Vous êtes sur la page 1sur 40

COMPARISON BETWEEN COMPUTER

SOFTWARE FOR ESTIAMTING DELAY AT


SIGNALIZED INTERSECTIONS IN JORDAN

By: Madhar M. Ta’amneh


Under the Supervision of
Dr. Bashar AL-Omari

Publication: Bashar Al-Omari & Madhar Ta'amneh "Validating HCS and


SIDRA Software for Estimating Delay at Signalized Intersections in
Jordan" Jordan Journal of Civil Engineering, Volume 1, No. 4, October
2007, pp. 375-393.
OBJECTIVES:
1- To make a comparison between HCS 2000, SIDRA
and CORSIM computer software for estimating delay at
signalized intersections in Jordan based on field delay
data.

2- To determine the most suitable software, which


represents the Jordanian conditions in estimating the
signalized intersection delay.

3- To find if these models can be calibrated to better


represent the Jordanian conditions.
METHODOLOGY
1- Site Selection
• A number of locations were visited to find
which ones fulfill the following
requirements:
a- Under-saturation Conditions
b- Low pedestrian volumes
c- No interruptions to traffic
• A total of 18 approaches at 5 signalized
intersection in Irbid City were selected.
2. Data Collection
Table 4.2. Free-flow Speed (85 Percentile)
Free-flow speed (Km/h)
Intersection
Name NB SB EB WB

Al-Karamah 71 66 63 54

Al-Naseem 60 53 40 52
Al-
Darawsheh N/A* 63 64 69
East
Entrance 61 59 66 65

East Theater N/A* 64 47 61


c.1.2. 15 Minutes Time Interval.
INTERSECTION CONTROL DELAY WORKSHEET
General Information Site Information
Analyst Madhar Ta'amneh Intersection Al-Karamah Intersection
Agency of Company JUST Area Type CBD Other
Date Perfomed Jurisdiction Irbid
Analysis Time Period 5 Minte Analysis Year 2002
Input Initial Parameters
Number of lanes, N 3 Total vehicle arriving 176
Free-flow speed, FFS (mi/h) 39.14 Stpopped-vehicle count 165
Survey count interval, Is (s) 15 Cycle length, C (s) 104
Input Field Data
Number of vehicle in queue
Count Interval
Clock Time Cycle Number 1 2 3 4 5 6 7 8 9 10
10:45 1 1 2 8 13 17 20 10
2 0 4 9 14 17 17 11 0
3 2 4 6 9 14 6
4 3 5 10 15 17 19 8
5 4 8 11 16 19 20 4
6 3 6 15 19 21 19
7 2 3 6 16 17 12 0
8 1 8 12 19 22 22 3
11:00 9 0 5 7 12 15 19 3 0

Total 16 45 84 133 159 154 39 0 0 0


Computation
Total vehicle in queue, Sum(Viq) 630 No. of cycles surveyed, Nc 8.65
Time-in-queue per vehicle, dvq 48.32 Fraction of vehicle stopping, FVS 0.94
No.of vehicle stopping per lane each cycle 6.36 Accel/Decel correction delay 6.56
Accel/Decel correction factor, CF 7 Control delay/vehicle 54.89
3. Software Validation

• HCS Validation.

•SIDRA Validation.

•CORSIM Validation.
4. Software Calibration

• Saturation Flow Rate.

•Bus Volumes/Dimensions.

5. Selection of the Best Software for


Jordanian Conditions.
6. Sampling Time Interval

* Usually 1,5, 10, 15, 30, 60 minutes


sampling time intervals are used.
* 15 minutes time interval will be used in this
study.
* Comparison between 5 & 15 minutes time
intervals will be conducted for HCS 2000
software to show the effect of sampling
time on the analysis.
DATA COLLECTION AND
REDUCTION

* A total of 54 hours of data were collected from


18 approach at 5 signalized intersection using
video camera.
1. Data Collection

• Traffic Volume
• Intersection Geometry
• Vehicle Queue Length
• Distribution of Arrival on all Approaches
• The Weather of the Day of Data Collection
• Pavement Condition.
• Pavement Type.
• Vehicle Type
• Pedestrian
• Free-flow Speed
• Basic Saturation Flow Rate
• Signal Cycle Time
• Parking activity
2. Data Reduction
* The Control Delay was Measured according to
the following (HCM Standard Procedure):
1. Queue length was determined each 15 sec.
2. Arriving and Stopped Vehicles were measured
3. Free-flow speed (85 percentile) was measured
4. Passenger Car Equivalent (PCE) is 2.0
5. Control delay measured at 5 and 15 minutes
6. Acceleration and Deceleration delays
7. The above data were interred to a special
worksheets in order to calculate control delay
incurred by vehicles.

• SPSS software was used for statistical analysis.


• The data was used to predict the delay using
HCS 2000, SIDRA and CORSIM.
1. Validation of HCS 2000
* (lane width, number of approaching lanes,
number of exit lanes, island width,
pedestrian volume, bus parking, signal
timing and traffic volumes) data were
entered to the software.

* A total of 54 hours of field data were


collected, producing 648 observations of 5
minutes time interval and 216 observations
of 15 minutes time interval.
Fifteen Minutes Time Interval
A- Control Delay Estimated Based on the Default
Parameters:
• At low delay ranges (up to 20 seconds) HCS 2000
over-estimates the control delay.
• For higher delay ranges, HCS 2000 has a mix of
under-estimates and over-estimates points with some
sever over-estimated points resulting from HCS 2000.
• This is confirmed by Figure 5.2, which show for
degree of saturation of lower then 0.8 HCS 2000 gives
a relatively good estimation.
• Paired t-test results:
Table 5.2. Comparison of Actual and HCS 2000 Control
Delay Prediction
Time
Interval N MEAN STDEV SE mean t-statistic p-value
15 216 -10.29 29.44 2 -5.135 0
• The hypothesis that the deviations are equal to
zero is rejected for 15 time intervals at 95%
confidence.
• HCS 2000 is not a good predictor of the actual
delay, especially at high degree of saturation.
B- Control Delay Estimated Based on 50% of the Bus
Volumes:

• At low delay ranges (up to 20 seconds) HCS 2000 still


over-estimates the control delay.

• For higher delay ranges, HCS 2000 has a mix of under-


estimates and over-estimates points with some more
over-estimated points resulting from HCS 2000.

• Over-estimation problem has now decreased due to bus


volumes reduction.
• Paired t-test Results:

Table 5.4. Comparison of Actual and HCS 2000 Control Delay


Prediction (50% B)
Time
Interval N MEAN STDEV SE mean t-statistic p-value
15 216 -3.27 15.552 1.058 -3.091 0.002

• The hypothesis that the deviations are equal to zero is


rejected for 15 time intervals at 95% confidence

• Delay prediction was enhanced due to bus volumes


reduction.
C- Control Delay Estimation Based on the 2200 pcphgpl
(Basic Saturation Flow Rate).
• Default value of the basic saturation flow (BSFR) rate
used by HCS 2000 was 1900 pcphgpl.
• Calibrated BSFR was 2200 pcphgpl.
• At low delay ranges (up to 20 seconds) HCS 2000
over-estimates the control delay.
• For higher delay ranges, HCS 2000 has a mix of
under-estimates and over-estimates points with some
more over-estimated points resulting from HCS 2000.
• Over-estimation problem has now decreased due to
new BSFR.
• The paired t-test was also used :

Table 5.6. Comparison of Actual and HCS 2000 Control Delay


Time Prediction (2200 pcphgpl)
Interval N MEAN STDEV SE mean t-statistic p-value
15 216 -1.78 15.37 1.05 -1.7 0.091

• The hypothesis that the deviations are equal to zero is


accepted at 95% confidence
• Delay prediction was enhanced due to BSFR increase
• Increase BSFR will increase the capacity. So (v/c)
will be reduced and hence the control delay will be
reduced.
D- Control Delay Estimation Based on Both 50% of
Bus Volumes and 2200 pcphgpl(BSFR).

• At low delay ranges (up to 20 seconds) HCS 2000


slightly over-estimates the control delay.

• For higher delay ranges, HCS has a mix of uner-


estimated and over-estimated points.

• Data points in this case were almost distributed


evenly around the 45° line.
• The paired t-test was also used:
Table 5.6. Comparison of Actual and HCS 2000 Control Delay
Prediction (2200 pcphgpl)
Time
Interval N MEAN STDEV SE mean t-statistic p-value
15 216 1.775 8.465 0.576 3.082 0.002

• The hypothesis that the deviations are equal to zero is


accepted at 95% confidence.

• Delay prediction was enhanced due to BSFR increase


and bus volumes reduction.
* Summary
Table 5.9. Regression Results of Actual and HCS 2000 Delay Time
Time Delay Intercept Average of
Interval based on Bo slope B1 R² % SEE Deviations
15 Default
Minutes values 28.94 0.15 21.6 9.43 -10.29
15
Minutes 50% busses 22.846 0.333 37.9 8.392 -3.27
15 2200
Minutes pcphgpl 23.503 0.333 34.2 8.623 -1.78

15 50%& 2200
Minutes pcphgpl 14.288 0.633 55 7.144 1.78
2. SIDRA Validation
A- Control Delay Estimation Based on the Default
Parameter:
• (up to 10 seconds), SIDRA slightly over-estimates the
control delay.
• In the (10 - 25) sec/veh ranges, SIDRA has a good
estimation of the control delay.
• In the (25 - 60) sec/veh, SIDRA has a mix of under-
estimated and over-estimated points.
• This is confirmed by Figure 5.8, which show for degree
of saturation of lower then 0.8 HCS 2000 gives a
relatively good estimation.
** Summary
Summary

Table 5.21. Regression Results of Actual and SIDRA Delay Time


Time Delay Intercept Average of
Interval based on Bo slope B1 R² % SEE Deviations
15 Default
Minutes values 22.891 0.346 33.6 8.678 -1.697
15
Minutes 50% busses 11.58 0.696 57.2 7 0.944
15 2200
Minutes pcphgpl 13.771 0.635 52.9 7.306 1.077

15 50%& 2200
Minutes pcphgpl 7.513 0.847 64.2 6.359 2.377
CORSIM Validation
A- Control Delay Estimation Based on the Default
Parameter:

• (up to 10 seconds), CORSIM has a good estimation of


the control delay.

• In the (10 - 60) sec/veh ranges, CORSIM under-


estimate the control delay with some scattering at large
delay values.
* Summary
Table 5.30. Regression Results of Actual and CORSIM Delay Time
Time Delay Intercept Average of
Interval based on Bo slope B1 R² % SEE Deviations
15 Default
Minutes values 21.17 0.51 38.1 8.381 7.05
15
Minutes 50% busses 15.17 0.732 50.5 7.5 7.6
15 2200
Minutes pcphgpl 13.8 0.805 53.4 7.27 8.43

15 50%& 2200
Minutes pcphgpl 9.9 0.949 61.3 6.62 8.48
Comparison Between 3 Software
Comparison Based on Both Buss Treatments
& 2200 pcphgpl (BSFR).

Traffic Intercept Slope Average of


Software Bo B1 R² % SEE Deviations
HCS 2000 14.29 0.633 55 7.14 1.775
SIDRA 7.51 0.847 64.3 6.36 2.377
CORSIM 9.9 0.949 61.3 6.62 8.48
• SIDRA has a larger R² and smaller SEE values. Thus it is
recommended to use for traffic analysis in Jordanian Traffic
Departments.
• The three software can be used after bring calibration for
saturation flow & bus volumes.
CONCLUSIONS
* HCS 2000 Software
1-Most of the time overestimate the control delay.
2- A lot of scattering around 45° line (R²=21.6%,
SEE=9.43).
3- 50% bus volume reduction produce (R²=37.9%,
SEE=8.39).
4- 2200 pcphgpl produce (R²=34.5%, SEE=8.62).
5- 2200 pcphgpl & 50% bus volume produce
(R²=55%, SEE=7.14).
* SIDRA Software
1- Predicted delay points almost distributed evenly
around 45° line at low delay ranges, but an over-
estimation occurs at high delay ranges (variation still
less than that for HCS 2000).
2- A lot of scattering around 45° line (R²=33.6%,
SEE=8.68).
3- 50% bus volume reduction produce (R²=57.2%,
SEE=7.02).
4- 2200 pcphgpl produce (R²=52.9%, SEE=7.31).
5- 2200 pcphgpl & 50% bus volume produce
(R²=64.3%, SEE=6.36).
* CORSIM Software

1- Most of the time under-estimate the control delays


with more scattering at high delay ranges.
2- Some scattering around 45° line at high delay
ranges (R²=38.1%, SEE=8.38).
3- Calibrating the bus lengths produce (R²=50.5%,
SEE=7.5).
4- 2200 pcphgpl produce (R²=53.5%, SEE=7.27).
5- 2200 pcphgpl & actual bus lengths produce
(R²=61.3%, SEE=6.62).
• SIDRA Software provides the best choice
for traffic analysis at Jordanian conditions
after calibrating saturation flow rate and bus
volumes. CORSIM software comes next
and then HCS 2000.

• 15 minutes time interval analysis gave


better results than the 5 minutes time
interval analysis.
RECOMMENDATIONS
1- Geometric design should be given more attention.
2- Re-timing the traffic signals.( for example, Irbid
East Entrance Intersection).
3- HCS2000, SIDRA, and CORSIM are not a good
predictors for delay at signalized intersections in
Jordan if the default parameters are used.
4- Studies are recommended to develop special and
comprehensive software for estimating delay at
signalized intersections in Jordan.
5- Studies are recommended on other software to
adjust variables that are related to driver behavior
and geometric parameters in Jordan.
6- Further research is recommended to find the
effect of the following factors on intersection
delay:
a- Pedestrian presence
b- Non-random arrivals
c- Vertical alignments
d- bus parking
e- PC parking
f- Weather and the streets surface conditions
250

200
HCS Delay (sec/veh)

150

100

50

0
0 50 100 150 200 250

Measured Delay (sec/veh)


Figure 5.1. HCS 2000 Vs. Field Control Delay Time

Table 5.1.Regression Results of Actual and HCS 2000 Delay Time


Time Interval Intercept Bo Slope B1 R² % SEE
15 Minutes 28.94 0.15 21.6 9.43
140

120

HCS Delay (sec/veh) at 0.5B


100

80

60

40

20

0
0 20 40 60 80 100 120 140

Measured Delay (sec/veh)

Figure 5.3. HCS2000 Vs. Field Control Delay Time


( 0.5B )

Table 5.3.Regression Results of Actual and HCS 2000 Delay Time


Time Interval Intercept Bo Slope B1 R² % SEE
15 Minutes 22.846 0.333 37.9 8.391
160

140

120
HCS Delay (sec/veh)
100

80

60

40

20

0
0 20 40 60 80 100 120 140 160
Measured Delay (sec/veh)

Figure 5.4. HCS 2000 Vs. Field Control Delay Time


( 2200 pcphgpl )

Table 5.5.Regression Results of Actual and HCS 2000 Delay Time


Time Interval Intercept Bo Slope B1 R² % SEE
15 Minutes 23.503 0.329 34.5 8.6203
90

80

HCS Delay (sec/veh) at 0.5B & 2200 pcphgpl


70

60

50

40

30

20

10

0
0 10 20 30 40 50 60 70 80 90

Measured Delay (sec/veh)


Figure 5.5. HCS 2000 Vs. Field Control Delay Time
( 0.5B & 2200 pcphgpl )
Table 5.7.Regression Results of Actual and HCS 2000 Delay Time
Time Interval Intercept Bo Slope B1 R² % SEE
15 Minutes 14.29 0.633 55 7.144
90

80

70

Estimated Control Delay (sec/veh)


60

50 HCS
SIDRA
CORSIM
40

30

20

10

0
0 10 20 30 40 50 60 70 80 90

Field Control Delay (sec/veh)

Figure 5.19: HCS 2000, SIDRA and CORSIM Vs.


Field Control Delay Time
( 0.5B & 2200 pcphgpl )

Vous aimerez peut-être aussi