Historical and Current Rheological Binder

Characterization vs. Binder Performance

Dr. David A. Anderson

Consultant
Dr. Geoffrey M. Rowe
ABATECH INC.
Dr. Donald Christensen
Advanced Asphalt Technologies, LLC

Petersen Conference
July, 2008

Why Characterize Binder Rheology

With Mathematical Models?
F Provide mathematical formula that can be
manipulated for purpose of calculation
F Provide quantitative parameters that can
be used to characterize rheology and
changes in rheology
F Provide rational basis for specification
criteria
F Provide link between binder and mixture

1
Models – What Is Needed?
F Needs vary with application
F From binder perspective binder need model that:
– Captures changes in binder rheology caused by long-
long-
term aging
Must be compatible with molecular/structure
changes at molecular level
F Links binder and mixture properties
F Is related to binder “quality” – a forgotten issue
F Many possibilities – focus on CA model

Explicit Models – Pre SHRP

F Jongepier and Kuilman
– Relaxation spectra as log normal distribution
F Dobson (1969)
– Based on empirical relationships between
modulus and phase angle
F Dickenson and DeWitt (1974)
– Based on hyperbolic representation
– Recognized relaxation spectra skewed

2
Models – Christensen-Anderson
F Developed during SHRP
F Genesis was the need to describe
relaxation modulus
– All other rheological functions can be
generated from relaxation modulus
F Christensen recognized that relaxation
modulus is skewed, not symmetric
– Concluded that skewed function was needed

CA Model

F Christensen-
Anderson - CA
model (1993)
– Relates G*(ω)
to Gg, ωc and R

3
Weibul Function to Model Relaxation
Spectrum
−( m +1)
m ⎧ x − a ⎫ ⎡ ⎧ x − a ⎫⎤
F ( x) = exp ⎨ ⎬ ⎢1 + ⎨ ⎬⎥
b ⎩ b ⎭ ⎣ ⎩ b ⎭⎦
F(x) = Probability density function
m = Skewness parameter
x = Independent parameter
b = Scale parameter
a = Location parameter

Cumulative Weibull Function

F Integrate to obtain cumulative function:
−m
⎡ ⎧ x − a ⎫b ⎤
P ( x) = 1 − ⎢1 − ⎨ ⎬ ⎥
⎢ ⎩ b ⎭ ⎥
⎣ ⎦
F or: −m
⎡ ⎧ x − a ⎫b ⎤
1 − P( x) = ⎢1 + ⎨ ⎬ ⎥
⎢ ⎩ b ⎭ ⎥
⎣ ⎦

4
CA Model for G*(ω)
F Substituting rheological parameters:
− R / log 2
⎡ ⎧ ω ⎫(log 2 / R ) ⎤
G * (ω ) = G g ⎢1 + ⎨ ⎬ ⎥
⎢ ⎩ ωc ⎭ ⎥
⎣ ⎦
G*(ω) = Measured complex modulus
Gg = Glassy modulus
R = Rhelogical Index (shape factor)
ω = Test frequency
ωc = Crossover frequency (location parameter)

CA Model for δ(ω)

F Rewriting and substituting rheological
parameters:

⎡ ⎧ ω ⎫(log 2) / R ⎤
δ (ω ) = 90 / ⎢1 + ⎨ ⎬ ⎥
⎢ ⎩ ωc ⎭ ⎥
⎣ ⎦
δ(ω) = Measured phase angle

5
Temperature Dependency
F WLF based on free volume concepts
– Good results above Tg
F Arrhenius based on rate theory
– Necessary below glass transition
temperature
F Polynomial
– Useful over small temperature range only
– Works well with BBR data

R and Delayed Elastic Response

1 / G * (ω ) = 1 / GE + 1 / GDE (ω ) + 1 / GV (ω )

1 / GDE (ω ) = 1 / G * (ω ) − 1 / GE − 1 / GV (ω )

Measured 1 / G g 1 /(ωηo )
at some T
and ω

6
Estimate for Rheological Index

F Model can be rewritten as:

log G * (ω ) = log G g + ( R / log 2) log[1 − δ (ω ) / 90]
F G*(ω) versus log[1- δ(ω) / 90)]
– slope R/log2
– Can be estimated from a single DSR measurement
F Relationship works well when:
10°< δ(ω) < 70°

Estimation of ηo and ωc
F Similar shortcuts
F Full mastercurve is not needed to
estimate model [parameters
F Useful in following aging studies

7
Other Models – van der Poel
F Predicts stiffness from Pen and vis
F Recognized hyperbolic relationship S vs.T
F Model implicit in
development of
nomograph
– Viscous asymptote
– Elastic asymptote

Mix models - Witczak

F Basic symmetric
1
sigmoid function y=
(1 + e − x )
F Basis of Witczak
model for asphalt
mixture E* data

8
 Witczak model α
log( E *) = δ +
1 + e β +γ (log tr )

F Two asymptotes
and the
central/inflection
point of the
sigmoid
F Model is limited
in shape to a
symmetrical
sigmoid

Temperature Susceptibility Parameters

F PIR&B (Pfeiffer and van Doormaal, 1936

– Ring and ball and pen
F PIlogPen (Huekelom and Klomp, 1964)
– Slope of log pen vs temperature
F PVN (McLeod, 1972)
– Pen at 25°C and viscosity at 60 or 135 60°C
F VTS (Puzinauskas)
– Viscosity at 60 and 135 60°C

9
Comments - Temperature Susceptibility
Parameters
F Based on measurements at two
temperatures
F Shear rates not same at multiple
temperatures
F Confound time and temperature effects
F Studies show that different indices are not
equivalent and vary differently with aging
– Indices reflect time and temperature
susceptibility

Temperature Susceptibility
F In old specifications captures via PI, PVN
etc
F How do we capture this in the newer
specifications/testing
F Example – five data sets from Lamont
Road

10
BBR data using CAM Model
F Analysis of 5 data examples from Lamont
test road
F Master curve construction using AASHTO
PP42
– CAM model

BBR data

F Five sections
considered
F Typical
example of
data
F Used all data
available

11
 Master Curve Construction

F Used
AASHTO
PP42 to
construct
master
curves

Shift factors
F Shift factors
obtained in
Arrhenius
form

12
 Temperature susceptibility
PEN = 0.0122 a1 - 248.7, R2 = 0.2825
PI = 0.0004 a1 - 12.783, R2 = 0.3546
PVN(135) = 0.0002 a1 - 6.7761, R2 = 0.6506
F The temperature PVN(60) = 0.0003 a1 - 8.5306, R2 = 0.7366140
susceptibility is 1
captured in the PVN60
slope of the shift 0.5 120
PVN135 PVN60

PEN at 25C, mm/10

factor PI
PVN135
F The better r2 0
PEN
100
results from the
PVN or PI
data obtained to -0.5 80
compute the PEN
empirical -1 60
parameter closest
to the rheological -1.5 40
measurements PI
-2 20

-2.5 0
25.5 26.0 26.5 27.0 27.5 28.0 28.5 29.0
Arrhenius gradient a1 (X.001)

Temperature intervals
F Similar to US absolute grade (PG)
F Used in Europe – particularly UK

13
PI can be related to temp interval

109

Modified Type 1
Modified Type 2
107 Unmodified

2MPa
G* (Pa)

105

2kPa
103

101
-10 0 10 20 30 40 50 60 70 80 90 100

Temperature (oC)

Penetration Index from DSR Data

2.50
2.00
1.50
Penetration Index

1.00
0.50
0.00
-0.50
-1.00
-1.50
-2.00
25 30 35 40 45 50 55
T @ 2kPa – T @ 2MPa oC

14
Model Parameter Changes with Aging

Model Parameter Changes with Aging

15
R and ωc May Both Change
F Changes in R reflect changes in
relaxation Modulus
– Shape of mastercurve
F Changes in Infinite number of
combinations ωc reflect hardening
– Location of mastercurve

Closing Comments
F Many different models are available for
characterizing rheological properties of binders
– Model selected depends on application
F Skewed function is needed to model
mastercurve
F Shortcut methods can be used to generate CA
model parameters
F Temperature Susceptibility parameters can be
related to CA model parameters
F R and or ωc can change with aging

16
Acknowledgements
Concepts described in this presentation are
not necessarily original but are borrowed
from the work of others that has appeared in
the literature over the course of the past 80
or more years. Acknowledgement is given to
those who preceded us, names such as
Puzinauskas, McLeod, van der Poel,
Hukelom and many others

17