Bai 1994

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

KNOWLEDGE-BASED EXPERT SYSTEM FOR

CONCRETE MIX DESIGN

By Yong Bai, ~ and Serji N. Amirkhanian,z Member, ASCE

ABSTRACT: In the United States, the American Concrete Institute method (ACI
211) is the most widely used concrete mix design. However, ACI 211 does not
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

consider all factors regarding concrete mix design. The acceptance of a freshly
mixed concrete depends on concrete quality control test results. If the results
indicate that the concrete does not meet the specification, mix design adjustment
must be made. Concrete mix design and adjustment are somewhat complicated
and time-consumingtasks that are performed best by experienced persons. In this
project, a rule-based expert system was developed to assist the user in concrete
mix design, including the necessary adjustments to the design. The expert system
can handle mix designs for normal-weight concrete, normal-weight-massconcrete,
normal-weightno-slump concrete, heavyweight concrete, and lightweightconcrete.
Besides workability, consistency, strength, durability, and density, the expert sys-
tem considersother criteria such as admixtures, transportation, and air temperature
that affect the concrete mix design. The system was tested on three projects. The
system's selection of concrete proportions compared favorably with the experts'
selections.

INTRODUCTION

Concrete mix design is the process of selecting the p r o p o r t i o n s of a con-


crete mix ( " A C I 116R-85" 1989). It involves satisfying a balance between
economics and specifications. The required characteristics, such as worka-
bility, consistency, strength, durability, and density, are g o v e r n e d by the
use to which the concrete will be put and by conditions expected to be
encountered at the time of placement. These are often, but not always,
reflected in concrete mix design specifications.
Generally, in the U n i t e d States, there are four m e t h o d s used for concrete
mix design: (1) Trial and error based on previous concrete mix designs; (2)
American Concrete Institute ( A C I ) m e t h o d ; (3) Portland C e m e n t Associ-
ation method; and (4) National Crushed Stone Association method. The
most widely used m e t h o d of a concrete mix is the A C I m e t h o d , given in
ACIManua1211 ('.'ACI 211.1-81" 1989). H o w e v e r , A C I 211 does not answer
all the questions regarding concrete m i x design. F o r instance, some criteria
such as admixtures, transportation, and t e m p e r a t u r e affect the concrete
proportions. The designers must use their experiences and j u d g m e n t to deal
with these issues.
Since many factors affect concrete mix design, it is almost impossible to
make the design right the first time. A d j u s t m e n t is always p e r f o r m e d based
on the results of concrete quality control tests. A control test is one that
gives results soon enough for the concrete designer to take corrective action
if results indicate that concrete is outside the specified limits ( " C o n t r o l "
1981). In general, control tests includes measuring the slump, air content,
unit weight, t e m p e r a t u r e , and compressive strength. E x c e p t for the corn-

~Grad. Student, Civ. Engrg. Dept., Clemson Univ., Clemson, SC 29634-0911.


2Assoc. Prof., Civ. Engrg. Dept., Clemson Univ., Clemson, SC.
Note. Discussion open until November 1, 1994. To extend the closing date one
month, a written request must be filed with the ASCE Manager of Journals. The
manuscript for this paper was submitted for review and possible publication on May
11, 1992. This paper is part of the Journal of Construction Engineering and Manage-
ment, Vol. 120, No. 2, June, 1994. 9 ISSN 0733-9364/94/0002-0357/$2.00 +
$.25 per page. Paper No. 4034.
357

J. Constr. Eng. Manage. 1994.120:357-373.


pressive strength test, these tests are conducted while the concrete is still
in the plastic state.
Concrete mix design and adjustments made to the mix design involve
complicated issues, and the correct ways to perform them can be achieved
with expert advice and experience. However, experts are not always avail-
able, nor do they always have the time to consult all possible references,
review available data, and so on. In some companies, they do not have
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

personnel with the experience to make the necessary decisions regarding


concrete mix design and the necessary adjustments. Therefore, development
of an expert system for concrete mix design, including the adjustments,
would be a valuable tool.
There have been a few expert systems developed for concrete mix design
(Celik et al. 1989; Clifton 1988; Malasri and Maldonado 1988); however,
most of these systems can handle only normal-weight concrete and do not
consider criteria such as admixtures, transportation, and air temperature
that can affect the concrete proportions. Also, they have no function to
perform the necessary adjustments to a mix design. After interviewing sev-
eral experts in the concrete industry who had some experience with com-
mercially available concrete mix design programs (e.g., CANMIX, SeeMIX,
etc.), it was determined that the development of an expert system is a
necessary and timely topic.

RESEARCH SCOPE AND METHODOLOGY


The scope of the research was to develop an expert system to help the
user in concrete mix design and adjustments. The developed expert system
can handle mix designs for normal-weight, normal-weight-mass, normal-
weight no-slump, heavyweight, and lightweight concrete. In this project,
only the most widely used admixtures are considered including: air-entrain-
ing, retarding, accelerating, water-reducing, water-reducing and retarding,
water-reducing and accelerating, high range water-reducing, high range water-
reducing and retarding, and fly ash admixtures.
The expert system development package for this research is EXSYS
Professional, a ruled-based expert system shell (EXSYS Inc. 1988). The
concrete mix design that is determined, by the expert system, is presented
on a spreadsheet (i.e., Q U A T T R O PRO).
The knowledge incorporated into the expert system is taken from tech-
nical literature (e.g., ACI manuals, ASTM standards, etc.), interviews with
experts, and a survey of several experts. Of the three experts who were
interviewed, two worked in ready-mixed concrete companies and one worked
in an engineering and design company. The survey questionnaires were
mailed to five other experts. These people were chosen because of their
knowledge of and their experience with concrete mix design and adjust-
ments.

KNOWLEDGE ACQUISITION
There were three stages in the knowledge-acquisition process: (1) Collect
knowledge from the literature; (2) interview experts; and (3) survey experts.

Knowledge Collection from Literature


Some knowledge incorporated into the expert system was obtained from
technical literature. Most was chosen from ACI manuals. The ACI manuals
were selected because the facts they contain represent a consensus of the
358

J. Constr. Eng. Manage. 1994.120:357-373.


knowledge and opinions of leaders in the concrete community. The ACI
manuals also give a sequence of logical, straightforward steps for concrete
mix design and necessary adjustments.

Interviews
The purpose of interviewing the experts was to collect and analyze in-
formation from one or more domain experts to formulate an expert system
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

knowledge base. Structured and unstructured interviews were performed in


this project.

Unstructured Interviews
Three unstructured interviews were conducted with one civil engineer
and two concrete quality control experts. In these interviews, experts were
approached with a series of questions regarding concrete mix design and
the necessary adjustments. They were asked to describe their knowledge
about the selection of concrete proportions, diagnosing the problems in mix
design and the adjustments, and their solutions.
Through the unstructured interviews, several conclusions were reached
including:

1. The ACI method for concrete mix design is the most widely used in
the concrete production industry.
2. Air temperature, transportation, and admixtures affect the concrete
proportions.
3. Rules-of-thumb, past experience, and intuition are very important in
concrete mix design and the necessary adjustments made to the mixture.
4. There are other commercially available programs for concrete mix
design; however, due to characteristics of an expert system, there is a need
to implement this type of technology in the concrete production industry.
5. The knowledge dealing with concrete mix design is available.

Structured Interviews
The objectives of the structured interviews were to solve several special
problems. Through the literature review and unstructured interviews, some
of the unsolved problems were clearly outlined. The structured interviews
focused on the following problems:

1. How do transportation, air temperature, and admixtures affect the


concrete proportions?
2. How does one perform concrete mix design adjustments?

The concrete slump will sometimes be lost during transportation, partic-


ularly for long hauling distances. Air temperature is another factor that
affects the concrete slump. High temperature causes concrete slump loss.
Table 1 indicates the relationship among air temperature, delivery time,
and concrete slump loss. For example, if the concrete mix design specifi-
cation requires 10.16 cm (4 in.) slump, the delivery time is one hour, and
if the air temperature is around 21. I~ (70~ the designer should make a
mix design that produces a 12.7 cm (5 in.) slump.
Transportation and air temperature also affect other properties of con-
crete such as air content. However, the experts indicated that these are not
critical factors. The designer does not need to consider these at the trial
359

J. Constr. Eng. Manage. 1994.120:357-373.


TABLE 1. Slump Loss (cm) due to Air Temperature and Delivery Time
Air Temperature
Delivery time Around 10~ Around 21.1 ~ Around 32.2~
(hour) (50~ (70~ (90~
(1) (2) (3) (4)
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

0.0 0.0 0.0 0.0


0.5 0.0 1.27 3.81
1.0 0.0 2.54 7.62
1.5 0.0 3.81 10.16

batching mix design stage. However, if the control tests indicate that the
concrete does not meet the specification, then adjustment of the mix design
must be performed.
Problems regarding admixtures and concrete mix design adjustment are
complicated. At the structured interview stage, these problems remained
unsolved. The writers realized that further research was needed.

Survey
The survey questionnaire, containing 29 multiple-choice questions, fo-
cused on two types of problems: concrete admixtures and mix design ad-
justment. The following is a summary of the survey findings.

Survey Findings on Admixtures


In general, most experts agreed that water-reducing admixtures (ASTM
C494 type A), water-reducing and retarding admixtures (ASTM C494 type
D), and water-reducing and accelerating admixtures (ASTM C494 type E)
can reduce water content about 7%. High range water-reducing admixtures
(ASTM C494 type F) and high range water-reducing and retarding admix-
tures (ASTM C494 type G) can reduce water content about 14%. According
to the experts' opinions, retarding admixtures (ASTM C494 type B) and
accelerating admixtures (ASTM C494 type C) do not affect the concrete
proportions, or their affect is so small that it can be neglected.
Experts also indicated that fly ash Class C and Class F can be used in
normal-weight, heavyweight, and lightweight concrete to replace the cement
from 15% to 30% by weight.
Regarding the air-entraining admixtures, the experts suggested that if the
concrete would be exposed to freezing and thawing cycles, moisture, free
water, deicing agents, and other aggressive chemicals, the use of air-en-
training admixtures becomes necessary.

Survey Findings on Slump Adjustment


The slump of concrete is influenced by many factors, the primary factor
being water content. Experts indicated that if the slump of the concrete is
incorrect, then the amount of water must be increased or decreased by 3.64
kg/0.76 m 3 (8 lb/cu yd) of concrete for each required increase or decrease
of 2.54 cm (1 in.) in slump. In general, when increasing the water content,
the cement content is also increased to keep the water-cement ratio constant.
When decreasing the water content, the cement content usually remains
unchanged. Also, adjustment should be made in fine aggregate weight as
necessary to maintain the required total effective displaced volume.
ACI 211.1-81 recommends that if the slump of the batch was incorrect,
360

J. Constr. Eng. Manage. 1994.120:357-373.


then increase or decrease the amount of water by 4.55 kg/0.76 m 3 (10 lb/cu
yd) of concrete for each required increase or decrease of 2.54 cm (~ in.) in
slump ("ACI 211.1-81" 1989). The amount of water required to increase
or decrease the slump differs between the experts' recommendations and
the ACI method. The experts explained that the ACI method is quite
general. Many factors affect the required water such as cement character-
istics, aggregate(s) type, and aggregate(s) grading. According to the experts'
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

opinions, wherever possible, the value obtained by observation and expe-


rience with particular materials should be used.

Survey Findings on Air Content Adjustment


There are several methods used to adjust the air content in a concrete
batch including changing the aggregate(s) grading, cement type, and the
concrete slump. Experts recommended that if the desired air content was
not achieved, the amount of air-entraining admixture must be increased or
decreased by 30 mL (1 oz) for each required increase or decrease of 1% in
air content, respectively. An increase or decrease in air content will be
accompanied by an increase or decrease in slump unless water is reduced
or increased to compensate.
For each percent increase or decrease in air content, water should be
decreased or increased by approximately 1.82 kg/0.76 m 3 (4 lb/cu yd). An
increase in water may be accompanied by a decrease in strength unless
compensated for by additional cement to keep water-cement ratio constant.
Also, adjustment should be made in fine aggregate weight as necessary to
maintain required total effective displaced volume.

Survey Findings on Temperature Adjustment


Monitoring the freshly mixed concrete temperature is necessary because
temperature change affects the concrete slump, the mixing water require-
ment, the air content, and the strength. Adjustment should be made if
concrete temperature is too high or too low. For high temperature [usually
above 29.4~ (85~ experts indicated that using ice as part of the mixing
water is most effective in reducing concrete temperature. Table 2 illustrates
the relationship among the amount of ice required in kilograms per cubic
meters (pound per cubic yard) of concrete, mixing water temperature (~
and reduction in concrete temperature (~ For instance, if there is a need
to reduce concrete temperature about 8.33~ (15~ and the mixing water
temperature is 26.7~ (80~ then, 38.63 kg (85 tb) of ice per 0.76 m 3 (1
cu yd) of concrete should be used in place of water.

TABLE 2. Relationship among Ice Required (kg) per 0.76 m 3 (1 cu yd) of Concrete,
Mixing Water Temperature, and Reduction in Concrete Temperature

Mixing water Reduction in Concrete Tern )erature


temperature ~ 5,55~ 8.33~ 11.11~
(~ (IO~ (15~ (20~
(I) (2) (3) (4)
15,56 (60) 31.78 45.4 61.29
21.11 (70) 29.51 40.86 56.75
26.67 (80) 27.24 38.59 52.21
32.33 (90) 24.97 36.32 49.94
37.78 (100) 22.70 34.05 47.76

361

J. Constr. Eng. Manage. 1994.120:357-373.


TABLE 3. Relationship among Mixing Water Temperature (~ Required Concrete
Temperature, and Average Materials' Temperature

Average materials' Required Concrete Temperature


temperature ~ 12.78~ 15.56~ 18.33~
(~ (55~ (60~ (65~
O) (2) (3) (4)
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

1.67 (35) 51.67 65.56 76.67


4.44 (40) 43.33 54.44 68.33
7.22 (45) 32.22 43.33 57.22
10 (50) 23.89 35.00 48.89

For low temperatures [usually below 10~ (50~ the adjustment meth-
ods include using hot water to mix concrete and heating aggregate before
mixing. When aggregates are flee of ice and frozen lumps, the desired
temperature of the concrete can usually be obtained by heating only the
mixing water. The temperature of the mixing water can be adjusted readily,
but the water should be available at a consistent, regulated temperature
and in sufficient quantity to avoid appreciable fluctuations in temperature.
Table 3 indicates the relationship among mixing water temperature (~
required concrete temperature (~ and average materials' temperature
(~ The average materials' temperature is the average temperature of
cement, water, and aggregates. For instance, if the specification requires a
freshly mixed concrete temperature of 15.6~ (60~ but the average ma-
terials' temperature is 4.44~ (40~ then the mixing water temperature
should be 54.44~ (130~ When the air temperature is consistently below
0~ (32~ it is usually necessary to heat the aggregates.

Survey Findings on Compressive Strength Adjustment


Many factors can affect the strength of a concrete batch. In general,
differences in strength can be traced to two fundamentally different sources
("ACI 214.1R-81" 1989): differences in strength-producing properties of
the concrete mixture and ingredients, and apparent differences in strength
caused by variations inherent in the testing.
In engineering practice, the strength of a concrete batch at a given age
and cured at a prescribed temperature is assumed to depend primarily on
two factors (Neville 1981): the water-cement ratio and the degree of com-
paction. If only the fully compacted concrete is considered, the strength of
concrete is governed by the water-cement ratio. If the compressive strength
at 28 days is lower than the requirement, the water-cement ratio must be
decreased or the cement content must be increased to increase the com-
pressive strength. Increasing the 28-day compressive strength of a concrete
by 689.5 kPa (100 psi) necessitates the addition of approximately 4.54 kg
(10 lb) of extra cement to the mix. The yield can be maintained by an
appropriate reduction of fine aggregate content.

Survey Findings on Unit Weight Adjustment


The unit weight test of freshly mixed concrete is used to calculate yield,
relative yield, and cement content. The relative yield is the actual yield
divided by the designed size of the batch and it is very important for both
the concrete producer and the contractor. If a relative yield is smaller than
1.0, it indicates that the actual batches are smaller than the designed batch
362

J. Constr. Eng. Manage. 1994.120:357-373.


size. In this case, the plant produces less concrete than the mix design calls
for, and the contractor should purchase more concrete to fill the forms. If
a relative yield is greater than 1.0, it indicates that the plant is over yielding
and may result in a cement content that is lower than specified. Relative
yield should be between 0.993 and 1.007 for good control (Waddell 1984).
If the slump, air content, temperature, and compressive strength all satisfy
the requirements, and only the unit weight differs significantly from the
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

value assumed initially for the calculations of proportioning, the unit weight
can be adjusted by increasing or decreasing the amount of fine aggregate.

Summary of Knowledge Acquisition


During knowledge acquisition, it was found that, in some cases, there
were different answers for the same problems. If this situation occurred,
the writers would ask some experts whether they had any comments about
other experts' opinions. In most cases, after discussing other experts' views,
the experts changed their answers, and common conclusions were reached.
However, some opinions remained unchanged. For instance, some experts
said that water-reducing admixtures (ASTM C494 type A) can reduce water
content about 7%, while others indicated 8%. In this case, however, experts
agreed that there is no significant difference between 7% and 8% because
other factors, such as cement characteristics and aggregate(s) grading, affect
the performance of water-reducing admixtures.
Knowledge acquisition was found to be the most difficult, laborious, and
time-consuming part of developing the expert system. One important thing
that the writers learned from knowledge acquisition is that a long-term
commitment from experts is essential. The commitment of the experts de-
pends on the experts' motivation for participation. The experts' motivation
for the project can be increased by narrowing the objectives of the research,
asking intelligent questions, and providing feedback.
Without expert help, it is almost impossible to develop a useful expert
system. Some expert systems have been developed from knowledge acquired
exclusively from texts and other published works. These systems have lim-
ited usefulness. The functionality of expert systems is derived in part from
their ability to capture and represent the experts' knowledge. In a vaguely
documented field such as concrete mix design, the knowledge cannot be
solely obtained from technical literature.

DEVELOPING THE SYSTEM


Rules Developed from Flowcharts
Several flowcharts were developed for the expert systems. These flow-
charts were used to develop the expert system rules. For instance, the
following rules (rules No. 1, No. 2, and No. 3) were created from the normal-
weight concrete mix design flowchart (Fig. 1).

Rule No. 1
IF: (1) The maximum slump required by specification is 12.7 cm AND
(2) Delivery time (the time needed to transport concrete from ready-
mix concrete plant to project site) is 0.5 h AND (3) Air temperature
is around 21.1~ (70~ THEN: [SLUMP] IS GIVEN THE VALUE
13.97 cm
Rule No. 2
IF: (1) [SLUMP] = 12.7 cm OR [SLUMP] = 15.24 cm AND (2) The
363

J. Constr. Eng. Manage. 1994.120:357-373.


START. /
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

I Asktypeofflyash, I

1
I Askpercentofflyash.

I Askmax.sizeofcoarseI_
aggregate. I

Askfinenessmodulusof I
fineaggregate.

1
I Selectvolumeofcoarse
]
aggregate.

FIG. 1. NormalWeight Concrete Mix Design Flowchart


nominal maximum size of coarse aggregate is 2.54 cm AND (3) Air
entraining is required by specification AND (4) The concrete exposure
condition is severe AND (5) Water-reducing admixture is ASTM C494
type F (water-reducing, high range admixture) OR (6) Water-reducing
admixture is ASTM C494 type G (water-reducing, high range, and
retarding admixture) THEN: (1) lAIR CONTENT] IS GIVEN THE
VALUE 6 AND (2) SS_WR(YBAI001.WK1, E27, [AIR CON-
TENT]) AND (3) ]MIXING WATER] IS GIVEN THE VALUE
121.22 kg AND (4) SS_WR(YBAI001.WK1, E30, [MIXING WATER])
Rule No. 3
IF: (1) [CLASS C] > = 0 AND (2) [CLASS F] > = 0 AND (3) [CRS
AGG] > 0 AND (4) [W/C RATIO] > 0 AND (5) [AIR CONTENT]
> 0 AND (6) [MIXING WATER] > 0 THEN: RUN NORMAL
WEIGHT CONCRETE MIX DESIGN SPREADSHEET Confi-
dence = 10/10

364

J. Constr. Eng. Manage. 1994.120:357-373.


No
n.I
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

deliv Askdellvepjtime.

I Askairtemperature. I Askairtemperature. I

1
I Selectthefinalslump. I I Selectthefinalslump.[

IYeS 9

I Ask~oncr=~x~,,rsI
I 0ond~,on. I

"~1 reducing
Asktypeadmi
ofwater
xture.
1 B

Selecttheamountof I
mixingwater.
I

FIG. 1. (Continued)

These rules are used to determine the concrete slump, the percent air
content, and the amount of mixing water. The command "SS_WR" allows
the expert system to transfer the data directly to a spreadsheet~

ConfidenceFactorSystems
Confidence factors are used in an expert system to account for varying
level of certainty. For the developed system, the confidence factors were
entered into the conclusions of rules in the knowledge base. The developed
system uses a confidence factor system of 0-10. In this system, 0 indicates
absolutely not and 10 indicates absolutely certain. The values of 1 - 9 rep-
resent degrees of confidence ranging from "very probably no" to "very
365
J. Constr. Eng. Manage. 1994.120:357-373.
I Selesttheamountof~ir I
contest. 1
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

Ask28-daycornp~ess~ve
strength. I

SelectW/CfromAC1211
Table5.3.4.a). I

T~le 5.3.4.(b)). I Selectthesm~JlerW/C.


I

I Selectthesm~lerW/C.
I- I Table5.3.4Ja)Isthefl~ I
I- W/Cv~ue. ]

1
\
FIG. 1. (Continued)

probably yes." For instance, rule No. 3 (mentioned before) shows that 10
is assigned to the confidence factor indicating that the conclusion is drawn
with 100% confidence. This is a subjective, not a statistical, value. These
confidence factors were chosen by the experts interviewed.

Developing Spreadsheets
The Quattro Pro spreadsheet was used to calculate the amount of concrete
ingredients. Some parts of the required data were transferred automatically
from the expert system, others were input directly into the spreadsheet.
Tables 4 and 5 indicate the required input data for mass concrete and normal,
heavy, light, and no-slump concrete, respectively. If all required data are
known, the spreadsheet will calculate the amount of concrete ingredients
automatically.

System Limitations
The best approach to perform a concrete mix design is to use proportions
previously established for similar concrete using the same materials. Where
such prior information is limited or unavailable, the developed expert system
366

J. Constr. Eng. Manage. 1994.120:357-373.


TABLE 4. Required Input Data for Mass Concrete
Specific
Items gravity Total moisture Absorption percent
(1) (2) (3) (4)
Cement Required Not required Not required
Fly ash (if used) Required Not required Not required
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

Fine aggregate Required Required Required


Coarse Aggregate
15.24 to 7.62 cm sieve Required Required Required
7.62 to 3.81 cm sieve Required Required Required
3.81 to 1.91 cm sieve Required Required Required
1.91 cm to 4.76 mm sieve Required Required Required

TABLE 5. Required Input Data for Normal, Heavy, Light, and No-Slump Concrete
Normal, heavy, and
Required input data no-slump concrete Light concrete
(1) (2) (3)
Cement specific gravity Required Required
Fly ash specific gravity (if used) Required Required
Batch size Required Required
Coarse aggregate specific gravity Required Required
Coarse aggregate total moisture Required Required
Coarse aggregate absorption Required Required
Coarse aggregate dry-rodded
unit weight Required Not required
Coarse aggregate dry-loose unit
weight Not required Required
Fine aggregate specific gravity Required Required
Fine aggregate total moisture Required Required
Fine aggregate absorption Required Required
Fine aggregate dry-loose unit
weight Not required Required

can be used to assist the user in selecting concrete proportions. The appli-
cability of the system is limited to the following cases:

1. It is assumed that the users will have the concrete mix design speci-
fications or similar documents and information about material properties.
2. All types of concrete will use type I cement with the exception of mass
concrete that will use type II cement.
3. Physical and chemical properties of portland cement and/or fly ash are
satisfactory according to the specifications.
4. Coarse and fine aggregates are graded within limits of generally ac.
cepted specifications.
5. The quality and properties of admixtures are satisfactory based on the
given specifications.
6, The concrete proportions will be selected without considering cost.

Limitations 2 and 6 could have been handled with more time. Limitations

367

J. Constr. Eng. Manage. 1994.120:357-373.


3, 4, and 5 were required since handling these problems was beyond the
scope of the work. The writers believe and the experts interviewed agreed
that the developed system could be used on a daily basis if the preceding
limitations were taken into consideration.

System Validation
The developed system was tested on three projects that were provided
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

by the experts. The writers performed the concrete mix design and the
necessary adjustments of these three projects on the developed expert sys-
tem. Experts were asked to perform the concrete mix design with its nec-
essary adjustments on both of the first two projects. The data of the first
two projects selected by the expert system were compared with data selected
by experts. The results of the third project selected by the expert system
were compared with data provided by Duke Power Co., Charlotte, N.C.

Project 1
Project 1 was a normal-weight concrete slab using type I cement. Table
6 is a list of the project information. Table 7 shows the comparison of the
trial batching mix design selected by the developed system and the experts.
Experts indicated that the concrete proportions selected by the system were
accurate enough for the first trial batching. The experts also indicated that
the differences among the three mix designs were not critical and for a
particular mix design there are always several answers that can satisfy the
requirements of the specification.
Experts were also asked to perform slump adjustment on the same mix-
ture. Both expert system and experts used the same way to adjust the slump.
Table 8 shows that 3.63 kg (8 lb) of water was added to the mixture to

TABLE 6. Mix Design Information for Project 1


Required information Conditions
(1) (2)
Fly ash Class F
Percent of fly ash 15% by weight
Maximum size of coarse aggregate 0.95 cm
Air entraining Required
Maximum slump 10.16 cm
Delivery time 1.0 h
Air temperature Around 32.2~
28-day compressive strength 27.58 MPa
Maximum water/cement ratio Not specified
Concrete exposure cohdition Mild
Chemical admixture ASTM C494 type D
Cement: specific gravity 3.15
Fly ash: specific gravity 2.45
Coarse aggregate: specific gravity 2.65
Coarse aggregate: total moisture 2%
Coarse aggregate: absorption 0.5%
Coarse aggregate: dry-rodded unit weight 1603 kg/m3
Fine aggregate: specific gravity 2.63
Fineness modulus of fine aggregate 2.6
Fine aggregate: total moisture 6%
Fine aggregate: absorption 0.7%

368

J. Constr. Eng. Manage. 1994.120:357-373.


TABLE 7. Trial Batching Mix Design Comparison between Expert System and
Experts for Project 1
Item Expert system Expert la Expert 2 b
(1) (2) (3) (4)
Cement (kg) 272.4 274.7 267.9
Fly ash (kg) 48.1 49.9 45.4
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

Water (kg) 110.3 108.9 106.7


Coarse aggregate (wet kg) 600.2 603.8 599.3
Fine aggregate (wet kg) 755.9 744.6 762.7
Air content (%) 4.5 5 5
aExpert 1 = John Rauschenbach, Zupan and Smith Sand and Concrete, Inc., Green-
ville, S.C.
bExpert 2 = John T. Carlton, Metromont Materials Corp., Greenville, S.C.

TABLE 8. Slump Adjustment for Project 1


Add item Expert system Expert 1~ Expert 2 b
(1) (2) (3) (4)
Cement (kg) 7.7 8.2 7.7
Water (kg) 3.6 3.6 3.6
Fine aggregate (wet kg) - 15.4 - 16.3 - 15.4
aExpert 1 = John Rauschenbach, Zupan and Smith Sand and Concrete, Inc., Green-
ville, S.C.
bExpert 2 = John T. Carlton, Metromont Materials Corp., Greenville, S.C.

TABLE 9. Final Mix Design Comparison Between Expert System and Experts for
Project 1
Item Expert system Expert 1a Expert 2 b
(1) (2) (3) (4)
Cement (kg) 280.1 282.8 275.6
Fly ash (kg) 48.1 49.9 45.4
Water (kg) 113.9 112.6 110.3
Coarse aggregate (wet kg) 600.2 603.8 599.3
Fine aggregate (wet kg) 740.5 728.2 747.3
Air content (%) 4.5 5 5
"Expert 1 = John Rauschenbach, Zupan and Smith Sand and Concrete, Inc., Green-
ville, S.C.
bExpert 2 = John T. Carlton, Metromont Materials Corp., Greenville, S.C.

increase the slump by 2.54 cm (1 in.). C e m e n t content was also increased


to keep the water-cement ratio constant. Fine aggregate was reduced to
keep the yield unchanged. Table 9 shows the final concrete proportions
selected by the developed system and the experts.

Project 2
Table 10 lists the mix design information for project 2. The project used
normal-weight concrete for a p a v e m e n t using type I cement. Table 11 shows
the trial batching concrete proportions selected by the developed expert
369

J. Constr. Eng. Manage. 1994.120:357-373.


TABLE 10. Mix Design Information for Project 2
Required information Conditions
(1) (2)
Fly ash Not used
Maximum size of coarse aggregate 3.81 cm
Air entraining Required
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

Maximum slump 7.62 cm


Delivery time 0.5 h
Air temperature Around 21.1~
28-day compressive strength 24.13 MPa
Maximum water/cement ratio 0.45
Concrete exposure condition Severe
Chemical admixture ASTM C494 type A
Cement: specific gravity 3.15
Coarse aggregate: specific gravity 2.68
Coarse aggregate: total moisture 3%
Coarse aggregate: absorption 0.5%
Coarse aggregate: dry-rodded weight 1,523 kg/m 3
Fine aggregate: specific gravity 2.64
Fineness modulus of fine aggregate 2.8
Fine aggregate: total moisture 5%
Fine aggregate: absorption 0.7%

TABLE 11. Trial Batching Mix Design Comparison between Expert System and
Experts for Project 2
Item Expert system Expert la Expert 2 b
(1) (2) (3) (4)
Cement (kg) 258.3 261.1 258.8
Water (kg) 70.8 72.6 70.4
Coarse aggregate (wet kg) 851.7 842.2 853.5
Fine aggregate (wet kg) 665.6 658.3 653.8
Air content (%) 5.5 6 6
aExpert 1 = John Rauschenbach, Zupan and Smith Sand and Concrete, Inc., Green-
ville, S.C.
bExpert 2 = John T. Carlton, Metromont Materials Corp., Greenville, S.C.

TABLE 12. Air Content Adjustment for Project 2


Add item Expert system Expert 1a Expert 2 b
(1) (2) (3) (4)
~ir-entraining admixture (mL) 30 30 30
Water (kg) - 1.8 - 1.8 - 1.8
Fine aggregate (kg) - 5.4 - 4.5 - 4.5
aExpert 1 = John Rauschenbach, Zupan and Smith Sand and Concrete, Inc., Green-
ville, S.C.
bExpert 2 = John T. Carlton, Metromont Materials Corp., Greenville, S.C.

370

J. Constr. Eng. Manage. 1994.120:357-373.


TABLE 13. Final Mix Design Comparison Between Expert System and Experts
for Project 2
Item Expert system Expert 1~ Expert 2 b
(1) (2) (3) (4)
Cement (kg) 258.3 261.1 258.8
Water (kg) 69.0 70.8 68.6
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

Coarse aggregate (wet kg) 851.7 842.2 853.5


Fine aggregate (wet kg) 660.1 653.8 649.2
Air content (%) 6.5 7 7
aExpert 1 = John Rauschenbach, Zupan and Smith Sand and Concrete, Inc., Green-
ville, S.C.
bExpert 2 = John T. Carlton, Metromont Materials Corp., Greenville, S.C.

TABLE 14. Mix Design Information for Project 3


Required information Conditions
(1) (2)
Type of concrete Normal weight
Type of cement I
Fly ash Class F
Percent of fly ash 2O%
Maximum size of coarse aggregate 1.27 cm
Air entraining Not required
Maximum slump 10.2 cm
Delivery time 0.5 h
Air temperature Around 21.1~
28-day compressive strength 34.48 MPa
Maximum water/cement ratio Not specified
Concrete exposure condition Mild
Chemical admixture ASTM C494 type D
Cement: specific gravity 3.15
Fly ash: specific gravity 2.45
Coarse aggregate: specific gravity 2.68
Coarse aggregate: dry-rodded weight 1,715 kg/m 3
Fine aggregate: specific gravity 2.63
Fineness modulus of fine aggregate 2.4

TABLE 15. Comparison between Expert System and Actual Data for Project 3
Item Expert system Actual data
(1) (2) (3)
Cement (kg) 256.5 254.2
Fly ash (kg) 64.0 68.1
Water (kg) 153.9 150.3
Coarse aggregate (SSD kg) 773.6 774.1
Fine aggregate (SSD kg) 513.9 522.6
a
Air content (percent) 2.5%
aDatum not recorded.

371

J. Constr. Eng. Manage. 1994.120:357-373.


system and the experts. According to the experts opinion, the proportions
selected by the system were close to the proportions selected by the experts.
Experts were asked to make air content adjustment for this project,
because air content needed to be increased by 1%. Both the system and
experts recommended that the amount of air-entraining admixture should
be increased by 30 mL (1 oz). An increase in air content will be accompanied
by an increase in slump unless the amount of water used is reduced. Also,
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

adjustment should be made in fine aggregate weight as necessary to maintain


required total effective displaced volume. Table 12 shows the adjustments
made to the mixture. The final concrete proportions selected by the system
and experts are presented in Table 13.

Project 3
Table 14 gives the mix design information for project 3. The writers
performed mix design on the developed expert system. The results were
compared with the data provided by Duke Power Co. Table 15 shows the
comparisons. Experts were consulted and all of them agreed that the con-
crete proportions selected by the system were satisfactory to make the trial
batching.

SUMMARY AND CONCLUSIONS


A rule-based expert system was developed to assist the user in concrete
mix design including the necessary adjustments to the design. The developed
expert system can handle mix designs for normal-weight, normal-weight-
mass, normal-weight no-slump, heavyweight, and lightweight concrete. Be-
sides workability, consistency, strength, durability, and density, the expert
system considers other criteria such as admixtures, transportation, and air
temperature that affect the concrete mix design.
The expert system was developed in two steps. The knowledge-acquisition
process and the development of the system. During the knowledge-acqui-
sition process, the writers performed an extensive literature review and
interviewed and surveyed several experts. The writers developed several
flowcharts based on the information contained in the literature and t h e
knowledge acquired from the experts. Based on the flowcharts, the writers
developed the expert system. The expert system was developed using EXSYS
Professional and Quattro Pro. The concrete mix design determined by the
expert system is presented on a spreadsheet.
The developed expert system was tested on three projects provided by
the experts. The writers performed the concrete mix design of three projects
on the developed expert system. The expert system's concrete proportions
selections compared favorably with the experts' selections.
The best approach to making a concrete mix design is to use proportions
previously established for similar concrete using the same materials. In
addition, rules-of-thumb and past experience should be used, wherever
possible, for concrete mix design adjustment. Where such prior information
is limited or unavailable, the developed expert system can be used to assist
the user in selecting the proportions.

ACKNOWLEDGMENTS
The writers would like to thank Jackie Rogers (Piedmont Olsen Hensley,
Inc., Greenville, SC), John Rauschenbach (Zupan and Smith Sand and
Concrete, Inc., Greenville), John Carlton (Metromont Materials Corp.,
372

J. Constr. Eng. Manage. 1994.120:357-373.


Greenville), and Curtis Arnold (Duke Power Co., Charlotte, NC) for their
invaluable input and patience.

APPENDIX. REFERENCES
"ACI 116R-85, cement and concrete terminology." (1989). A CI manual of concrete
practice, Part 1, American Concrete Institute, Detroit, Mich.
Downloaded from ascelibrary.org by New York University on 05/18/15. Copyright ASCE. For personal use only; all rights reserved.

"ACI 211.1-81 standard practice for selecting proportions for normal, heavyweight,
and mass concrete." (1989). A C I manual of concrete practice, Part 1, American
Concrete Institute, Detroit, Mich.
"AC1214.1R-81, use of accelerated strength testing." (1989). A C1 manual of concrete
practice, Part 2, American Concrete Institute, Detroit, Mich.
Celik, T., Thorpe, A., and McCaffer, R. (1989). "Development of an expert system."
Concrete Int. Des. and Constr., 11(8), 37-41.
Clifton, J. R., and Kaetzel, L. J. (1988). "Expert systems for concrete construction."
Concrete Int. Des. and Constr., 10(11), 19-24.
"Control tests on fresh concrete." (1981). Concr. Constr., 26(7), 557-561.
EXSYS Inc. (1988). "EXSYS professional manual." Albuquerque, N.M.
Malasri, S., and Maldonado (1988). "Concrete mix designer." Computer Applications
in Concrete Technology, SPl13-3, American Concrete Institute, Detroit, Mich.
Neville, A. M. (1981). Properties of concrete, 3rd Ed., Pitman Publishing, Inc.,
Marshfield, Mass.
Waddell, J. J. (1984). Concrete manual, International Conference of Building Of-
ficials, Whittier, Calif.

373

J. Constr. Eng. Manage. 1994.120:357-373.

You might also like