TRB Presentation

8/13/2019 TRB Presentation

1/27

EVALUATION OF WIDELY USED HYDROPLANING

RISK PREDICTION METHODS USING FLORIDASPAST CRASH DATA

Presented by:

Waruna Jayasooriya, M.Sc.

Ph.D candidate

Department of Civil and Environmental

Engineering

University of South Florida

Manjriker Gunaratne, Ph.D., P.E.

Professor and Chairman

Department of Civil and Environmental

Engineering

University of South Florida

1


2/27

Introduction to Hydroplaning

When a tire moving at a certain speed, layer of water

builds between the tire and the road surface

loss of traction and preventing the vehicle from

responding to control inputs

i.e. Steering, braking or accelerating

2


3/27

Introduction cont...

Viscous hydroplaning (Sliding)

Dynamic hydroplaning

3


4/27

Objectives

Develop a methodology to capturer the dynamic

hydroplaning crashes

Estimate the accuracy and reliability of most profound

hydroplaning prediction models

Estimates the accuracy of all cross combination

Develop an application to estimate all the combination

4


5/27

Stages of hydroplaning speed prediction

5


6/27

Stages of hydroplaning speed prediction

cont...

6


7/27

Literature review

Equatio

n

Number

SourceModel

StructureEquation Form Variables Applicability Limitations

4.1

British - Road

Research

Laboratory (4)

Empirical

Mean texture depth (MTD) was

not considered.

4.2Empirical form of

PAVDRN (8) Empirical None

4.3Gallaway (2)

(TxDOT method)Analystical None

4.4NZ modified

Equation (3)Empirical None

4.5Analytical form of

PAVDRN (8,10)Analytical None

MTDS

nLIt

6.0

5.01.36

7


8/27

Equation

Number Source

Model

Structure Equation Form Variables Applicability Limitations

5.1 NASA (11) EmpiricalAverage water film thickness is limited to

7.62 mm.

5.2 Ivey, et al. (12) Empirical None

5.3 NASA (11,13) EmpiricalAverage water film thickness is limited to

7.62 mm.

5.4Agrawal and Henry

(14,15)Empirical

Maximum water film thickness of 2.4 mm

5.5 Wambold et al. (16) Empirical

5.6 Horne, et al. (13) EmpiricalApplicable for the truck tires with a

fixed water film thickness.

5.7 PAVDRN (10,15) Analytical

Five pavement sections can be

analyzed by using PAVDRN: (a)

tangent section, (b) horizontal curve,

(c) transition section, (d) vertical crest

curve, and (e) vertical sag curve.

5.8Gallaway (2)

(TxDOT method)Analytical

Limited to vehicle speeds of less than

55 mph, 10% SD used as an indicator

of hydroplaning.

Tread depth of 2/32 inches used in

design.

5.9

USF Gunaratne et al.

(7) simplified form

based on Ong and Fwa

(6)

Empirical/

Finite

Element

Applicable to the light vehicles that

employ tires that are compatible with

locked-wheel tester tires.8


9/27

Methodology

1. Development of hydroplaning crash database

a) Florida Department of Transportation (FDOT)

databases

b) Weather information

c) Field observations

2. Analyze each hydroplaning crash with the

existing models and its combination

3. Evaluate the sensitivity and prediction accuracy

of the model combinations9


10/27

Databases (FDOT)

Crash Analysis and Reporting System (CARS) database:

Pavement Condition Survey (PCS) database:

Geographical Information System (GIS) database:

Vehicle, Passenger and Driver (VPD) Information database:

Police long-form (reports) database:

As-built plans

10


11/27

11

CARSDatabase

Police

Long-form

Weather Sta.

Database

GISDatabase

As-built

Plans

VPDDatabase

PCSDatabase

1-Interstate

Network

5-Estimation of

WFT, Use of

PAVDRN

equation

2-Screening Parameters

Weather Condition: Rainy

Surface Condition: Wet, Slippery

Max.Posted Speed: >= 40mphLighting Condition

Drug or Alcohol ExcludedSite Loc. Ramps Excluded

6-Compare lagging

distance with SSD

Wider roadway

section & HOV lanedesignation

Identification of Inward

sloped sections

4-Identification of closest

weather stations

3-Location details

Side of the roadPavement distresses Vehicle information

Driver, Passenger details

7-Classification of Hyd.

Crash type

Hydroplaning Crash

Database 2006 - 2011

Rainfall Intensity,Visibility, Wind

Dynamic hyd.

Viscous hyd.

Vehicle damage,Tire condition,

Vehicle movement

damage severity

IRI, Rut number, Ride

rating, Crack rating

WFT

Flow Chart Key

Database

Screening Criteria

Outcome variable

Combined Database


12/27


13/27

Separation of Hydroplaning Crash Type

(Manual)

13

(a) Dynamic hydroplaning event (b) Viscous hydroplaning event


14/27

Assignment of Rainfall Intensity for Each

Crash Weather data interpolation

Inverse squared distance interpolation

Thiessen polygon Method

14


15/27

Sample of Hydroplaning Database

15

Crash ID CrashRate Traffic Densit rate for 10ACCISEV HYD TYPE Travel LanACCLANE Num of LaTravel SpeMAXSPEE CS Trav la Rainfall InWFT (mm Visibility ( Pav Mat

71094140 0.00442982 225.7428 44 2 1 2 S 2 60 70 2.0% 0 2.5 #DIV/0! OGFC

85275710 0.004016064 249 40 1 1 2 S 2 65 70 2.0% 0.34 2.95 0.34 OGFC

90465020 0.007297752 137.0285 73 1 1 1 S 3 50 70 2.0% 0.07 2.65 0.07 OGFC

92895220 0.002024038 494.06175 20 1 1 3 S 4 30 45 3.0% 1.05 1.6 0.420476 OGFC

95164550 0.001498641 667.271 15 1 1 2 S 2 70 70 2.0% 0.01 2.6 0.01 OGFC

143399430 0.000873675 1144.591 9 3 0 1 1 3 45 55 2.0% 0.4 0.7 0.225595 OGFC

603599490 0.000512509 1951.1865 5 3 0 4 4 5 30 65 3.0% 0.15 0.7 0.083824 OGFC

700168610 0.001268081 788.593 13 1 1 2 S 2 55 70 2.0% 0.21 2.85 0.21 OGFC700625690 0.000735328 1359.936963 7 1 1 1 M 3 30 45 2.0% 0 0 #DIV/0! OGFC

700786450 0.002607905 383.4495 26 1 1 1 M 3 50 55 2.0% 0 0 #DIV/0! PCC

701512880 0.00697715 143.325 70 1 1 3 2 4 40 50 3.0% 0 0 #DIV/0! PCC

704357700 0.000677199 1476.6705 7 1 0 2 2 4 60 65 2.0% 0.18 0.6 0.143381 OGFC

704556370 0.001511001 661.81275 15 1 1 2 2 3 70 70 2.0% 0.04 2.65 0.021543 OGFC

704970550 0.002305248 433.7928 23 3 1 5 S 5 65 65 3.5% 0.08 0.6 0.056359 OGFC

704972150 0.005057694 197.7185595 51 1 0 2 2 4 65 55 2.0% 0.08 0.4 0.037447 OGFC

707872010 0.000436085 2293.13 4 2 0 1 1 3 15 55 2.0% 0 0 #DIV/0! OGFC

707882860 0.000779837 1282.32 8 1 1 2 S 3 65 55 2.0% 0.48 1 0.392594 OGFC

707899220 0.000605179 1652.4025 6 3 0 1 1 3 40 55 2.0% 0.24 2.75 0.24 OGFC

709963360 0.002688129 372.006 27 2 1 2 S 2 70 70 2.0% 0.27 2.9 0.233073 OGFC

710451930 0.001465532 682.346 15 1 1 1 1 3 70 70 2.0% 0.21 2.75 0.168089 OGFC

711178700 0.002385804 419.146 24 1 1 1 S 3 55 70 2.0% 0 2.5 #DIV/0! OGFC

711184800 0.001269218 787.887 13 2 1 2 S 3 70 70 2.0% 0.81 3.15 0.652484 OGFC


16/27

Content of Hydroplaning Database

Hydroplaning type

Posted speed/ Travel speed

Rainfall intensity Number of lanes/Travel lane

Pavement material

Dense graded asphalt concrete (DGAC)

Portland Cement Concrete (PCC) Open graded friction course (OGFC)

Pavement distresses (Roughness, IRI)16


17/27

Sensitivity Analysis of Model combinations

The specific values of the variables closest to their averages in

the entire dynamic hydroplaning crash database were assumed

as the base values listed below.

Rainfall intensity, I = 2 inches/hr.

Pavement surfacing type = DGAC

Travel lane (from median of the road) = 2

Longitudinal slope, s = 0.5 %

17


18/27

Summary Results of Sensitivity Analysis

Model Combination1 2 3 4 5 6 7 8 9 10 11 12

Stage 1

(WFT prediction)Gallaway (Eqn. 4.3) British-RRL (Eqn. 4.1) NZ modified (Eqn. 4.4) PAVDRN (Eqn. 4.5)

PAVDRN

(Eqn. 5.7)

Gallaway

(Eqn. 5.8)

USF

(Eqn. 5.9)

PAVDRN

(Eqn. 5.7)

Gallaway

(Eqn. 5.8)

USF

(Eqn. 5.9)

PAVDRN

(Eqn. 5.7)

Gallaway

(Eqn. 5.8)

USF

(Eqn. 5.9)

PAVDRN

(Eqn. 5.7)

Gallaway

(Eqn. 5.8)

USF

(Eqn. 5.9)

Rainfall

Intensity(in/h) 0.1 (-95%) 41.9%* 6.6% 16.6% 51.2% 6.2% 13.0% 29.1% 3.8% 7.9% 50.4% 6.2% 12.8%

0.5(-75%) 11.3% 3.0% 8.5% 20.5% 2.7% 5.7% 12.6% 1.7% 3.6% 20.3% 2.7% 5.7%

1(-50%) 0.4% 1.5% 5.2% 9.7% 1.3% 2.8% 6.1% 0.9% 1.8% 9.6% 1.3% 2.8%

3(50%) -1.8% -0.8% -1.1% -5.2% -0.8% -1.6% -3.4% -0.5% -1.0% -5.2% -0.8% -1.6%

4(100%) -3.1% -1.4% -1.8% -8.8% -1.3% -2.7% -5.7% -0.8% -1.8% -8.7% -1.3% -2.7%

Pavement

Type PCC - - - - - - - - - 20.5% 2.7% 5.7%

OGFC 6.9% -0.2% -0.3% 0.5% 0.1% 7.6% 0.0% 0.0% 7.4% 5.0% 0.7% 9.0%

TravelLane 1 8.8% 2.6% 7.7% 20.5% 2.7% 5.7% 19.7% 2.7% 5.5% 18.1% 2.4% 5.1%

3 -2.8% -1.3% -1.7% -9.6% -1.4% -3.0% -9.1% -1.4% -2.8% -8.3% -1.2% -2.6%

4 -4.7% -2.3% -2.9% -12.3% -2.4% -7.8% -15.2% -2.3% -4.8% -11.0% -2.1% -7.3%

5 -5.5% -3.0% -3.7% -13.7% -3.2% -8.7% -16.6% -3.1% -9.1% -12.3% -2.8% -8.1%

18


19/27

Model prediction accuracy

19


20/27

Comparison of model combinations with

accuracy and sensitivity model combination 1

shows an average

sensitivity with the highest

prediction accuracy. model combinations 4 and

10 also show high

sensitivity to rainfall

intensity and travel lane

while exhibiting a lowprediction accuracy.

20


21/27

Descriptive Statistics of Error Distribution

Model Combination 1 2 3 4 5 6 7 8 9 10 11 12

Mean (mph) 4.34 6.99 14.23 16.03 8.97 20.88 12.00 8.54 19.71 12.39 8.63 19.88

Median (mph) 0.00 4.32 13.29 13.48 6.98 19.82 10.63 6.33 18.54 10.69 6.53 18.47

Mode (mph) 0.00 0.00 0.00 0.00 0.00 20.48 0.00 0.00 16.62 0.00 0.00 11.35

Std. Dev. 7.16 7.55 9.33 14.20 8.11 9.29 10.90 7.99 8.85 11.12 8.06 8.79

5% percentile (mph) 0.00 0.00 0.00 0.00 0.00 7.44 0.00 0.00 7.48 0.00 0.00 7.74

95% percentile (mph) 20.56 22.15 31.04 42.99 24.65 37.49 32.80 23.93 35.96 33.71 24.12 36.03

Range (mph) 42.65 32.61 44.30 69.22 35.42 51.79 51.32 34.73 49.22 58.50 34.91 48.30

ICC* 0.733 0.114 0.408 0.427 0.133 0.260 0.518 0.082 0.245 0.420 0.040 0.184

*Intraclass Correlation Coefficient

21


22/27

Error distributions in each model

combination

Based on the ICC and the

model prediction accuracy

the model combinations canbe ordered as 1,4,10 and 3

in terms of accuracy.

fFrequency distribution of

hydroplaning crashes,Error in speed prediction with

respect to predicted threshold

hydroplaning speed22


23/27

Development an application on analysis of

minimum Hs

Input parameters

Pavement properties

Geometrical properties Vehicle characteristics

23


24/27

Conclusions

Floridas CARS database was fortified with using (1) PCS (2)

CARS (3) GIS (4) VPD databases and (5) Police long-forms.

Two stage process of prediction generated twelve possible

model combinations.

Sensitivity analysis performed on the threshold hydroplaning

speed on the key parameters.

Reliability of each model combination was evaluated by using

intra-class correlation coefficients of the resulting error

distributions.

24


25/27

Conclusions cont..

Prediction accuracy of model combinations was alsoevaluated

A computer program (HP) was developed to compute thepossible hydroplaning speed for each model combination

Model combinations were more accurate in predictinghydroplaning risk.

Provide tools to evaluate the hydroplaning risk on a givenhighway under typical adverse weather conditions withimproved reliability.

25


26/27

Thank you

26


27/27

Acknowledgement

This research project was funded by Florida department of

transportation (FDOT)

27

Documents

TRB Presentation