Page 1 of 4
European Journal of Business &
Social Sciences
Available at https://ejbss.org/
ISSN: 2235-767X
Volume 07 Issue 05
May 2019
Available online:https://ejbss.org/ P a g e | 1060
Studies on Safety Performance of Twolane Rural Highways Under Mixed
Traffic
D. Anbiah, PG Student
Dept. of Civil Engineering
PRIST University
Thanjavur, India
Abstract—The traffic on the Indian highways is heterogeneous
in nature. Vehicles with widely varying static and dynamic
characteristics occupy the common carriageway space. The speed
differentials between the vehicles of the same class and different
classes significantly contribute to road crashes on the highways.
The operating speeds of the different vehicle types vary widely.
The design speed is different from the operating speeds
observed on the highways. Design of the road geometrics for
safety considering vehicles with widely varying static and
dynamic characteristics is a challenging task to the designers.
The present research work attempts to develop safety prediction
models for inter-city highways catering to heterogeneous traffic
considering the operating speeds of vehicles and relating it to
crashes duly considering the roadway geometric elements. A
typical four lane highway is considered in the research study.
The data on roadway geometrics at different sections along the
highway, the traffic details and other elements contributing to
the road crashes are collected along with the road accident data
for a period of four years. Statistical analysis was carried out to
develop safety prediction models. The models may be used in
assessment of safety performance of four lane rural highways
under heterogeneous traffic conditions and in conditions.
Index Terms: Traffic survey, Road Crash, Heterogeneous
traffic, Highway design, Regression Modelling
I. INTRODUCTION
According to World Health Organization (WHO 2013),
over 1.2 million person arekilled and 20 to 50 million person
are injured annually in road crashes worldwide. As far as
developing countries are concerned, 20 to 200 deaths occur per
10,000 motor vehicles, whereas only 2 to 5 deaths happen in
the case of developed countries. India is no exception with over
1.3 million people gets killed in over 4.89 million road crashes
every year. The increase in the number of vehicles has
however, far outweighed the projected capacities and adaptive
capabilities of the existing road infrastructure leading to
deterioration in traffic performance and increasing difficulties
in maintaining an acceptable and sustainable traffic-safety
standards.
Fatality rate, which is defined as the number of persons
killed per 100,000 inhabitants per year, is 16.6 in 2014 in the
country. Statistical data from India reveals that highways are
facing a disparate rate of crashes, i.e., almost 53.7% of road
crashes occurred on inter-city highways in the year 2014 when
compared to urban roads (Transport Research Wing (TRW),
Ministry of Road Transport & Highways, 2014). The
heterogeneous traffic under high-volume conditions, share the
available road space without sufficient safe lateral as well as
longitudinal clearances. The lane-less movement further adds
to the complexity of analyzing/modeling heterogeneous traffic.
This motivates the need for a level of safety analysis under
heterogeneous traffic conditions.
Road crashes are complex events and are influenced by
many factors such as road geometric design, traffic volume and
composition, speed, weather, motivation for travelling, drivers‘
physical and mental conditions etc. Highway geometry should
be designed for traffic safety and efficiency. Traffic-safety
analysis was initially recognized in 1960‘s. Later several
research works related to traffic safety analysis were performed
by different research groups in different parts of the world by
considering its seriousness to the public health, which resulted
in different ways for evaluating safety performance. Safety
Performance Function (SPF), is defined as the relationship
between crash and the factors influencing crashes. SPFs are
established in order to evaluate safety performance of highway
or roadway under study.
Fig. 1. Typical View of Road Accident in Two Lane Rural Roads
Page 2 of 4
European Journal of Business &
Social Sciences
Available at https://ejbss.org/
ISSN: 2235-767X
Volume 07 Issue 05
May 2019
Available online:https://ejbss.org/ P a g e | 1061
II. DATA DESCRIPTION
After reconnaissance survey, 85 km long four-lane divided
inter-city roadway in Tamil Nadu state, in India, was identified
as the study stretch. In order to develop models for evaluation
of safety performance, data related to road geometry, traffic,
volume, composition, speed etc and crash characteristics is
essential.
Road inventory surveys were carried out to collect the
roadway geometric data. The geometric data were collected
using Hawkeye, automated data collection equipment
developed by Australian Road Research Board and the outputs
are linked to both spatial (GPS) and linear references. The
equipment has three high-resolution digital cameras by which
digital images were collected with a 150 to 180 degree field of
view (centered on the travel lane) and the data were extracted
using Hawkeye processing toolkit by standard rating form.
The extracted data includes gradient, cross slope, horizontal
curve radius, vertical curve radius, pavement width, shoulder
width, access points, median opening and all particulars related
to geometry of a highway. In order to collect traffic
characteristics, video capturing technique was used and also
data were collected from official records. Traffic composition
data was extracted manually from videos and it comprises of
five major categories of vehicles from city journey vehicles
(two-wheelers, three-wheelers) to multi-axle trucks plying on
the roadway.
The presence of different categories of these motorised
vehicles operating at different speeds and varied physical
dimensions were affecting the average operating speed along
the roadway, thereby playing a vital role on the safety. Traffic
data collection includes composition of vehicles, operating
speed along the highway for different categories of vehicles,
and volume count. City journey vehicles like two-wheelers are
the more vulnerable road users among the vehicle category
plying on the study corridor..
III. CHEMICAL COMPOSITION OF ASPHALT
Unlike synthetic polymers, factors that govern the
viscoelastic behavior of asphalt binders are not well-defined,
which is mainly because of their unclear molecular structure.
Chemical composition of the asphalt binders affects their
durability, hardening and viscoelastic properties Asphalt is a
complex mixture of hydrocarbons and also contains small
amounts of sulfur, oxygen, nitrogen and trace quantities of
metals such as vanadium, iron, nickel, calcium and
magnesium.
Generally, asphalt binders contain two main phases: an oily
phase and a non-oily phase. Oily phase in asphalt is composed
of saturated hydrocarbons (paraffinic and maltenes), cyclic
products (naphthene and aromatics) and resins (polar
aromatics), while non-oily phase composed of asphaltenes,
carbenes, etc. To analyze and identify asphalt composition, the
socalled SARA method subdivides a large number of asphalt
molecules into a limited number of fractions based on the
affinity in their chemical composition. The SARA fractions are
defined as saturates (S), aromatics (A), resins (R) (which are
from the maltene fraction), and asphaltenes (A). The
representative structures of these four fractions. The
complexity, molecular weight, heteroatomic content, polarity,
and aromaticity increase in the order of S < A< R < A.
Asphaltenes are the most polar fraction of asphalt due to
the presence of functional groups such as carboxylic acids,
phenolics, amines, and thiols. These polar components play a
significant role in the adhesion of asphalt binder to the polar
surfaces of aggregates. Viscoelasticity of asphalt binders are
influenced by the ratio of asphaltenes to the other constituents
such as saturates, resins, and maltenes.
A.Internal Structure of Polymers
Polymers are very large molecules which are made of
hundreds or thousands of atoms. Polymers are formed by
successive linking of one or more types of small molecules
(i.e., monomers) into chain or network structures. The
fascinating area of polymers is too big to cover in a single
article. Thus, we restrict ourselves to the polymers commonly
used to modify asphalt binders and mixtures with an emphasis
on their ability to change the rheology, morphology, and
mechanical properties of binders. In general, the structure of a
polymer can be a straight chain or many variations linked and
cross-linked chains. Copolymers, which are composed of
several monomer units, are the most common asphalt modifiers
used in paving industry. Demonstrates some typical
architectures of copolymers. Copolymers are classified into
four types by monomer placement: random copolymers,
alternating copolymers, block copolymers and graft
copolymers. The architecture of these copolymers can be
demonstrated using two or more monomer units. Random
copolymers are identified by a random sequence arrangement
of two or more monomer units while alternating copolymers
are identified by the regular alternating of monomer units.
Block and graft copolymers are composed of several segments
and they differ in the intersegment linkage site. Block
copolymers are identified by their terminally connected
structures while graft copolymers are synthesized by
chemically linked pair of homo-polymers and have a comb
type structure. Block copolymers obtained by joining hard and
soft segments (usually in a three-block sequence of hard-soft- hard) is one of the favorable architectures used for polymer
modification of asphalt binders.
Page 3 of 4
European Journal of Business &
Social Sciences
Available at https://ejbss.org/
ISSN: 2235-767X
Volume 07 Issue 05
May 2019
Available online:https://ejbss.org/ P a g e | 1062
Fig. 2. Traffic Composition along the Study Stretch
B.Crash history
Crash history is the major input for the model development.
The data on crashes was collected from records for the study
period (2009 to 2012), which includes the number, type and
cause of crash, vehicles involved, severity, location and so on.
The dependence of the number of crashes at each location
to the operating speed and radius of horizontal curve. From the
figure it can be observed that the variation in the geometric
profile of highway under study has an influence on the crashes.
Fig. 3. Plot of crashes with operating speed and radius along the study stretch
The data collected for four years (2009-2012), were cleaned
for developing models to evaluate safety performance under
heterogeneous traffic condition for four-lane intercity
highways. The data were segmented based on fundamental
segmentation approaches i.e., length of roadway and annual
average daily traffic (AADT). In the present work,
segmentation based on fixed length of 500m length is adopted.
A total of 170 segmental data are used in the crash analysis. To
explore the relationship between the number of road crashes as
dependent variable and geometric design characteristics and
AADT (exposure variable) as independent variables, Poisson
and/or Negative Binomial (NB) regression and Zero-inflated
models were considered.
Goodness of fit of the developed models are evaluated
using likelihood ratio and Akaike‘s Information Criteria (AIC).
The model which has the minimum AIC value is
considered as the best model.
IV. RESULTS AND SUMMARY
A. Safety Performance Function
Poisson regression model was developed and it was found
that most of the variables identified for the model development
are statistically significant. But greater variability is observed
(i.e., 7.32), which clearly indicates the over-dispersion of data.
Hence attempts were made to develop Poisson-gamma and
NB regression models with number of crashes as dependent
variable and upward gradient(%), downward gradient(%),
cross slope (%), horizontal curvature (1/km), vertical curvature
(1/km), operating speed (km/hr), presence of median opening
(1 – if present; otherwise 0), presence of access points (1 – if
present; otherwise 0), and annual average daily traffic as
independent variables.
The models developed using Zero-Inflated regression
(count data parts, i.e., ZIP and ZINB) (due to the presence of
zeros and over-dispersion). In the case of NB regression
models few variables such as downward gradient, operating
speed, and AADT are found to be statistically significant for
most of the models at 1%, 5% and 10% significance levels.
The value of dispersion parameter, α is 0.274.
The log-likelihood value was found to be -554.153, which
is lowest of the three values. The mean AIC value was
estimated to be 1132.3, which is also the lowest value among
the other values. Log-likelihood ratio and Akaike‘s Information
Criteria values obtained using NB regression model were lower
than that of Poisson regression and Poisson-gamma models,
which shows the ability of NB model to predicting crashes as
shown in eqn. 8. Since the value of vuong statistics is less than
1.96, it is evident that negative binomial model is superior to
ZINB models.
B.Best Fit Models for Safety Analysis
Out of the models developed, the best model was identified
based on the goodness of fit test values. The identified models
(shown in Table 6), will enables the transportation engineers to
evaluate the safety performance of the existing highways and
suggest counter measures to obtain a particular level of safety.
More over this application can also be used by engineers and
planners for the design/re-design of new/existing highways.
The first part of the study enables the user to evaluate
safety directly from the available geometric design
characteristics, traffic characteristics and other roadway
features. If all these data are available, then safety level of the
highway under study can be evaluated using the model
developed. The second part is applicable when the speed data
