International Journal of Mathematics and

Computational Science
Vol. 1, No. 1, 2015, pp. 1-4
http://www.publicscienceframework.org/journal/ijmcs

Modeling Asphalt Concrete Mix Design Using Least

Square Optimization
Saad Issa Sarsam*

Department of Civil Engineering, College of Engineering, University of Baghdad, Baghdad, Iraq

Abstract
This paper presents a computerized method for asphalt concrete job-mix formula design; the preparation of overall gradation of
aggregates for mix design is considered as an essential step in the design of asphalt concrete mixture. The target gradation should
be dense, uniform, and should not go for one extreme boundary to another or move along one of the boundaries. On the other
hand, the maximum size of aggregates (start point of the gradation curve) and the percent filler (end of the gradation curve) are
fixed for each type of asphalt concrete mix. The parabola fit using least square method was adopted in this paper; it cares for job
mix formula mathematical equation smoothing. The developed program uses parabola fit method to find all possible equations
that combines the various material gradations as per the specification requirements, the optimization process will select and print
the best six formulas (eq. 1 to eq.6) ascended according to the sum of errors squares. Decision can be made by the user, and given
to the computer to choose one of the six equations after considering the economic issue. For several months with every day
execution of the program, it proved that it is efficient and quick enough to be used in such mix design.

Keywords

Asphalt Concrete, Design, Job-Mix Formula, Least Square, Optimization

Received: December 15, 2014 / Accepted: December 31, 2014 / Published online: January 23, 2015
@ 2015 The Authors. Published by American Institute of Science. This Open Access article is under the CC BY-NC license.
http://creativecommons.org/licenses/by-nc/4.0/

1. Introduction
The asphalt concrete mix for base, binder and wearing courses, employed in previous test report calculated by traditional
methods. Another work by [2] studied the possibility of
is composed basically of aggregate, coarse, medium, and fine
implementing the stepwise optimization in the design of
in sizes, mineral filler and asphalt cement. The several mineral
asphalt concrete mix; he concluded that it is suitable for such
constituents are to be sized, uniformly graded and combined in
such proportions that the resulting blend (job mix formula) purpose with accepted accuracy. On the other hand, [3]
investigated the possibility of using the Mathematical
meets the grading requirements for the specific layer type
Modeling technique in combining the aggregates of different
under contract. Such combinations are considered as 100% by
sizes to meet the specification requirements; he concluded that
weight of total aggregate in the mix. A study presented by [1]
Modeling the aggregate combination is efficient and quick for
describes the application of general curve fitting procedure for
AASHO Road test, which can be used for linear or non-linear application in the mix design.
models (least square residuals and minimum absolute [4] Implemented computerized optimization for aggregate
residuals). He presented results that are obtained when the test gradation, various mathematical expressions have been tried
data are summarized with models that are simpler in form and for such purpose, finally it was concluded that the Modeling
more comprehensive using a computer program than those phenomena is recommended replacing the current trial and

* Corresponding author
E-mail address: saadisasarsam@uobaghdad.coeng.edu.iq
2 Saad Issa Sarsam: Modeling Asphalt Concrete Mix Design Using Least Square Optimization

error process usually adopted. grading. The process used is a linear programming technique.
[5] Presented the specification requirements for various He concluded that optimum cost mixes can be achieved by
using the obtained model to select and combine aggregates,
asphalt concrete layers and stated that the job mix formula
filler and bitumen.
should not vary from the lower limit of the specification
requirements on the sieve to the higher limit on the adjacent
sieve but should be uniformly- graded. Usually a dense 2. Methodology of Job Mix
gradation formula is employed in executing pavement of
asphalt concrete type as illustrated in tables 1.
Formula Design
Asphalt concrete mix design equation is usually found All of the gradation specification requirements for base,
manually using trial and error method. Although good results binder or wearing courses were stored in a file. The program
may be obtained, but it is certainly not the best and also first asks about the kind of specification to follow, then it gets
consumes time. Difficulties appear using computer are the it out from the file and tries to find an optimum job-mix
high process time for finding the optimum solution and the formula according to specific objective function and certain
decision factors affecting the optimality sum of which are constrains. The objective function is to find the smoothest
logical. The asphalt concrete is designed to produce a stable curve inside the specification, while constrains are:
material of maximum durability and minimum cost. Such a.) The formula curve should be inside the specification area.
requirements are specified in terms of a target grading and
b.) The formula curve should be away from both extremes of
permitted range of asphalt content, stiffness and voids.
the specifications (i.e. in the middle of the specification area if
Various sizes of aggregates (coarse, medium, fine, sand, and
possible) so that the largest range of tolerance could be
filler) are combined theoretically using trial and error method
obtained.
to get the final gradation (job mix formula) which must satisfy
the specification requirements. It was felt that using a c.) Logical constrains from economical point of view,
computer program in such analysis may give more accurate satisfying all about objective function and constrains, may
and fast results. [6] Studied the setting up of a mathematical give a job mix formula that is not economical, for example it
model to combine the aggregates in order to meet a desired require high percentages of filler material.

Table 1. SCRB (2004) specification requirements for asphalt concrete

Percent finer by weight

Sieve size (mm)
Base course A Base course B Base course C Binder course Wearing course
37.5 100-90 100 100 100 100
25 95-77 100-87 100 100 100
19 90-68 95-80 100-92 100-88 100
12.5 83-55 90-70 95-82 87-65 95-75
9.5 75-47 85-65 92-75 80-55 88-65
4.75 65-33 75-50 82-60 64-37 75-50
2.0 50-20 65-33 70-42 45-23 55-32
1.0 --------- --------- --------- 34-17 42-24
0.6 --------- --------- --------- 27-13 35-18
0.425 30-10 40-17 45-20 ------- ------
0.250 -------- ---------- --------- 20-8 25-10
0.180 22-5 25-10 28-10 ------- -------
0.125 --------- ---------- --------- 15-6 20-8
0.075 10-3 10-3 10-3 10-5 12-6

square errors of all possible combinations are executed to get

3. Mathematical Derivation of the best formula. The mathematical function is derived from
Job Mix Formula each combination of points using least square fitting. The job
mix formula is assumed as:
Least square optimization method were used to get the
optimum objective function, the possible combination of Y = A0 + A1X + A2X2 (1)
points inside the specification are used to construct the mix Where X, Y, are the possible combination points, A0, A1, A2
gradation formula, then sum of squares errors of these points are constrains derived using least square fitting method.
from calculated mathematical function like second order
equation is calculated. Then the combinations of all sums of Y = % passing each sieve; X = size (mm).
 International Journal of Mathematics and Computational Science Vol. 1, No. 1, 2015, pp. 1-4 3

The constrains are calculated using equations 2, 3, and 4 as another smooth function with a new value of ERRK. The
follows: program always maintains all information (i.e. A0, A1, A2,
ERRK and its points about the smoothest six formulas
A = (2) (formulas with least values of ERRK).
As shown in figure 1, all six-gradation formulas points with
A = (3)
mixing values are printed and plotted. Equation 1 represents
the formula with minimum ERRK, which means the
A = (4)
best-derived formulas; equation 2 formula has largest ERRK
Where: w0 = Ʃ Yi; w1 = Ʃ Yi Xi; w2 = Ʃ Yi Xi2; w3 = Ʃ Yi Xi3; and so on.
w4 = Ʃ Yi Xi4; To satisfy the third constrain given in this paper, which is a
v =∑ I = n; v = ∑ Xi; v2 = Ʃ Xi;v3 = Ʃ Xi2; logical constrain, one can choose any one of the six formula,
for example equation 6 with 7% filler because equation 1
v4 = Ʃ Xi3; n = no. of points formula assumes 9% filler, which is not economical. From the
After finding the function y = f(x) from (n) points (x1, y1) ……. graph in figure 1 it is obvious that there is no big difference in
(x n, y n) [note that X i represents particle size (mm) and Y smoothing between equation 1 curve and equation 6 curve.
represents percentage passing], summation of error squares is Table 2 shows the grain size distribution for each of the six
calculated using equation5 and 6. equations selected by the least square optimization procedure
ERRK = ∑ f Xi − Yi$ (5) and exhibited as an output. On the other hand, Table 3
demonstrates the percentages of each type of mineral
ERRK = ∑ A + A Xi + A Xi − Yi$ (6) aggregates required for each equation as supplied by the
A0, A1, A2, ERRK and all points (x1, y1) ……. (x n, y n) are software.
kept for later stage in the program. Again, the program follows
all above calculations but for new set of points to find also
Table 2. Grain size distribution output for each equation

Percent finer by weight

Sieve size (mm)
Equation 1 Equation 2 Equation 3 Equation 4 Equation 5 Equation 6
25 90.0 90.0 90.0 92.5 92.5 92.5
19 82.3 82.3 82.3 84.47 84.47 94.47
12.5 76.14 76.14 76.14 77.06 77.06 77.06
9.5 70.07 70.07 70.07 70.11 70.11 70.11
4.75 56.39 56.38 56.37 56.39 56.38 56.37
2.0 44.45 44.30 44.15 44.45 44.30 44.15
0.425 24.62 23.99 23.36 24.62 23.99 23.36
0.18 14.00 13.12 12.25 14.00 13.12 12.25
0.075 9.35 8.55 7.74 9.35 8.55 7.74

Table 3. Type and percentage of materials required for each equation

% of material Equation 1 Equation 2 Equation 3 Equation 4 Equation 5 Equation 6

Coarse aggregate 20 20 20 15 15 15
Medium aggregate 5.0 5.0 5.0 10 10 10
Fine aggregate 25 25 25 25 25 25
Sand 41 42 43 41 42 43
Mineral Filler 9.0 8.0 7.0 9.0 8.0 7.0

Figure 2 illustrates the final setup of the selected equations and software, the gradation of asphalt concrete constituent
materials are presented; the percentages of each type of
the design gradation. It shows that the tolerance of the J.M.F.
aggregate, which provide the optimized job mix formula, are
is sited inside and within the specification requirements. On
also shown.
the other hand, Table 4 demonstrates the final output of the
4 Saad Issa Sarsam: Modeling Asphalt Concrete Mix Design Using Least Square Optimization

Fig. 2. Final Job mix formula design for Asphalt concrete Base course
Fig. 1. Best six job mix formulas using Least Square Optimization method.

Table 4. Final output showing the optimized job mix formula

Percent finer by weight

Sieve size (mm) Mineral Aggregates Material types Optimized Formula J.M.F. Tolerance Specifications
Coarse Medium Fine Sand Filler J. M. F. Upper Lower Max. Min
37.5 100 100 100 100 100 100 100 100 100 100
25 50.2 100 100 100 100 90.0 87.0 96.0 87.0 100
19 20.5 64.0 100 100 100 82.3 80.0 88.3 80.0 95.0
12.5 0.9 19.2 100 100 100 76.1 70.1 82.1 70.0 90.0
9.5 0.0 0.8 80.1 100 100 70.1 65.0 76.1 65.0 85.0
4.75 0.0 0.0 27.2 99.0 100 56.4 50.4 62.4 50.0 75.0
2.0 0.0 0.0 2.4 85.0 100 44.2 40.2 48.2 33.0 65.0
0.420 0.0 0.0 1.0 37.0 100 23.4 19.4 27.4 17.0 40.0
0.180 0.0 0.0 0.2 12.2 100 12.2 10.0 16.2 10.0 25.0
0.075 0.0 0.0 0.2 4.2 84.8 7.7 6.2 9.2 3.0 10.0
Percentages 20.0 5.0 25.0 43.0 7.0

4. Conclusions
References
Based on the limited investigation, the following conclusions
[1] Irick P. (1963) “Application of general curve fitting procedure
may be drawn: to AASHO road test data” HRB, Highway research record No.
1. The least square optimization method for asphalt concrete 42, USA.
mix design produces exactly the optimum objective function. [2] Sarsam S. (2013) “Stepwise Optimization Modeling of Asphalt
Concrete Mix Design” Research Journal of Computation and
2. It provides the best six job mix formula after taking the Mathematics, RJCM 2013, 1(2): P (28-32); Sciknow
specifications constrains into consideration. Publications Ltd. USA.

3. It gives suggestion about the best job mix equation through [3] Sarsam S. (1997) “The design of asphaltic concrete mix using a
mathematical model” Journal of Engineering and Technology
optimization process, and allows the user to consider
Vol. 16 No. 3, University of technology, Baghdad, Iraq.
economic issues in deciding the final mix design equation.
[4] Sarsam S. (2012) “Computerized Optimization of Asphalt
4. The software output provides a summary of the tried Concrete Mix Design” Proceedings, International Conference
equations and demonstrates the final optimized job mix on Management Sciences and Information Technology (MSIT
2012) Changsha, China, July 1-2, 2012
formula with the required tolerance.
[5] SCRB (2004) “Standard specification for roads and bridges”
5. For several months and every day executing the program
Ministry of Housing and construction, Iraq
many times, it was proved that the least square optimization
method is efficient and quick enough for five aggregate [6] Yeaman J. (1970) “The design of asphaltic concrete using
mathematical models and linear programming techniques”
material sizes mixing, and is recommended to be used always. Proceeding, ARRB – Vol.5 Part 5 Australia.