International Journal of Pavement Engineering

ISSN: 1029-8436 (Print) 1477-268X (Online) Journal homepage: https://www.tandfonline.com/loi/gpav20

Pavement serviceability evaluation using whole

body vibration techniques: a case study for urban
roads

Luis Fuentes, Rafael Camargo, Gilberto Martínez-Arguelles, Julius J. Komba,

Bhaven Naik & Lubinda F. Walubita

To cite this article: Luis Fuentes, Rafael Camargo, Gilberto Martínez-Arguelles, Julius J. Komba,
Bhaven Naik & Lubinda F. Walubita (2019): Pavement serviceability evaluation using whole body
vibration techniques: a case study for urban roads, International Journal of Pavement Engineering,
DOI: 10.1080/10298436.2019.1672872

To link to this article: https://doi.org/10.1080/10298436.2019.1672872

Published online: 30 Sep 2019.

Submit your article to this journal

Article views: 20

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=gpav20
INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING
https://doi.org/10.1080/10298436.2019.1672872

Pavement serviceability evaluation using whole body vibration techniques: a case

study for urban roads
a
Luis Fuentes , Rafael Camargoa, Gilberto Martínez-Arguellesa, Julius J. Kombab, Bhaven Naikc and Lubinda
F. Walubitaa,d
a
Department of Civil and Environmental Engineering, Universidad del Norte, Barranquilla, Colombia; bUniversity of Pretoria | Council for Scientific and
Industrial Research (CSIR), Pretoria, South Africa; cRuss College of Engineering and Technology, Ohio University, Athens, OH, USA; dTTI – The Texas A&M
University System, College Station, TX, USA

ABSTRACT ARTICLE HISTORY

Quantification of the pavement serviceability is most often based on the international roughness index Received 14 May 2019
(IRI). However, in urban environments it can become challenging to drive at 80 km/h as vehicle Accepted 18 September 2019
operating speeds are much lower, among other limitations associated with the use of IRI as an
KEYWORDS
indicator of ride quality in the urban context. This study was conducted to formulate an alternative Pavement serviceability
pavement serviceability evaluation criteria for low-speed roads (30–60 km/h) in urban settings. index; ride quality; whole
Deterministic and probabilistic models were developed to predict the PSI based on the whole-body body vibration; international
vibration (WBV) concepts, and thereafter, the results were compared with the ISO 2631 standard that roughness index
addresses human exposure to multiple mechanical shocks. The ISO 2631 standard uses a WBV-based
frequency-weighted root-mean-square acceleration parameter (aw) for evaluating discomfort in a multi-
axis environment. Based on the estimated models, an awz criterion of 0.98 m/s2 at a probability
acceptance of 85%, for a vehicle operating speed of 50 km/h, was proposed for urban roads in this
study. Overall, the study demonstrated that for accurate estimation of ride quality and comfort (i.e. PSI)
of low-speed urban roads, the evaluation criteria should be based on low vehicle speeds that are more
representative of urban field conditions.

1. Introduction
developing countries including Colombia (ABC 2011, ICPA
In general, the decision-making process of a road agency 2014, INVIAS 2015).
should be supported by a robust Pavement Management Sys- In the AASHO Road Test, it was shown that the roughness
tem (PMS). A PMS is an objective process to evaluate, design, of the longitudinal pavement profile is the variable that best
and programme timely maintenance and rehabilitation (M&R) describes users’ ratings, and therefore, the serviceability of a
activities to prolong the service life of the road network. For pavement (AASHO 1962a). Since then, several attempts have
instance, performance indicators such as pavement surface been made to relate the pavement roughness to the serviceabil-
roughness, cracking, rutting, etc., represent a key element to a ity concept in different contexts, mainly on rural and interstates
PMS. Based on these indicators, highway agencies can prioritise roads. The International Roughness Index (IRI) is one such
decisions and optimize the limited resources, while guarantee- popular index used widely in the PMS’s around the world
ing an acceptable level of service of the road network. Gener- (Múčka 2017a).
ally, road authorities try to define and maintain the In a study performed by Fuentes et al. (2019), the authors
performance indicators at acceptable levels to sustain/improve presented a summary of the pavement serviceability models
the ride quality (among others) on their road infrastructure. available in the literature along with their relevant details.
By definition, pavement serviceability is a concept that was One of the main findings of their study is the widespread
developed by the American Association of State Highway usage of the IRI for the serviceability estimations in different
Officials’ (AASHO) Road Test to assess the ability of a given contexts, purposes, and interpretations. In their study, Fuentes
pavement section to serve the traffic in its current condition et al. (2019) also formulated deterministic and probabilistic
(AASHO 1962a). Thus, it is a measure of the comfort and safety serviceability models for urban roads, and proposed some IRI
experienced by motorists/users when travelling on a road and pavement condition index (PCI) thresholds for the
(Carey and Irick 1960). For this reason, the present serviceabil- users’ acceptance of the pavement condition for developing
ity index (PSI) is one of the common indicators used worldwide countries.
in the design of pavement structures, as in the case of the However, since the IRI development in 1986 by the World
AASHTO 93 design method (AASHTO 1993). Despite the Bank (Sayers et al. 1986b), different studies have described
advances made in the design of pavements, such as the many limitations about the application of the IRI to urban
mechanistic-empirical (M-E) method, the AASHTO 93 design roads. One of the most reported limitation is the difference
method is still widely used today, literally in most of the between the simulation speed of the quarter car model

CONTACT Luis Fuentes [email protected]

© 2019 Informa UK Limited, trading as Taylor & Francis Group
2 L. FUENTES ET AL.

(80 km/h) and the typical operating vehicle speeds on urban estimations specific to urban roads with low operating
roads (30–60 km/h) (Abudinen et al. 2017). This discrepancy speeds.
can undesirably lead to an overestimation of the discomforts (2) Propose tentative thresholds for pavement serviceability
perceived by the users under real-time travel conditions. assessment of urban roads using the aw parameter as a
In consideration of the above challenges/limitations, the more representative characterisation with respect to travel
current approaches consist of complementing the evaluation quality; and lastly, compare the proposed aw thresholds to
of the users’ perception using parameters that adequately the ISO 2631 standard.
describe the users’ experience when travelling on the roads.
The ISO 2631 standard proposes a methodology to quantify The proposed approach was tested for low speed urban
the body vibrations with an index called -frequency weighted roads in the City of Barranquilla (Colombia). In the subsequent
root mean square (RMS) acceleration (aw)- (ISO 1997). The section, a literature review is presented followed by the study
aw index allows for the evaluation of the vibrational methodology. Formulation and development of the determinis-
effects induced by the pavement profile (surface) irregularities tic and probabilistic models are then presented including stat-
(or surface distresses) on human comfort at operational vehicle istical analysis and proposition of the acceptance thresholds/
speeds. This flexibility of the aw index inherently addresses the criteria. A synthesis and discussion of the results are then pre-
main limitations of the IRI when making estimations of servi- sented followed by a summary of conclusions and recommen-
ceability on low-speed roads, with operating vehicle speeds less dations for future work.
than 80 km/h. The use of representative indicators, such as the
aw index, for real-time travel conditions, is a fundamental com-
ponent of the PMS and is very critical for rational decision- 3. Review of literature and technical standards
making processes by the road authorities.
A literature review including the associated technical standards
is presented and discussed under this section. Aspects covered
include the pavement serviceability, ride quality, pavement
2. Objectives and scope of study
roughness, IRI, whole-body vibration concepts, and ride quality
Based on the above background and as a supplement to the IRI related studies on urban roads.
criteria, this objective of this study was two-fold:

(1) Analyze the relationship between user’s subjective percep- 3.1. Pavement serviceability
tion regarding the comfort level provided by a road and the
During a period of two years (1958–1960), the AASHO devel-
vertical accelerations induced by the road profile using the
oped the experimental phase of the AASHO Road Test. A full-
ISO 2631 whole-body vibration index concept. To accom-
scale trial whose main objective was the evaluation of the per-
plish this objective, deterministic and probabilistic models
formance of two types of structures with defined characteristics,
were formulated and utilised for pavement serviceability
namely pavements (rigid and flexible) and bridges (reinforced
concrete and structural steel) subjected to dynamic loads of
known magnitude and frequency (AASHO 1962a).
Considering that the fundamental function of a pavement is
to serve traffic (withstand traffic loading and provide comforta-
ble ride quality), and that, it is the users who can best evaluate
this service, the AASHO proposed the concept of serviceability.
The serviceability of a road represents the quality of the road as
perceived by users, in terms of comfort (ride quality) and can be
related to the different measurable distresses (e.g. roughness,
cracking, and rutting) on the pavement surface (Paterson
1987).
Due to the subjective nature of the serviceability concept, it
was necessary to numerically characterise the users’ opinions
and, as far as possible, find their relationship with quantifiable
physical characteristics (i.e. roughness, cracking, rutting, and
patching areas) of the pavement. Two fundamental indicators
emerged from this process, namely the Present Serviceability
Rating (PSR) and the Present Serviceability Index (PSI). The
PSR represents the average of user ratings for each pavement
section on a 0–5 scale (i.e. 0 ≤ PSR ≤ 5), using the rating
form shown in Figure 1 (for all pavement types). The PSI rep-
resents the result of the relationship between the PSR and pave-
ment objective parameters such as the IRI, PCI, and/or Rut
Depth (RD in the case of flexible pavements). In other words,
Figure 1. Rating form for serviceability study (ASTM 2018a). the PSI is used to describe/predict the serviceability (PSR)
 INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 3

through statistical models. That is, the PSI is essentially a pro- . The preparation and provision of appropriate/adequate rat-
duct of adjusting statistical models to estimate the perception of ing forms.
users based on objective parameters of the pavement surface. . Maintain consistency by the formulation and issuing of
As a result of the AASHO Road Test, the first pavement ser- appropriate instructions to the raters regarding the use of
viceability models were introduced in 1960 (Carey and Irick rating forms.
1960). The models proposed served as the basis for the pave-
ment design procedure adopted by AASHTO. Equation (1) In general, because of its subjective nature and to ensure
presents the model proposed for flexible pavements, while there is consistency between raters, the above considerations
Equation (2) is used for rigid pavements. should be adhered to when conducting pavement rating
studies. An important use of the resulting pavement rating
. Flexible pavements: data is to determine the ability of hypothesised functions of
√






the physical pavement parameters, such as profile roughness,
PSI = 5.03 − 1.91log(1 + SV) − 1.38(RD)2 − 0.01 C + P distresses indicators, etc., to provide appropriate and realistic
R2 = 0.84, SEE = 0.38 estimations of the users’ perception about ride quality
.
(ASTM, 2018a).
Rigid pavements:
√







PSI = 5.41 − 1.78log(1 + SV) − 0.09 C + P
3.2. Ride quality and pavement roughness
R2 = 0.92, SEE = 0.32
Ride quality refers to the level of comfort experienced by occu-
pants of vehicles as they travel on a section of roadway. This
Where SV = slope variance over the pavement section from level of comfort is strongly dependent on the in-vehicle
CHLOE profilometer measurements; RD = mean rut depth vibrations, especially vertical accelerations, induced by the
(in); C = cracking (ft/1000 ft2); P = patching (ft2/ 1000 ft2); R2 pavement surface roughness (Ahlin and Granlund 2002).
= coefficient of determination; and SEE = standard error of esti- Unevenness of the pavement surface causes vibrations on the
mate – where 1 ft ≅ 0.305 m and 1 ft2 ≅ 0.093 m2. users’ whole-body during vehicle motion, adversely affecting
In the AASHO Road Test, it was demonstrated that from the the ride quality and comfort. For these reasons, the study of
different variables contemplated in the field experiment to the pavement-vehicle-human interactions and the physical
evaluate the overall pavement condition, the surface roughness analysis of the vibration phenomena are both critical aspects,
of the longitudinal pavement profile, described in terms of SV, particularly for quantitatively relating pavement surface
was the variable that best described the users’ ratings (AASHO roughness to ride quality (Cantisani and Loprencipe 2010).
1962b). Since then, different concepts have been explored to As discussed subsequently, studies conducted to formulate
relate the pavement surface roughness to pavement serviceabil- mathematical representations of pavement surface roughness
ity. The IRI is one such concept and plays a key role in the pave- relative to travel quality, and support the PMS decision-making
ment condition assessment in many countries around the processes, yielded the well-known and widely used IRI concept
world (Múčka 2017a). (Múčka 2016, 2017a).
ASTM E1927-98 (2018a) describes a procedure for obtain-
ing subjective numerical ride ratings for a group of pavement
sections. The standard defines the guidelines for the field exper- 3.2.1. The international roughness index (IRI)
iment, particularly with aspects related to the suitable selection Based on the International Road Roughness Experiment
of pavement test sections and panel raters. In the ASTM stan- (IRRE) in Brazil, the World Bank, in 1986, published a docu-
dard, the average value of ride quality for each pavement sec- ment entitled ‘Guidelines for Conducting and Calibrating
tion is called the Mean Panel Rating (MPR). The MPR is Roughness Measurements’ – a calibration manual to introduce
equal, by definition, to the PSR developed in the AASHO and extend the IRI internationally (Sayers et al. 1986a). The IRI
Road Test. Key considerations in ASTM E1927-98 standard is based on a mathematical model that relates the vertical
include the following aspects (for both flexible and rigid responses of a quarter-car model (QCM) as it traverses a longi-
pavements): tudinal pavement profile at a constant speed of 80 km/h (Sayers
1995). For the simulation, the parameters of the QCM are set
. A minimum number of 20 pavement sections should be and normalised to represent a typical passenger car called
selected for each pavement type. ‘the Golden Car’. The simulated motion between the sprung
. Each section should have homogeneous physical character- and unsprung masses (passengers and vehicle components
istics throughout its length. masses) is accumulated and divided by the distance travelled
. The group of test sections should be well distributed by dis- by the QCM during a simulated ride at the standardised
tress level. speed of 80 km/h – see Equation (3) below.
. Pavement sections should be of equal length, long enough to 
1 L/V
provide an appropriate time of panel raters exposure. IRI = |ż s − ż u |dt (3)
L 0
. The panel size should be selected as a function of acceptable
error in MPR units. Where: IRI = the pavement surface roughness indicator pre-
. The rating panel should be representative of the road user sented in units of slope (mm/m, m/km or in/mi); ż s , ż u =
population. time derivatives of height (vertical coordinate) of sprung and
4 L. FUENTES ET AL.

unsprung masses, respectively; L = pavement profile length; (FHWA 2015), in urban areas, can overestimate the actual
and V = simulation speed (i.e. 80 km/h). surface-roughness condition of the pavements, leading to
For details on the mathematical processing of the longitudi- inappropriate and costly decisions by the road authorities.
nal pavement profiles to calculate the IRI, interested readers are . Lastly, for ride comfort evaluation, the mathematical formu-
directed to the ASTM E1926-98 (2015) standard. The algor- lation of the IRI has some disadvantages. For the same IRI
ithm developed in the IRRE study and reported in the World value, the users can be exposed to totally different travel
Bank technical reports (Sayers et al. 1986a, 1986b) and sum- experiences if these are performed at different speeds (Kro-
marised by Sayers (1995), consists of solving in time domain, pác and Múcka 2006, Cantisani and Loprencipe 2010, Abu-
the differential equations that represent the QCM, riding on dinen et al. 2017). In this way, it is possible to obtain
the pavement section profile (measured through any of the significantly different users’ perception of a set of pavements
methods described by the ASTM E950/950M-09 (2018b) and with the same IRI value. In addition, the QCM is used to cal-
ASTM E1364-95 (2017) standards), at 80 km/h. culate the suspension relative motion of a simulated vehicle
In a study carried out by Fuentes et al. (2019), the authors, ride. However, the ride comfort is related to vehicle
based on the traditional approach of predicting serviceability vibration rather than displacements (Múčka 2016).
using the IRI, formulated probabilistic and deterministic servi-
ceability models for urban roads in Colombia. However, some Given these challenges and limitations associated with the
studies have been reported that the IRI presents certain limit- IRI concept, this study aims to propose the use of a recognised
ations that warrants addressing, particularly on low-speed alternative indicator, which can be more suitable for modelling
roads (Ahlin and Granlund 2002, La Torre et al. 2002, Arhin user perception about the pavements in the urban areas. As dis-
et al. 2015, Múčka 2016, Abudinen et al. 2017). Some of cussed subsequently, the envisioned alternative/supplementary
these challenges and limitations include the following: indicator is based on the whole-body vibration concept.

. The IRI was developed on rural roadways based on pavement 3.2.2. The whole-body vibration concept
sections of 320 metres (1050 feet). However, on roads in The vehicle industry practice shows that ride quality assess-
urban locations, the length of pavement sections is defined ment is preferably based on car vibrations, rather than suspen-
mainly by intersections or junctions, which are usually sion strokes, which corresponds to the displacement
found every 150 metres (500 feet). Since the IRI is sensitive accumulation encompassed in the IRI model (Ahlin and Gran-
to the base length over which it is calculated, it is common lund 2002). In the same way, road agencies should identify and/
to obtain IRI values significantly higher than what the pave- or formulate appropriate indicators to manage the road net-
ment condition actually reflects (La Torre et al. 2002). works and guarantee acceptable comfort levels. Considering
. The IRI model adopts a constant modelling speed of 80 km/h. the IRI limitations in urban roads, alternative approaches are
However, on urban roads, it is typical for vehicle operating required for travel quality assessment of pavements, particu-
speeds to be significantly lower. For example, in the Colombian larly those located within the city limits.
case (City of Barranquilla), the vehicle speeds on urban roads The ISO 2631 standard (ISO 1997) proposes a methodology
within the city limits are usually less than 60 km/h. In addition, to quantify the vibrations that have an influence on human
motorists comfort is strongly related to the vehicle speed and, comfort and health. This method allows for modelling of the
therefore, IRI values can overestimate (or underestimate) the whole-body vibration (WBV) into a single parameter from
users’ real subjective perception about the road condition. In the analysis of accelerations perceived by users when travelling
order to address this limitation, some authors have proposed on a road. Specifically, the RMS value of the frequency-
different methodologies to calculate the IRI using different weighted acceleration time history, called aw, is one such par-
speeds other than the one defined in the ASTM standard (Per- ameter for effectively accomplishing this (ISO 1997). The fre-
era and Kohn 2004, Loizos and Plati 2008). quency weighting is the mathematical process to represent
. The IRI is influenced by particular characteristics associated the manner in which vibration affects human health and com-
with urban roads such as drainage provisions, frequent inter- fort differently depending on the vibration frequency and load-
sections, speed humps, roundabouts, frequent stop-go zones ing time history.
(i.e. braking/accelerating), pedestrian stops/crossings, and It is important to highlight that the ISO 2631 (ISO 1997)
numerous interfaces with utility access boxes that affect the defines a procedure for processing the acceleration data from
IRI significantly without representing greater impact on the a real travel experiment. During a car trip, pavement surface
perception of the users about the road quality (Fuentes unevenness causes translational and rotational vibration on
et al. 2019). In other words, these characteristic factors hardly the occupants in all directions. However, as in the case of the
influence the users’ judgement of the road quality and ride IRI and other roughness indicators, the most used approaches
comfort despite being quantitatively detrimental in terms of in the literature for roughness assessment consist of the riding
magnifying the IRI values. The high IRI values for urban mathematical simulations based on a measured road profile. In
areas can exaggerate the scope of M&R activities required if this way, it is not only possible to define the simulation par-
these are not programmed based on appropriate/adequate ameters according to the real travel conditions, but also study
indicators and thresholds for each specific context (Arhin hypothetical scenarios of interest. In the case of the QCM,
et al. 2015). For example, the adoption of conventional IRI which is used subsequently in this study, the model can only
scales or international IRI thresholds such as the roughness describe the vertical accelerations experienced by the sprung
thresholds used by the Federal Highway Administration and unsprung masses (Sayers 1995). To avoid confusion and
 INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 5

differentiate from aw, the parameter to model the vibrations in for some applications where the IRI cannot be accurately
this study will be denoted as awz and mathematically expressed approximated, the authors recommended implementing direct
as shown in Equation (4): calculations based on accelerations measurements.





















 La Torre et al. (2002) developed a correlation between user
2
awz = (Wk,i × aRMS
iz ) (4) acceptability of pavement condition, user ratings on a PSR
i scale, and pavement surface roughness. The study was based
Where: on data from 17 sections of flexible and articulated pavements.
aRMS = vertical RMS acceleration values on the vehicle occu- The roughness was characterised by a modified IRI* model (the
iz
pants’ body calculated by means of the power spectral density IRI model using a base length of 50 m and a riding speed of
(PSD) for each ith octave thirds band; 50 km/h instead of 80 km/h) for a more approximate represen-
Wk,i = frequency weighting factors for aRMS tation of the real travel conditions on urban roads.
iz , according to
Table 3 of ISO 2631-1 standard guide/manual; and Cantisani and Loprencipe (2010) proposed and calibrated a
awz = vertical frequency-weighted acceleration. ‘full car model’ to represent a complete real vehicle. This model
The vertical accelerations of the sprung mass (passengers) was used to calculate the WBV (awz values) for a sample of pave-
can be extracted from the QCM in order to compute awz ment sections and various speeds (30–90 km/h). In total, 124
according to the ISO standard (ISO 1997). As it can be pavement sections extracted from the Strategic Highway Research
observed, the awz model, proposed in Equation (4), has some Program (SHRP) Long-Term Pavement Performance (LTPP)
advantages over the traditional IRI model (as reviewed in the were analysed. The authors developed correlations between the
literature), including the following: standardised IRI model and the awz values for each speed con-
sidered. Based on the indicative comfort values of the ISO 2631
. The IRI computation is standardised to a reference speed of standard (ISO 1997), they proposed speed-related IRI thresholds
80 km/h, while the awz can be computed for any specific to evaluate ride quality on low-speed roads.
operational speed conditions of the given road (Cantisani Arhin et al. (2015) developed a model to predict the IRI based
and Loprencipe 2010, Abudinen et al. 2017). on the PSR in urban areas. 122 flexible pavement sections selected
. The parameters of the QCM for the IRI computation are from the District Department of Transportation (DDOT) were
fixed to the golden car parameters. However, for the awz used in the study. The authors also proposed IRI thresholds
mathematical processing, the accelerations can be measured based on user perception and compared with the IRI values
in a field experiment or estimated from calibrated simu- defined by the FHWA. In their recommendations, the authors
lation models that represent real travel conditions (Cantisani emphasise the need to define suitable thresholds based on the
and Loprencipe 2010, Abudinen et al. 2017). actual pavement surface roughness perceived by motorists.
. For some applications, the IRI-based models can provide Aleadelat et al. (2016) developed a pavement serviceability
misleading estimates when assessing travel quality. It has prediction model for county paved roads based on the analysis
been shown that it is possible to experience different levels of 30 flexible pavement sections. The proposed models
of vibrations when travelling on pavements with the same included recognised pavement indicators such as the IRI,
IRI values (Ahlin and Granlund 2002, Kropác and Múcka PCI, and Rut Depth (RD). Overall, the literature review indi-
2006, Cantisani and Loprencipe 2010). The awz allows to cated that the models proposed by the Aleadelat et al. (2016)
differentiate and quantify this effect. provided a more realistic representation of the conditions of
county roads compared to the statewide model used on the
Lastly, it is important to note that some authors have proposed state’s highway system.
correlations between the awz and the IRI, these have been sum- Abudinen et al. (2017) described a methodology for the tra-
marised and analysed by Múčka (2017b). vel quality assessment for urban roads. The authors, using a
database with information on more than 80 rigid pavements,
estimated linear regression models to relate the IRI with the
3.3. Ride quality related studies on urban roads
awz values for various riding speeds (30–100 km/h). In their
Over the last decades, several authors have proposed different findings, the authors proposed speed-related IRI thresholds
approaches to assess the level of service offered by pavement for pavement management activities.
to users in the urban-road context (Ahlin and Granlund Loprencipe and Pantuso (2017) proposed an alternative
2002, La Torre et al. 2002, Yu et al. 2006, Cantisani and Lopren- pavement conditions assessment methodology that accounts
cipe 2010, Arhin et al. 2015, Abudinen et al. 2017, Kırbaş and for other pavement distresses occurring in urban roads such
Karaşahin 2018, 2019, Loprencipe et al. 2019, Nguyen et al. tree roots and artificial elements (catch basins and manholes)
2019). These studies have shown a wide variety of analytical not considered in the ASTM D6433 distress catalogue. The
techniques for ride quality data processing, with the aim of authors introduced new deduct value curves to complete the
enhancing the decision-making processes within a PMS. ASTM D6433 Distress Identification Catalogue to assess the
Ahlin and Granlund (2002) estimated some relationships urban road pavement surfaces.
between pavement surface roughness, in terms of IRI and a Kırbaş and Karaşahin (2018, 2019) investigated the influence
PSD indicator, vehicle speeds and the WBV experienced by of a distress indicator, the PCI, on ride comfort. In their studies,
users based on analytical ride simulations. These models can PCI and vibrational measurements were performed on flexible
be used to convert the WBV limit values to correspondingly urban asphalt-concrete pavements, according to the PAVER sys-
approximate the IRI values for an IRI-based PMS. However, tem and the ISO 2631 standards. Using logistic regression,
6 L. FUENTES ET AL.

Table 1. Studies related to ride quality on urban roads.

Reference source Speed Indicator Criteria Proposed thresholds
(La Torre et al. 2002) 50 km/ IRI* (m/ Percentage of users satisfied (%) 4.1 (85%), 5.2 (75%), 6.9 (60%), 8.2 (50%)
h km)
(Cantisani and 50 km/ IRI (m/km) Pavement condition based on indicative comfort <2.98 (Very Good), 2.98–5.95 (Good/Fair), 5.95–8.51
Loprencipe 2010) h values of ISO 2631 standard (Mediocre), >8.51 (Poor)
(Arhin et al. 2015) – IRI (m/km) FHWA pavement condition scale and IRI-based Freeways: <2.0 (Good), 2.0–3.4 (Acceptable)
serviceability models Arterials: <2.9 (Good), 2.9–4.4 (Acceptable)
Collectors: <3.0 (Good), 3.0–5.0 (Acceptable)
(Abudinen et al. 2017) 50 km/ IRI (m/km) Pavement condition according to IRI- awz models and <2.05 (Very Good), 2.05–4.1 (Good), 4.1–5.86 (Regular),
h ISO 2631 comfort values >5.86 (Bad)
(Kirbaş and Karaşahin 50 km/ PCI Indicative comfort values of ISO 2631 Standard <51 (Fairly Uncomfortable), 51–78 (A Little Uncomfortable),
2018) h >78 (Not Uncomfortable)
(Kirbaş and Karaşahin 50 km/ PCI Indicative comfort values of ISO 2631 standard <45 (Fairly Uncomfortable), 45–78 (A Little Uncomfortable),
2019) h >78 (Not Uncomfortable)
(Fuentes et al. 2019) 50 km/ IRI(m/km) Probabilities of pavement condition acceptance (%) 3.9 (95%), 5.9 (85%), 6.9 (75%), 8.7 (50%), 10.5 (25%)
h
(Fuentes et al. 2019) 50 km/ PCI Probabilities of pavement condition acceptance (%) 71 (95%), 61 (85%), 55 (75%), 46 (50%), 37 (25%)
h

artificial neural network (ANN), and fuzzy logic techniques, they Phase II of the research approach comprised of an objective
proposed some threshold values of PCI for ride comfort. evaluation where performance indexes such as the IRI and the
Finally, Table 1 summarises the findings of the different PCI were evaluated on each pavement section. The pavement sec-
studies related to ride quality on urban roads, emphasising tions evaluated presented an average IRI value of 6.1 m/km with a
on the proposed thresholds for pavement sections that have standard deviation of 1.6, and an average PCI of 66 with a stan-
an operating speed of 50 km/h. dard deviation of 18, respectively. Detailed results and data docu-
mentation for these evaluations can be found in Fuentes et al.
(2019). However, for a better subjective characterisation of the
users’ perception of ride quality in the urban-road context, an
3. Study methodology and research approach additional parameter, namely the vertical frequency-weighted
Urban roads in the City of Barranquilla (Colombia) served as RMS acceleration (awz), was calculated in this study. The awz
the case study and encompassed over 50 city roads – all with was used for the serviceability estimations in this study in order
rigid (concrete) pavement structures. The field data measured to propose an alternative approach for the functional evaluation
and collected included road profiles, pavement distresses of pavements within the PMS’ framework in urban areas.
(PCI), users’ perceptions (PSR), and acceptance responses For field measurements and data collection, pavement profiles
regarding the condition of the pavement sections (Fuentes were measured continuously using a SurPro walking profiler
et al. 2019). A two-phase approach was devised and executed along the outside wheel path. Thereafter, the vertical whole-
for collecting the field data. Phase I comprised of a subjective body vibrational values, i.e. awz, were simulated/modelled using
evaluation where a riding quality experiment was conducted the QCM at the typical vehicle operating speeds prevalent on Bar-
following the guidelines defined by ASTM E1927 (ASTM ranquilla’s urban roads, namely 20–90 km/h, with a step incre-
2018a). In particular, the ASTM E1927 standard procedures ment of 10 km/h. Thus, the vehicle properties for awz
for the adequate selection of road test sections and confor- simulations in this study were based on the Golden Car (QCM)
mation to the evaluation panel (raters) requirements were com- defined by the ASTM standard (ASTM 2015, 2017, 2018a,
pliantly adhered to. More than fifty (50) rigid pavement 2018b). Using an algorithm developed in the MATLAB program-
sections, generally 3.5 m wide with lengths ranging between ming platform (MATLAB 2017), the awz values were computed
50 and 200 m, were selected from a representative group of and determined in accordance with the ISO 2631 standard pro-
roads, based on certain physical characteristics (surface rough- cedure (ISO 1997).
ness and distresses), of the City of Barranquilla, Colombia.
Note that over 90% of urban roads in the City of Barranquilla 4. Pavement serviceability modelling and statistical
comprise of rigid pavement structures and thus, only rigid analysis
pavement sections were included in the study matrix.
In addition, the rating panel size was defined to be 15 members This section discusses the pavement serviceability modelling
based on 0.3 MPR maximum error with a normal distribution, and estimations including statistical analysis. Model formu-
according to the ASTM specification (ASTM 2018a). The raters lation included both the deterministic and probabilistic
were representatively selected from the road user population methods. The tentatively proposed serviceability thresholds
based on age, sex, driving experience, and occupation; taking based on the whole-body vibration concept are also presented
into account that these variables could influence the perception and discussed in this section.
of users regarding road condition (Arellana et al. 2019). The
panel members (raters) were trained before the experiment fol-
4.1. Deterministic modelling
lowing the ASTM guidelines. The average rating of the panel
members for each pavement section was reported as the MPR As documented in Fuentes et al. (2019), the data collected
and subsequently, used for serviceability estimations. was compiled in a database that included the estimated awz
 INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 7

values at typical operating speeds on Barranquilla’s urban

roads (20–90 km/h with an incremental step of 10 km/h),
MPR, and acceptance responses regarding the road
infrastructure.
Figure 2 depicts the relationship between awz and PSI at the
indicated vehicle speeds by means of four exponential models.
As theoretically expected, pavement serviceability decreases as
the vibrational levels represented by awz increases.
It is important to highlight that the exponential models
defined by Equation (5) to (8), as derived from Figure 2, are
restricted to a PSI maximum value of 5.0 when awz is 0.0.
This is based on the theoretical argument that a pavement in
optimal conditions is one that does not generate discomfort
to its users.

PSI = 5e−1.164(awz(30) )
R2 = 0.70 Figure 3. Sensitivity analysis for the coefficient of determination.
−0.785(awz(40) )
PSI = 5e
R2 = 0.72 Mathematically, it should be considered that exponential
models constitute a particular form of linear regression models.
PSI = 5e−0.478(awz(50) )
A linear model accounts for both a deterministic component
R2 = 0.84 and a probabilistic (random error) component (Arhin et al.
PSI = 5e−0.382(awz(60) ) 2015). However, for practical reasons, one usually uses
regression models to obtain the expected value of the depen-
R2 = 0.74
dent variable (in this case PSI), assuming that the error term
Where awz is expressed in m/s2. Additionally, Figure 3 shows has a mean equal to zero – hence, ignoring the random nature
the sensitivity analysis for the coefficient of determination con- of the dependent variable.
sidering a wide range of modelling/simulation speeds The ability of the awz parameter to reflect the user percep-
(20–90 km/h). It is also worth noting that as illustrated in tion in terms of pavement serviceability, for each considered
Figure 3, the coefficient of determination (R2) between PSI speed, was quantified using R2 concept. The R2 value calculated
and awz for different speeds is 50% and higher. The highest for a simple linear regression is a statistical measure of how well
R2 value is 84% for 50 km/h and the lowest (50%) corresponds the regression line describes the real data points (Navidi 2008).
to 90 km/h. These R2 values indicate a good correlation-ship As one can observe from Figure 3, based on the highest R2 value
and suggest that PSI can be estimated to a minimum accuracy (i.e. 84%), the model that best adjusts the user ratings in the
of 50% from awz. Based on Figure 3 and Equations (5) to (8), present study is defined by Equation (7), which corresponds
the highest PSI prediction accuracy and statistical confidence, to a modelling speed of 50 km/h. In summary, the high R2
would therefore be for 50 km/h vehicle speed with an R2 values are associated with the typical operating characteristics
value of 84%. of the urban roads considered in the study. Hence, the user’s
perceptions are defined by the experience when they travel
on the roads in a context where the speeds are between 20
and 90 km/h. On the other hand, when one consider speeds
away from typical operating conditions (around 50 km/h) the
simulated values of awz can provided a travel experience differ-
ent from the one that users actually experience when travelling
on a pavement, and therefore, the predictive capacity of the
model is diminished.
The above results/findings are consistent with the contri-
butions made in previous studies regarding the importance
of the 50 km/h speed when studying urban roads, particularly
in developing countries (La Torre et al. 2002, Kırbaş and
Karaşahin 2018, Fuentes et al. 2019). In fact, the user comfort
is highly dependent on the vehicle speed. Therefore, the user
ratings are influenced by their travel experience, particularly
in a situation where the average speed limit is 50 km/h. For
all these reasons, the pavement serviceability analysis of
urban roads in this study will thus be based on simulated ver-
tical accelerations at 50 km/h (awz(50)). Figure 4 shows the dis-
Figure 2. Relationship between PSI and awz. tribution of the reported average user ratings for each rigid
8 L. FUENTES ET AL.

Table 3. Estimation results of ordered logit model.

Parameter Estimate S.E. Z p-value Odds ratio 95% CI (α = 0.05)
τ1 −10.52 2.41 −4.36 <0.001
τ2 −7.11 1.36 −5.23 <0.001
τ3 −4.14 0.76 −5.45 <0.001
τ4 −1.17 0.54 −2.18 0.03
bawz −3.74 0.78 −4.78 <0.001 0.02 (0.01, 0.11)
Legend: S.E. = Standard Error; Z = Standard Score; CI = Confidence Interval; α =
Significance Level.

in Table 3 according to Equation (10).

1(Very Poor) if −1 , PSI∗ ≤ t1
2(Poor) if t1 , PSI∗ ≤ t2
PSI = 3(Fair) if t2 , PSI∗ ≤ t3 (10)
4(Good) if t3 , PSI∗ ≤ t4
5(Very Good) if t4 , PSI∗ ≤ 1
The model parameters were estimated using the maximum
Figure 4. Relationship between PSI and awz(50). likelihood procedure (Hosmer et al. 2013) in the SPSS statistical
software (SPSS 2017), and the relevant results are summarised
in Table 3.
pavement section and their corresponding simulated awz(50) The coefficient bawz in Table 3 can be used to estimate the
values. odds ratio. In logistic regression models, the odds ratio (OR) is
From Figure 4 and Equation (7), it is possible to define the used to measure the association between the dependent variable
awz ranges based on the subjective serviceability scale shown in (i.e. PSI) and the independent variable (i.e. awz) (Hosmer et al.
Table 2. 2013). Specifically, it tells you how an increment of 1 m/s2 in
awz has an effect on the PSI. However, considering the range
of the awz values (the order of magnitude seen in Figure 4), a
5.2. Probabilistic modelling change of 1 m/s2 is too large and 0.1 m/s2 is more realistic
change for a given pavement in consecutive years. Hence, to pro-
The model defined in Equation (7) can be used to determine
vide a useful interpretation, the OR for a change of 0.1 m/s2 in
the serviceability of urban roads based on the vertical
awz is OR(0.1) = exp(−3.74 × 0.1) = 0.69. This estimate implies
vibrations experienced by users, quantified in terms of the
a 31% reduction in the odds of assigning a higher PSI level per
awz parameter. However, this model was proposed using a
0.1 m/s2 increase in awz, or mathematically, 31% reduction in
deterministic approach, and in its formulation, the subjectiv-
the odds of (PSI > k) over (PSI ≤ k), where k represents one of
ity of the ratings in the range of the measured data was not
the five possible levels for the categorical variables of the PSI.
considered.
Additionally, Equation (11) can be used to estimate the
To address this limitation, a probabilistic approach was used
cumulative probability of observing a specific PSI level based
to predict the likelihood of obtaining a given serviceability rat-
on the discomfort experienced by users due to the prevailing
ing based on the vertical accelerations. Specifically, an ordered
pavement condition. In this sense, one could modify Equation
logit model was formulated to study the relationship between
(11) to obtain the probability of observing a specific PSI level of
PSI and awz(50). The ordered logit model was chosen consider-
a pavement section given the simulated acceleration.
ing the ordinal nature of the dependent variable (PSI subjective
levels), which was modelled using an unobserved variable e(tk − bawz × awz(50) )
(PSI*), according to Equation (9). P(PSI ≤ k|awz(50)) = (11)
1 + e(tk − bawz × awz(50) )
PSI∗ = bawz × awz(50) + 1 (9) The negative value of bawz implies that the probability of ranking
Where: a pavement section with a high PSI value (4 or 5) decreases as the
bawz = the regression coefficient; and awz(50) increases, i.e. increasing discomfort. As an example,
1 = the random disturbance variable. according to the estimated logit function, a pavement whose
The predicted ordinal PSI level can be obtained using the irregularities generate vertical accelerations of awz(50) = 0.15 m/
results of Equation (9) and the estimated thresholds τ found s2, presents the following probabilities of being rated by users
as follows: Probability (PSI = 1) = 0.0%, Probability (PSI = 2) =
0.1%, Probability (PSI = 3) = 2.6%, Probability (PSI = 4) =
Table 2. awz thresholds for serviceability based on deterministic approach.
32.4%, and Probability (PSI = 5) = 64.9%, respectively.
PSI Pavement condition awz(50)
Additionally, Figure 5 presents a sensitivity analysis con-
(4, 5] Very good <0.46
(3, 4] Good (0.46, 1.06)
ducted using Equation (11) that allows the understanding of
(2, 3] Fair (1.06, 1.91) the effect of the discomforts generated by the pavement in
(1, 2] Poor (1.91, 3.36) the users, measured by awz, in their subjective criteria when
[0, 1] Very poor >3.36
evaluating a pavement with a specific PSI level.
 INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 9

Figure 5. Sensitivity analysis for ordinal logistic model based on the awz(50).

It is important to note that the PSI levels were grouped into to obtain the critical awz(50) values.
five categories according to the qualitative scale of the service-
ability. Therefore, the results must be interpreted cautiously con- 1
P(Z = 1|awz(50)) = (12)
sidering that each category considers a range of values. For 1 + e− (b0 + b1 × awz(50) )
example, PSI = 1 corresponds to user ratings of 1 or below
(≤1), PSI = 2 corresponds to ratings between 1 and 2 (1 < Where Z is the dependent variable; P (Z = 1) is the predicted
PSI ≤ 2), and so on, until the last category, which is PSI = 5. probability that the binary variable takes a numerical value of
As depicted in Figure 5, pavements in excellent conditions (awz one – that is, accept the vibration levels; and b0 , b1 are
values close to zero) are more likely to be rated with high PSI values regression coefficients.
(4–5), while pavements in poor conditions are more likely to be The maximum likelihood procedure was then used to esti-
quantitatively/numerically rated with PSI values less than two. mate the model parameters, namely b0 and b1 (Hosmer
A similar analysis to that used in the deterministic approach et al. 2013). The relevant details of the modelling are presented
can be used to define the awz ranges based on the subjective servi- in Table 5. In general, the results show a good fit for the model
ceability scale. Using Equation (11), it is possible to determine the with a p-value significantly less than 0.05 at 95% confidence
awz(50) ranges for which the probability of users qualifying the level.
pavement at each PSI level reaches the highest values – see From Table 5, the estimated OR for an increase of 1 m/s2 in
Table 4. awz is 0.04. However, as discussed before, 0.1 m/s2 was used
considering that it represents a more realistic change of pave-
ment condition in consecutive years. Therefore, the odds
5.3. Formulation of WBV-based acceptance criteria and ratio OR(0.1) = exp(−3.22*0.1) = 0.72, implies a 27% reduction
thresholds in the odds for accept pavement condition per 0.1 m/s2 increase
in awz. In other words, for every increase of 0.1 m/s2 in the ver-
An analysis was conducted using the information of the users
tical accelerations experienced by users, the ratio of the prob-
regarding the acceptance or rejection of a rigid pavement sec-
ability that users accept a given pavement section over the
tion given its prevailing condition with the aim of proposing
probability that they reject it decreases 27%.
and recommending some acceptance thresholds for vertical
Additionally, the predicted probability of acceptance was
accelerations. In the field experiments, raters were asked to
computed using Equation (12) with the parameters from
identify whether the pavement sections were acceptable in
Table 5 as inputs. The corresponding results are graphically
terms of riding comfort. The two possible answers and response
plotted in Figure 6.
outcomes, namely ‘accept’ or ‘reject’ were modelled in the SPSS
As recommended in previous investigations by Fuentes et al.
statistical software (SPSS 2017) as a binary variable Z.
(2019), a value of P = 0.85 was used to define the acceptability
Considering Z = 1 for acceptance, and Z = 0 for rejection, a
binary logit model, expressed in Equation (12), was formulated
Table 5. Estimation results of acceptance logit model.
p- Odds 95% CI (α =
Table 4. awz thresholds for serviceability based on probabilistic approach. Parameter Estimate S.E. (W 2) value ratio 0.05)
PSI Pavement condition awz(50) b0 4.904 1.242 15.601 <0.001
(4, 5] Very Good <0.35 b1 −3.22 1.109 8.428 0.004 0.04 (0.004, 0.351)
(3, 4] Good (0.35, 1.11) Likelihood Ratio Test Statistic: 25.15
(2, 3] Fair (1.11, 1.89) P-value: < 0.001
(1, 2] Poor (1.89, 2.80) Legend: S.E. = Standard Error; W = Wald Statistic; CI = Confidence Interval; α = Sig-
[0, 1] Very Poor >2.80 nificance Level.
10 L. FUENTES ET AL.

Table 7. ISO 2631 indicative acceleration values for ride comfort.

aw (m/s2) Expected comfort level
<0.315 Not uncomfortable
0.315–0.63 A little uncomfortable
0.5–1.0 Fairly uncomfortable
0.8–1.6 Uncomfortable
1.25–2.5 Very uncomfortable
>2.0 Extremely uncomfortable

an approximation to these values, ignoring the other accelera-

tion components.
Figure 7 summarises the most relevant findings of this study.
The graph shows the results obtained from the deterministic and
probabilistic modelling including the proposed awz thresholds
for pavement condition acceptance and the values suggested
by the ISO 2631 standard (ISO 1997) for the evaluation of the
vibrations that affect the travel quality and ride comfort.
It is interesting to note that the thresholds summarised in
Figure 6. Predicted probability of acceptance based on the awz(50). Figure 7 are consistent and/or overlapping with the ISO 2631
standard (ISO 1997). For example, for awz values greater than
thresholds. This value is also convenient because, below this 2 m/s2, which mean approximately PSI qualitative categories
value (i.e. less than 85%), the probability decreases significantly of ‘Poor’ and ‘Very Poor’ pavement condition according to
for a small change in the predictor variable – see Figure 6. From the deterministic and probabilistic approaches, correspond to
Figure 6, one can also observe a critical value of awz(50) = 0.98 m/ unacceptable acceleration values according to the acceptance
s2. However, depending on the local conditions and available threshold, and an ‘Extremely Uncomfortable’ level of comfort.
resources, the road authorities can define thresholds for different Similarly, awz values lower than 0.5 m/s2, qualitatively describe
acceptance probabilities. Table 6 summarises some proposed roughly a pavement in ‘Very Good’ condition, with a high prob-
values of interest that can be used in the PMS decision-making ability of acceptance (P ≥ 0.85), and a ride experience qualitat-
process to improve the ability of the road infrastructure to satis- ively described as ‘Not Uncomfortable’.
factorily serve its users in terms of safety and comfort. Considering that Colombian local road agencies have yet to
establish a PMS that defines the policies for the programming,
timing, and scheduling of M&R activities, the thresholds pro-
5. Synthesis and discussion of the results posed in this study as summarised in Figure 7 could be tenta-
The definition of ride comfort varies depending on personal tively used to define a prioritisation criteria that take into
expectations and perceptions about the pavement conditions considerations how users perceive the urban road conditions.
and user travelling experience. The ISO 2631 standard (ISO
1997) gives approximate indications of the likely reactions to
6. Conclusions and recommendations
various magnitudes of the WBV for application in public trans-
port, as presented in Table 7. However, considering the subjec- Using the City of Barranquilla’s urban rigid (concrete) pave-
tivity in the opinion of users in different contexts such user ments as the datum source, this study was conducted to
travel experience, comfort thresholds are not defined in the
ISO 2631 standard (Kırbaş and Karaşahin 2018, Arellana
et al. 2019). In fact, the limit values for the acceleration’s ranges
are often in conflict due to multiple overlaps in the aw values –
see Table 7. For example, according to the ISO, aw values
between 0.8 and 1.0 m/s2 can be qualitatively described as
‘Fairly Uncomfortable’ or ‘Uncomfortable’. It is important to
note that the values proposed by the ISO 2631 standard in
Table 7 are based on acceleration signals in three orthogonal
directional axes and therefore, the awz parameter represents

Table 6. Proposed pavement acceptance

thresholds based on the awz parameter.
Predicted probability awz(50)
of acceptance (%) (m/s2)
95 0.61
85 0.98
75 1.18
50 1.52
25 1.86 Figure 7. Travel quality and ride comfort thresholds (rigid pavements).
 INTERNATIONAL JOURNAL OF PAVEMENT ENGINEERING 11

formulate some alternative pavement serviceability criteria for considered full PSI-based models that incorporate IRI, PCI,
low-speed roads (30–60 km/h) in an urban setting. Determinis- and RD (in case of flexible pavements). Similarly, evaluation
tic and probabilistic modelling of PSI, in the SPSS statistical of the simulated frequency-weighted acceleration response in
software, were executed based on the WBV concepts, and a frequency domain such as the power spectral density for
thereafter, the results were compared with the ISO 2631 stan- different velocities is also recommended for future studies.
dard. The key findings and recommendations drawn from Lastly, although the results and findings reported herein per-
the study, based on an analysis of over 50 rigid (concrete) pave- tain to the rigid (concrete) pavements evaluated in this study,
ment sections, are summarised as follows: the methodology and concepts adopted can be replicated to
urban flexible pavements.
. The formulated frequency-weighted root-mean-square
acceleration (awz) parameter, based on the WBV concept
and accounts for simulated vertical acceleration, offers a Acknowledgements
more realistic representation for modelling the serviceability The authors gratefully acknowledge and thank all those who helped with
and ride quality of low-speed urban roads. The awz par- field measurements, pavement profile data collection, and documentation.
ameter is sensitive for any surface distresses, irregularities The City of Barranquilla (Colombia) is also duly acknowledged for facili-
or disturbances that may produce a reaction at the tire-pave- tating the study and using their jurisdictional rigid (concrete) roads as the
case study.
ment interface. The robustness and flexibility of the awz par- The contents of this paper reflect the views of the authors who are
ameter allow it to accommodate and compute PSI values for responsible for the facts and accuracy of the data presented herein and
any operating vehicle speed within an urban setting. Plaus- do not necessarily reflect the official views or policies of any agency or
ible results were obtained in this study for a speed range institute. This paper does not constitute a standard, specification, nor is
of 30–60 km/h. it intended for design, construction, bidding, contracting, tendering, cer-
tification, or permit purposes. Trade names were used solely for infor-
. For the vehicle speeds considered in this study (i.e. 30– mation purposes and not for product endorsement, advertisement,
60 km/h), deterministic modelling yielded high PSI promotions, or certification.
prediction accuracy (84%) and statistical confidence for an
operating vehicle speed of 50 km/h. This finding aligns
well with the fact that the speed limit for most urban Disclosure statement
roads in cities such as Barranquilla rarely exceeds 50 km/h, No potential conflict of interest was reported by the authors.
i.e. speed limit < 50 km/h.
. Various serviceability thresholds based on both determinis-
tic and probabilistic modelling were successfully formulated Funding
for the awz parameter, which corresponding to a PSI scaling
This work was supported by Departamento Administrativo de Ciencia,
range of 0–5.0 and pavement condition rating range of ‘Very Tecnología e Innovación: [Grant Number 1215-745-59105].
Poor’ to ‘Very Good’.
. Using the WBV concept for an operating vehicle speed of
50 km/h, an awz criterion and threshold of 0.98 m/s2 at a ORCID
probability acceptance of 85% was proposed. Thus, in lieu
Luis Fuentes http://orcid.org/0000-0002-7811-8821
of the traditional perspective for travel quality assessment
(i.e. the IRI-based models), this study recommends to con-
sider this tentative criterion proposed herein for quantifying References
and rating the serviceability and ride quality of low-speed
urban rigid (concrete) roads. AASHTO, 1993. AASHTO guide for design of pavement structures.
Washington, DC: The Association.
ABC, 2011. Manual de Diseño de Conservación Vial. La Paz:
Overall, this study has demonstrated that for accurate esti- Administradora Boliviana de Carreteras.
mation of the ride quality and comfort, in terms of PSI, of Abudinen, D., Fuentes, L.G., and Muñoz, J.S.C., 2017. Travel quality
low-speed urban roads, the evaluation criteria should corre- assessment of urban roads based on international roughness index:
spondingly be based on low vehicle speeds that are more repre- case study in Colombia. Transportation Research Record: Journal of
the Transportation Research Board, 2612, 1–10.
sentative of urban field conditions. For road agencies such as Ahlin, K. and Granlund, N.O.J., 2002. Relating road roughness and vehicle
the City of Barranquilla, the proposed awz criteria and speeds to human whole body vibration and exposure limits.
thresholds can potentially serve as cost-effective supplements International Journal of Pavement Engineering, 3 (4), 207–216.
for efficient PMS decision making and optimize M&R timing. Aleadelat, W., Saha, P., and Ksaibati, K., 2016. Development of serviceabil-
Although the WBV-based parameter used for pavement ser- ity prediction model for county paved roads. International Journal of
Pavement Engineering, 19 (6), 526–533.
viceability estimations offers a more accurate representation of American Association of State Highway Officials (AASHO), 1962a. Special
real travel conditions than traditional approaches, the model report 61E: the AASHO road test. Report 5: pavement research.
includes several simplifications that could be the subject of Washington, DC.
future studies. For example, the estimation of more complex American Association of State Highway Officials (AASHO), 1962b. Special
vehicle models that accounts for both translational (in all direc- report 61G: the AASHO road test. Report 7: summary report.
Washington, DC.
tions) and rotational vibration that are experienced in a real Arellana, J., et al., 2019. Multivariate analysis of user perceptions about the
vehicle ride. Additionally, while the awz was primarily focused serviceability of urban roads: case of Barranquilla. International Journal
on IRI-based models in this study, future studies should of Pavement Engineering, 1–10. doi:10.1080/10298436.2019.1577420.
12 L. FUENTES ET AL.

Arhin, S.A., Noel, E.C., and Ribbiso, A., 2015. Acceptable international Transportation Research Record: Journal of the Transportation
roughness index thresholds based on present serviceability rating. Research Board, 1806, 131–139.
Journal of Civil Engineering Research, 5 (4), 90–96. Loizos, A. and Plati, C., 2008. An alternative approach to pavement rough-
ASTM, 2015. ASTM E1926-08: standard practice for computing inter- ness evaluation. International Journal of Pavement Engineering, 9 (1),
national roughness index of roads from longitudinal profile measure- 69–78.
ments. West Conshohocken: ASTM International. Available from: Loprencipe, G. and Pantuso, A., 2017. A specified procedure for distress
www.astm.org. identification and assessment for urban road surfaces based on PCI.
ASTM, 2017. ASTM E1364-95: standard test method for measuring road Coatings, 7 (65), 26.
roughness by static level method. West Conshohocken: ASTM Loprencipe, G., Zoccali, P., and Cantisani, G., 2019. Effects of vehicular
International. Available from: www.astm.org. speed on the assessment of pavement road roughness. Applied
ASTM, 2018a. ASTM E1927-98: standard guide for conducting subjective Sciences, 9 (9), 1783.
pavement ride quality ratings. West Conshohocken: ASTM MATLAB, 2017. MATLAB, version R2017a. The MathWorks Inc.
International. Available from: www.astm.org. Available from: https://la.mathworks.com/products/matlab.html.
ASTM, 2018b. ASTM e950 / E950M - 09: standard test method for measur- Múčka, P., 2016. Current approaches to quantify the longitudinal road
ing the longitudinal profile of traveled surfaces with an accelerometer- roughness. International Journal of Pavement Engineering, 17 (8),
established inertial profiling reference. West Conshohocken: ASTM 659–679.
International. Available from: www.astm.org. Múčka, P., 2017a. International roughness index specifications around the
Cantisani, G. and Loprencipe, G., 2010. Road roughness and whole body world. Road Materials and Pavement Design, 18 (4), 929–965.
vibration: evaluation tools and comfort limits. Journal of Múčka, P., 2017b. Road roughness limit values based on measured vehicle
Transportation Engineering, 136 (9), 818–826. vibration. Journal of Infrastructure Systems, 23 (2), 04016029.
Carey, W.N. and Irick, P.E., 1960. The pavement serviceability-perform- Navidi, W.C., 2008. Statistics for engineers and scientists. New York:
ance concept. Bulletin, 250, 40–58. McGraw-Hill Higher Education.
FHWA, 2015. Status of the nation’s highways, bridges, and transit: conditions Nguyen, T., et al., 2019. Bus ride index – a refined approach to evaluating
& performance. Report to Congress. U.S. Department of Transportation. road surface irregularities. Road Materials and Pavement Design. doi:10.
Fuentes, L., et al., 2019. Modelling pavement serviceability of urban roads 1080/14680629.2019.1625806.
using deterministic and probabilistic approaches. International Journal Paterson, W.D.O., 1987. Road deterioration and maintenance effects:
of Pavement Engineering. doi:10.1080/10298436.2019.1577422. models for planning and management. Washington, DC: The World
Hosmer, D.W., Lemeshow, S., and Sturdivant, R.X., 2013. Applied logistic Bank.
regression. New York: John Wiley & Sons, Incorporated. Perera, R. and Kohn, S., 2004. Effects of variation in quarter-car simulation
ICPA, 2014. Manual de Diseño y Construcción de Pavimentos de speed on international roughness index algorithm. Transportation
Hormigón. Buenos Aires: Instituto del Cemento Portland Argentino. Research Record: Journal of the Transportation Research Board, 1889
INVIAS, 2015. Manual de Diseño de Pavimentos Asfálticos en Vías con (1), 144–151.
Medios y Altos Volúmenes de Tránsito. Bogota: Ministerio de Sayers, M.W., 1995. On the calculation of international roughness index
Transporte de la República de Colombia. from longitudinal road profile. Transportation Research Record, 1501,
ISO, 1997. ISO 2631-1: mechanical vibration and shock – evaluation of 1–12.
human exposure to whole-body vibration - part 1: general requirement. Sayers, M.W., Gillespie, T.D., and Paterson, W.D.O., 1986a. Guidelines for
Geneva: ISO. Available from: www.iso.org. conducting and calibrating road roughness measurements. Washington,
Kırbaş, U. and Karaşahin, M., 2018. Investigation of ride comfort limits on DC: The World Bank.
urban asphalt concrete pavements. International Journal of Pavement Sayers, M.W., Gillespie, T.D., and Queiroz, C.A., 1986b. The international
Engineering, 19 (10), 949–955. road roughness experiment: establishing correlation and a calibration
Kırbaş, U. and Karaşahin, M., 2019. Determination of pavement perform- standard for measurements. Washington, DC: The World Bank.
ance thresholds for comfortable riding on urban roads. Journal of SPSS, 2017. IBM SPSS statistics for windows, version 25.0. IBM
Testing and Evaluation, 47 (1), 57–77. Corporation. Available from: https://www.ibm.com/analytics/spss-
Kropác, O. and Múcka, P., 2006. Effect of obstacles on the indicators of road statistics-software.
unevenness. Proceedings of the Institution of Mechanical Engineers, Part Yu, J., Eddie, Y.J.C., and Yau, J.T., 2006. Development of speed-related
D: Journal of Automobile Engineering, 220 (10), 1333–1343. ride quality thresholds using international roughness index.
La Torre, F., Ballerini, L., and Di Volo, N., 2002. Correlation between Transportation Research Record No. 1974, TRB, Washington, DC,
longitudinal roughness and user perception in urban areas. 47–53.