writes down the pile No. information of each waypoint in the section of presorting, and forms the section of presorting.
In this step 4, to step 3 form level and smooth after data sequence ds, carry out preliminary segmentation and be called the section of presorting that detects data.This phase method of presorting mainly is to serve as the control index with extreme difference (refer in one group of data maximum value and minimum value poor); Promptly allow under the extreme difference condition the ds section of presorting in given maximum; And after the section of presorting, the extreme difference of each section all is not more than maximum and allows extreme difference.The section of presorting principle flow chart is as shown in Figure 2:
At first, make i=1, set up an interim empty array TA, the data total amount of establishing among the data sequence ds in the step 3 is n, carries out following operation:
(a) the interim array TA of initialization is just with the whole element zero clearings among the array TA;
(b) with i data ds among the data sequence ds
iInsert interim array TA;
Whether the extreme difference of (c) judging array TA is less than or equal to maximum is allowed extreme difference rmax;
(d) if the extreme difference of array TA is allowed extreme difference smaller or equal to maximum, then the value with i adds 1; If this moment, i was smaller or equal to n, then turn back to (b); If this moment, i was greater than n, then segmentation finishes, the pile No. information that arrangement is returned;
(e), then write down the corresponding pile No. k of i-1 data if the extreme difference of array TA is allowed extreme difference greater than maximum
I-1If this moment, i was less than n, then turn back to (a); If this moment, i equaled n, then segmentation finishes, the pile No. information that arrangement is returned.
The maximum of mentioning in the step 4 allows that extreme difference rmax has determined number of fragments.Maximum allows that extreme difference rmax is big more, and the number of fragments that the section of presorting obtains is few more, and maximum allows that extreme difference rmax is more little, and the number of fragments that the section of presorting obtains is many more.Maximum allows that the value of extreme difference rmax should be by follow-up design scheme decision; In other words; If a certain detection index is after the difference value in two highway sections reaches a certain threshold; Can be so that the design scheme generation marked change of two-way section, then this threshold maximum of promptly can be used as the section of presorting is allowed extreme difference rmax.This method recommends the maximum of flexure index to allow that extreme difference rmax gets 10-30 (0.01mm), and the maximum of rut index allows that extreme difference rmax gets 0.5mm-2mm, and the maximum of other index allows that extreme difference rmax value chooses with reference to preamble said.
According to the pile No. information of returning, can carry out segmentation to the data sequence d that obtains in the step 2 with the pile No. that obtains as waypoint.Data sequence d can obtain a comparatively meticulous segmentation through after the section of presorting, but because this segmentation is often too meticulous, and be unfavorable for schematic design.Especially at some local location, section length is often very little.This means that these less segmentations only need Local treatment to get final product, there is no need to classify it as section separately.Therefore, meet the needs of practical application, be necessary the section of presorting result is merged, thereby obtain being convenient to the new segmentation of schematic design for making segmentation.
Step 5: search for the first time exceptional value: 4 sections of presorting that obtain set by step, the detection data that step 2 obtains are carried out segmentation, exceptional value is carried out in each segmentation searches, and in detecting data mark in addition.
Before the section of presorting to step 4 merges processing, must find out the abnormal data in the respectively section of presorting of step 4 earlier, prevent that it is combined operation and produces and disturb.With reference to " highway subgrade road surface on-the-spot test rules " (JTG E60-2008); If detecting the size and the road conditions level of index is inversely proportional to; All detection data that add the twice standard deviation more than or equal to this section average are as exceptional value in then will this section, promptly shown in the formula (5), and record in addition; Be directly proportional with the road conditions level if detect the size of index, then will this section in all detection data that deduct the twice standard deviation smaller or equal to this section average as exceptional value, promptly shown in the formula (6), and record in addition.
The data sequence d that step 1 is obtained, 4 segmentations that obtain set by step, carry out as judging:
Formula (5)
Formula (6)
In formula (5) and formula (6):
x
IjJ data among the expression data sequence d in the i segmentation;
The average of i data segments among
expression data sequence d;
S
iThe standard deviation of i data segments among the expression data sequence d.
According to formula (5) and formula (6), each data in each segmentation of data sequence d are judged, if x
IjSatisfy formula (5) or formula (6), then be used as for the first time exceptional value and search the exceptional value that obtains, the remainder data except that exceptional value for the first time exceptional value search the normal value that obtains, they are labelled in data sequence d respectively.
Step 6: merge for the first time segmentation: the section of presorting that obtains according to step 4; The detection data that step 2 obtains are carried out segmentation; After getting rid of the exceptional value that obtains in the step 5; Based on two sample Andrei Kolmogorovs-Si Moluofu Kolmogorov-Smirnov check, the section of presorting that step 4 is obtained merges, and record merges the pile No. information of each waypoint of back.
In step 6; The check of two sample Andrei Kolmogorovs-Si Moluofu Kolmogorov-Smirnov can be judged the whether same distribution of subordinate (being not necessarily normal distribution) of two groups of samples; Adopt this control device, can the adjacent segment that the data sequence has a same distribution tentatively be merged.
The merging principle flow chart of step 6 is as shown in Figure 3, merges principle and is described below:
At first, make i=1, set up an interim empty array TA, the m group data sequence that the normal value by among the data sequence d that is obtained by step 5 is formed, m is for presorting hop count, establishes this m and organizes data sequence and be respectively X
1, X
2.X
M-1, X
mThe m that also can obtain being made up of total data among the data sequence d organizes data sequence, and m establishes this m group data sequence and is respectively Z for presorting hop count
1, Z
2.Z
M-1, Z
m, Z wherein
iShould be by X
iThe exceptional value that reaches in the i section that is obtained by step 5 is formed.Carry out following operation:
(a) the interim array TA of initialization is just with the whole element zero clearings among the array TA;
(b) with i data sequence X
iInsert interim array TA;
(c) to array TA and i+1 data sequence X
I+1Carrying out significance is two sample Andrei Kolmogorovs-Si Moluofu Kolmogorov-Smirnov check, just (TA, X of the H=ks2test among Fig. 3 of p
I+1, p), if two samples pass through check, then H=0 thinks that hypothesis is accepted, and shows X
I+1Belong to same distribution with TA; If two samples are not through check, then H=1 thinks that hypothesis is rejected, and shows X
I+1Do not belong to same distribution with TA;
(d) as if H=0, then the value with i adds 1; If this moment, i was less than m, then turn back to b; If this moment, i equaled m, then segmentation finishes, the pile No. information that arrangement is returned;
(e) if H=1 then writes down i data sequence Z
iIn, the pairing pile No. Ki of last data; Make i equal i+1, if i then turns back to (a) less than m at this moment; Otherwise segmentation finishes, the pile No. information that arrangement is returned.
The significance p that mentions in the step 6 is a probable value, and it has determined the severe of two sample Andrei Kolmogorovs-Si Moluofu Kolmogorov-Smirnov check.In general, p is big more, and then check is harsher, and promptly hypothesis is rejected more easily.For common practical applications, the p suggestion gets 0.05.
According to the pile No. information of returning, can carry out segmentation once more to the data sequence d that obtains in the step 2 with the pile No. that obtains as waypoint, the segmentation that at this moment obtains is exactly through the segmentation after merging for the first time.
Owing to only use merging method based on the check of two sample Andrei Kolmogorovs-Si Moluofu Kolmogorov-Smirnov, distinguished owing to distribute different though can occur adjacent two sections, its average or intermediate value do not have the situation of significant difference.Yet, in practical engineering application, to have only when significant difference appears in the typical value of two adjacent segment, its design scheme just can change.In other words, if indifference between the typical value of adjacent segment can be divided into them in the same paragraph and go.So, need adopt rank test here, i.e. the method for median test, the adjacent segment that each other average or intermediate value are not had significant difference is dropped into row and is merged.
Step 7: search for the second time exceptional value: set by step 6 obtain waypoint pile No. information, the detection data that step 2 obtains are carried out segmentation, exceptional value is carried out in each segmentation searches, and in detecting data mark in addition.
In this step 7, the process of searching exceptional value is: be inversely proportional to if detect the size and the road conditions level of index, then will this section in all detection data that add the twice standard deviation more than or equal to this section average as exceptional value, record in addition; Be directly proportional with the road conditions level if detect the size of index, then will this section in all detection data that deduct the twice standard deviation smaller or equal to this section average as exceptional value, record in addition.
Before the segmentation that step 6 is obtained merges once more, be combined the interference of operation for preventing abnormal data, need carry out exceptional value to the data sequence d that step 2 obtains and search.The difference of step 7 and step 5 is, the roadway segment in the step 5 is by obtaining in the step 4, and the roadway segment in the step 7 is obtained by step 6.
6 obtain the data sequence d segmentation that pile No. information obtains step 2 set by step; Carrying out the second time of exceptional value searches: be inversely proportional to if detect the size and the road conditions level of index; All detection data that add the twice standard deviation more than or equal to this section average are as exceptional value in then will this section; Be shown in the formula (7), and record in addition; Be directly proportional with the road conditions level if detect the size of index, then will this section in all detection data that deduct the twice standard deviation smaller or equal to this section average as exceptional value, promptly shown in the formula (8), and record in addition.
Formula (7)
Formula (8)
In formula (7) and formula (8):
x
IjJ data among the expression data sequence d in the i segmentation;
The average of i data segments among
expression data sequence d;
S
iThe standard deviation of i data segments among the expression data sequence d.
According to top method, each data in each segmentation of data sequence d are judged, if x
IjSatisfy formula (7) or formula (8), then be used as for the second time exceptional value and search the exceptional value that obtains, the remainder data except that exceptional value for the second time exceptional value search the normal value that obtains, they are labelled in data sequence d respectively.
Why step 7 is necessary, and this is because for the first time exceptional value is searched the segmentation that the exceptional value that obtains just obtains to step 4, and for the second time exceptional value to search the exceptional value that obtains be the segmentation that obtains to step 6.This means that also for same data sequence d, as long as step 4 is different with the segmentation that step 6 obtains, the exceptional value of carrying out so obtaining after the exceptional value search operation just has the possibility that changes.
Step 8: merge for the second time segmentation: the pile No. information of 6 waypoints that obtain set by step; Detection data to step 2 obtains are carried out segmentation, after the exceptional value that obtains in the eliminating step 7, based on rank test; The segmentation that step 7 is obtained merges, and record merges the pile No. information of each waypoint of back.
In step 8; The value that each data can be effectively distinguished in rank test has two groups of samples of notable difference; Compare with two sample Andrei Kolmogorovs-Si Moluofu Kolmogorov-Smirnov check, whether it more pays close attention to two groups of differences between the sample remarkable.According to this character, can the segmentation that step 6 obtains further be merged.
Merging principle flow chart in the step 8 is as shown in Figure 4, merges principle and is described below:
At first, make i=1, set up an interim empty array TA.The r that step 7 obtains being made up of the normal value among the data sequence d organizes data sequence, and r is the segments that step 6 obtains, and establishes this r group data sequence and is respectively Y
1, Y
2.Y
R-1, Y
rThe r that also can obtain being made up of total data among the data sequence d organizes data sequence, and r establishes this r group data sequence and is respectively W for presorting hop count
1, W
2.W
R-1, W
r, W wherein
iShould be by Y
iThe exceptional value that reaches in the i section that is obtained by step 7 is formed.Carry out following operation:
(a) the interim array TA of initialization is just with the whole element zero clearings among the array TA;
(b) with i data sequence Y
iInsert interim array TA;
(c) to array TA and i+1 data sequence Y
I+1Carry out the rank test that significance is u, just (TA, the Y of the H=ranksum among Fig. 4
I+1, u), if two samples pass through check, then H=0 thinks that hypothesis is accepted, and shows Y
I+1And the difference between the TA is not remarkable; If two samples are not through check, then H=1 thinks that hypothesis is rejected, and shows Y
I+1And there were significant differences between the TA;
(d) as if H=0, then the value with i adds 1; If this moment, i was less than r, then turn back to (b); If this moment, i equaled r, then segmentation finishes, the pile No. information that arrangement is returned;
(e) if H=1 then writes down i data sequence W
iIn, the pairing pile No. k of last data
iMake i equal i+1, if i then turns back to (a) less than r at this moment; Otherwise segmentation finishes, the pile No. information that arrangement is returned.
The significance u that mentions in the step 8 is a probable value, and it has determined the severe of rank test.In general, u is big more, and then check is harsher, and promptly hypothesis is rejected more easily.For common practical applications, the u suggestion gets 0.05.
According to the pile No. information that step 8 EO is returned, can carry out segmentation once more to the data sequence d that obtains in the step 2 with the pile No. information that obtains as waypoint, the segmentation that at this moment obtains is exactly the segmentation after merging for the second time.
Step 9: search the Local treatment point: the 8 pile No. information that obtain set by step, the detection data that step 2 obtains are carried out segmentation, search the Local treatment point.
The segmentation that obtains through step 8 is the final segmentation of the detection data sequence d that step 2 obtains.Local treatment point also should be searched in final segmentation.In this step 9; The process of searching Local treatment point is: be inversely proportional to if detect the size and the road conditions level of index; All detection data that add the twice standard deviation more than or equal to this section average are as the Local treatment point in then will this section; Promptly adopt average to add the twice standard deviation as the threshold of judging Local treatment point, shown in (9), and record in addition; If detecting the size of index is directly proportional with the road conditions level; All detection data that deduct the twice standard deviation smaller or equal to this section average are as the Local treatment point in then will this section; Promptly adopt average to deduct the twice standard deviation as the threshold of judging Local treatment point, shown in (10), and record in addition.
Formula (9)
Formula (10)
In formula (9) and formula (10):
x
IjJ data among the expression data sequence d in the i segmentation;
The average of i data segments among
expression data sequence d;
S
iThe standard deviation of i data segments among the expression data sequence d.
If x
IjSatisfy formula (9) or formula (10), then being used as needs the Local treatment point, and it is labelled in data sequence d.
Step 10: calculate every kind of typical value that detects index in each segmentation: the 8 pile No. information that obtain set by step; The detection data that step 2 obtains are carried out segmentation; And get rid of the Local treatment point of searching in the step 9, calculate every kind of typical value that detects index in each segmentation.
Can obtain by the normal value among the data sequence d by step 9, just get rid of the data sequence behind the Local treatment point, the t group data sequence of composition, t is the segments that step 9 obtains, and establishes this t group data sequence and is respectively N
1, N
2.N
T-1, N
t, the t that also can obtain being made up of total data among the data sequence d organizes data sequence, and t establishes this t group data sequence and is respectively M for presorting hop count
1, M
2.M
T-1, M
t, M wherein
iShould be by N
iThe Local treatment point that reaches in the i section that is obtained by step 9 is formed.
In this step 10; Calculating every kind of detection index in the process of the typical value of each segmentation is: be inversely proportional to if detect the size and the road conditions level of index; Then the average with the detection data of having got rid of Local treatment point in this segmentation adds that one times of standard deviation is as typical value, shown in (11); Be directly proportional with the road conditions level if detect the size of index, then the average with the detection data of having got rid of Local treatment point in this segmentation deducts one times of standard deviation as typical value, shown in (12).
Formula (11)
Formula (12)
In formula (11) and formula (12):
The average of having got rid of the detection data of Local treatment point in
expression i section;
S
iRepresent to have got rid of in the i section standard deviation of the detection data of Local treatment point;
RV
iRepresent the typical value of certain index in the i section.
Adopt normal value to carry out the typical value measuring and calculating, can avoid the interference of abnormal data to data overall distribution situation, because got rid of after the exceptional value, the data in a certain segmentation are smoother generally, the situation that can not occur suddenling change.Adopting the method measuring and calculating typical value of one times of standard deviation of average plus-minus, is to have taken all factors into consideration design, and the feasibility of construction draws.