Professional Documents
Culture Documents
The purpose of this report is to introduce the use of a simplified version of the statistical concepts as outlined in ACI 214 for the specification, control, and evaluation of the production of concrete. For a more elaborate discussion of the concepts, see the ?Recommended Practice f o r the Evaluation of Strength Test Results o Concrete? f @CI 214).
Keywords: coefficient of variation; compression tests; compressive strength;
concrete construction; concretes; cylinders; evaluation; quality control; sampling; standard deviation; statistical analysis; variations.
INTRODUCTION The strength test is widely used in specifying, controlling, and evaluating concrete quality. Quality concrete must be able to: 1) carry loads imposed upon it; 2) resist deterioration; and 3) be dimensionally stable.
Copyright American Concrete Institute Provided by IHS under license with ACI No reproduction or networking permitted without license from IHS
--`````,```,````,``````````,,`-`-`,,`,,`,`,,`---
AC1 214.3R-88
(Reapproved 1997)
Simplified Version of the Recommended Practice for Evaluation of Strength Test Results of Concrete
Reported by AC1 Committee 214
V. Ramakrishnan
Chairman Steven H. Gebler Eugen O. Goeb Gilbert J. Haddad David F. Harrald Peter A. Kopac Kenneth R. Lauer H. S. Lew V. M. Malhotra Larry W. Matejcek T a n n R. Naik Robert E. Neal Robert E. Philleo Francis J. Principe Owen Richards Orrin Riley Ephraim Senbetta P. N.Balagun Secretary
S. N. Shanmughasundram Shyam N. Shukla Luke M. Snell Roger L. Sprouse Rodney J. Stebbins Michael A. Taylor J. Derle Thorpe* Don J. Wade
CONTENTS
T?here are several tests that can be made with plastic and hardened concrete, but the strength test is generally accepted as a measure of the quality of concrete being placed on a project. Although the strength test is not a direct measure of concrete durability or dimensional stability, it provides an indication of the water-cement ratio of the concrete. The water-cement ratio, in turn, directly influences the strength; durability; wear resistance; dimensional stability; and other desirable properties of concrete. The strength test is also used to measure the variability of concrete. By using statistical methods based on the strength test, realistic specifications can also be prepared. VARIABILITY OF CONCRETE Portland cement concrete is subject to numerous factors that affect its strength and other properties. These may include variations in the manufacture of portland cement; preparation of aggregates; batching, mixing, and curing of concrete; and finally in the preparation, handling, and testing of the cylinders. The major variables are listed in [?.:N-: i. These variables must be considered when speci@ing, producing, or controlling the strength of concrete. NORMAL DISTRIBUTION Test data from large concrete projects with many tests show a grouping around the average strength. A
AC1 Committee Reports, Guides, Standard Practices, and Commentanes are intended for guidance in designing, planning, executing, or inspecting construction and in preparing specifications Reference to these documents shall not be made in the Project Documents If items found in these documents are desired to be part of the Project Documents they should be phrased in mandatory language and incorporated into the Project Documents
*Principal author of this report. 1988, American Concrete Institute. Copyright ?I? All rights reserved including rights of reproduction and use in any form or by any means, including the making of copies by any photo process, or by any electronic or mechanical device, printed, written, or oral, or recording for sound or visual reproduction or for use in any knowledge or retrieval system or device, unless permission in writing is obtained from the copyright proprietors.
214.3R-2
~ i 3 S i psi l
10
Variations due to fabrication techniques Cylinder molding Poor quality molds Handling and curing of newly made cylinders
m3
z
o o o
3600
o
4000
o
4400
o
4800 5200
Changes in curing Temperature variation Variable moisture Delays in bringing cylinders to the laboratory Poor testing procedures Care of cylinders, transportation and capPing Improper placement in testing machine Testing machine platens out of specifications Incorrect speed of testing
O
2400 2300 3200
Fig. 2-Strength tests plotted in Fig. I with normal distribution curve superimposed on data
10
O O O o O O 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0
n W
3 z
O
o o o o o o o o o
0 0 0 0 0 0 0 0 0 0 0
2400 2300 3200 3600 4000 4400 4800 5200
2400
2800
3200
3600
4000
4400
4800
5200
psi)
Fig. 3-Normal distribution curve represents variation of individual test results plotted in Fig. I and 2 distribution on the plot of individual strengths. As shown in Fig. 2, this curve smooths out the plot by reducing the effect of individual differences through averaging. The center of the curve is located at the average of all the tests. The area under the curve represents 100 percent of the tests. Fiq. I! shows the normal distribution curve used to represent all of the tests, rather than using the individual tests plotted in their respective cells. This curve will be used to represent all of the strength tests without the individual plotted tests throughout the remainder of this report. The example illustrated in this report is the compressive strength test, but the procedures outlined here may be used on test Licensee=Aramco HQ/9980755100 data from any test used to determine the strength of Not for Resale, 01/22/2006 07:56:58 MST conCrete.
typical grouping is illustrated in Fi$; .!. To produce Fig. 1, the strength tests are divided into cells. The cell width for Fig. 1 is 200 psi. For example, the seven tests that fall between 3900 and 4099 psi have been plotted in the cell listed as 4000 psi.* Similarly, all other strengths from the series of tests have been plotted in their respective cells. Since the grouping of tests on each side of the average is nearly symmetrical, it is called a normal distribution. It is possible to superimpose a normal
Copyright American Concrete Institute Provided by IHS under license with ACI No reproduction or networking permitted without license from IHS
214.3R-3
fi5=340
\
psi
COMPRESSIVE STRENGTH
(psi)
Fig. 4-Plot of three normal distribution curves with the same average. Curves illustrate low variabdi@ s = 340psi , medium variabilit ( = 560psi), and igh s variabi i@( = 800 psi) (ce1) s width is 400 psi)
fi
Where X, is each strength test from i = 1 to i = n, and EX, represents the sum of all the tests. More than one concrete cylinder is usually made from the same sample of concrete. A strength test is defined as the average strength of all cylinders made from the same sample of concrete cured under the same conditions and tested at the same age. Variations in strength between cylinders cast from the same sample may be attributed to variability of the concrete within the sample, cylinder molding, care, transportation, and testing procedures. If testing procedures are uniform, variations observed between strength tests can be attributed to variations in materials, batching, mixing, transportation, and curing. Standard deviation s The standard deviation is a measure of the variability or scatter of test values around the average. The higher the standard deviation, the greater the variability. The standard deviation can be calculated using one of the following equations
The normal distribution curve can be either steep or flat, depending on the grouping of the individual tests on each side of the mean. Three normal distribution curves are illustrated in Fig. 4. The shape of the curve indicates the variability in the test data. The steep curve represents a series of strength tests with low variability. Strength tests are closely grouped around the average. The flat curve represents a series of tests with high variability. Whatever the shape of the normal distribution curve, it reveals significant information about the test data it represents. The top of the curve is located at the average of all the tests. STATISTICAL EVALUATION Concrete samples should be obtained by randomly sampling batches of a given class of concrete used on the project. Different portions of a structure may require different concrete mix designs. Each mix is often called a class of concrete. Each class of concrete usually has its own proportions, and should be sampled and evaluated separately. If the number of randomly obtained samples of a given class of concrete is large enough, the strength test results can be statistically evaluated. The statistical values calculated from the test data, which are useful in the evaluation, follow. Average X The average (often called the mean) is the approximate middle value between extremes. The average can be calculated using the following formula
s =
(X, j X) + (X X) + . . .
-
( X , - X)
n -1
(3)
where
s = standard deviation
X , , X,: . . . X , = individual strength tests n = total number of tests X = average strength of all the tests The preceding formula may be represented by
1 to i
n , and
Some calculators compute the average and standard deviation in the same operation. If a calculator with statistical functions is not available, the standard deviation may be determined from the following formula using a simple four-function calculator
,.
X = X,
+ X + X ,
n
...
+ X
(1)
where C(X,) is the sum of the square of each test, and (EX,) is the sum of all the tests squared.
Where X is the average strength; XI, X 2 , X 3 , . . X , are individual strength tests; and n is the total number Copyright American Concrete Institute Provided by IHS under license with ACI of or networking permitted without license from IHS tests. No reproduction
214.3R-4
Many computer programs are available to calculate these statistical values. The normal distribution curve will break, .e., change from a concave shape to a convex shape at two points, one on each side of the curve. These points will mark the limits of zones that will include approximately one-third of ail the tests below the average and one-third of all the tests above the average. These points are located a distance equal to one standard deviation each side of the average. These points can be determined by calculating the standard deviation for the set of tests.
159 %
+-t
I
68.2%
341%
I
I
341%
INTERPRETATION OF RESULTS
When the relationships between the individual test results, the normal distribution curve, and the statistical values produced from the test data are understood, it is possible to draw conclusions about the variability of the test data. The area under the normal distribution curve represents 100 percent of the tests. A series of zones can be created under the curve by drawing vertical lines, each spaced a distance equal to a standard deviation on each side of the vertical line drawn at the average. Fig. 5 shows a normal distribution curve with the percentage of tests expected to fall within each zone of the curve. Each zone can be identified by standard-deviation limits on each side of the average. Fifty percent of the tests fall on each side of the center of the curve, or average strength. The zone bounded by one standard-deviation limit on each side of the average (I s) includes 68.2 percent of the tests. As soon as the average and the standard deviation are calculated, the shape of the normal distribution curve that represents the data can be visualized. Moving a second standard-deviation limit on each side of the average will include an additional 27 percent of all the tests. Therefore, a total of 95.2 percent of all the tests fall within two standard-deviation limits (I 2s) on each side of the average. An additional 2.4 percent of all the tests fall between two and three standard-deviation limits on each side of the average strength for a total of approximately 100 percent of the tests. These three standard-deviation limits each side of the average strength (I 3s) are normally considered to be the limits that include almost all test values. Engineers are not normally concerned with strengths that are too high. Therefore, only the standard-deviation limits below the average cause concern when evaluating the strength of concrete. F q . 5 shows 15.9 percent of the tests, or approximately l in 6, will be below one standard-deviation limit; 2.4 percent of the tests, or approximately 1 in 42, will be below the two standarddeviation limits.
--`````,```,````,``````````,,`-`-`,,`,,`,`,,`---
2400
2800
3200
3800
4000
4400
4800
5200
Fig. 5 - Percentages of tests expected to fall in each zone of normal distribution curve
distribution curve, if the average strength of the concrete is approximately equal to the specified strength, one-half of the concrete will have a strength less than the specified strength. Because it is usually not acceptable to have one-half of the strength tests lower than specified strength, the average strength must be higher than the specified strength by some factor. It is possible to use the statistical tools introduced here in all phases of concrete production - strength has been used as an example. Similar principles can be applied to other important characteristics of concrete such as entrained air, which relates to durability. The specification writer, in consultation with the engineer, selects a specified strength and the percentage of low tests that are considered acceptable for the class of concrete. AC1 31 8, Building Code Requirements for Reinforced Concrete, provides guidelines for selecting the acceptable number of low tests. An example of a statement for strength in the specification might read: The average of all strength tests shall be such that not more than one test in ten (IO percent) shall fall below the specified strength f. of 3500 psi.
214.3R-5
10
PS
10
\-l
I c i i 3 4 W psi Si346pri
<B
I<B
w
I-
f
5
8
[ r
m
rcceptabia lumber of low tests
3 z
n
3200
O
2400 2800
3600
4000
4404
4800
2400 5200
2800
3200
3600
4000
4400
4600
5200
COMPRESSIVE STRENGTH
(psi)
(Plotted in ceils of 200 psi) Fig. 6-Illustration of determination of required average strength f, = 3000 psi, product of probability factor p, and standard deviation s = 800 psi. The sum of f: and ps = f,,, which is 3811 psi in this example
Fig. 7-Plot shows effect of standard deviation on calculating required average strength. In each figure probability factor is 1.28, allowing I O percent of tests to fall below specified deviation s strength
--`````,```,````,``````````,,`-`-`,,`,,`,`,,`---
able percentage of low tests, the probability factor can be determined using the properties of the normal distribution curve. The probability factors for various perRequired Required centages of low tests are given in T a S e 2. average average Percentage of strength Percentage of strength The standard deviation is obtained by analyzing the low tests low tests concrete producers data. Since the standard deviation 1; + 1.60s 1; + 0.00s 50.0 5.5 for a project is not known at the beginning of a proj1 + 0.10s ; J; + 1.70s 46.0 4.5 ect, Chapter 4 of AC1 318 permits the substitution of a 1; + 0.20s 1; + 1.80s 42.1 3.6 1; + 0.30s J + 1.90s 38.2 2.9 standard deviation calculated from at least 30 consecuf, + 0.40s 1; + 2.00s 34.5 2.3 tive strengths on concrete produced at the proposed 1; + 0.50s 1; + 2.10s 30.9 1.8 concrete plant using similar materials and conditions. 1; + 2.20s 1; + 0.60s 27.4 1.4 ,. -. f, + 0.70s 1; + 2.30s 1.1 24.2 k-q; .7 shows the effect of the standard deviation on J; + 0.80s i+ 2.40s 0.8 21.2 the required average strength. The steep curve plotted + 0.90s J + 2.50s 0.6 18.2 from strength data with low variability, .e., a low f, + 1.00s f, + 2.60s 0.45 15.9 J; + 1.10s 1; + 2.70s standard deviation (s = 344 psi), shows that the re0.35 13.6 f + 1.20s J; + 2.80s 11.5 0.25 quired average strength should be 3440 psi. Eq. (8) is 0.19 f, + 1.30s 9.7 1; + 2.90s used to calculate the value. For the flat curve plotted f, + 1.40s 1; + 3.00s 0.13 8.1 from strength data with high variability, .e., a high f+ 1.50s 6.7 standard deviation (s = 806 psi), a required average The probabil y factor used in the calculation of ? required average strength can be c ermined from this table using the accepiable percentage of strength of 4032 psi is needed. In each case, 10 percent low tests. Source: William A. Cordon, Concrete Quality. AC1 Echiridion E704-4, of tests can be expected to fall below the specified 1973,p. 8. strengt h. Most concrete comes from plants with continuous where testing programs. Quality-control personnel from these f,, = required average strength, psi plants can supply standard deviation data on each class f,! = specified strength, psi of concrete. Since three standard-deviation limits are = probability factor based on the percentage of generally considered to include all tests, the engineer tests the designer will allow to fall below A who unrealistically refuses to recognize the variability s = expected standard deviation for the project, psi that does exist, even in carefully controlled concreting operations, and demands that no tests fall below the specified strength, must realize that the required averUse of the normal distribution curve to obtain the age strength must then be three standard-deviation limrequired average strength is illustrated in Fig. S. To its above the specified strength. Even with the required calculate the required average strength, the engineer average strength at three standard-deviation limits must decide the specified strength and what percentage Copyright American Concrete Institute Provided by IHS under license with ACI above the specified strength, there is a slight chance of of tests falling below the specified strength will be al- Licensee=Aramco HQ/9980755100 MST No reproduction or networking permitted without license from IHS Not for Resale, 01/22/2006 07:56:58 a test falling below the specified strength. lowed. When the decision has been made on an accept-
214.3R6
,--. ,.,l.%l j ~,
. c 2 ~ - ~ ii?c indicates that using three standard-deviation , limits does not completely insure that no test will fall below the specified strength. The predicted percentage of low tests where the average strength exceeds the specified strength by three standard deviation limits is 0.13 percent, or 1.3 tests in 1000. When the engineer understands the implications of the three standard-deviation limits, he may want to consider using several different probability factors for a given project, depending on the critical nature of the strength of each class of concrete. Tabiv 3 lists criteria . for selecting different probability factors based upon the risk if the concrete strength falls below the specified strength.
Table 3 Recommendations for 3 to be used s in computing the required averas strength based on critical nature of stren! h of concrete.
iecommended values for p s for computing required average strength ( f , = fl + DS) Concrete strength below f;' cannot be tolerated (minimum strength specification) Strength is critical Strength belowf;' is not critical but a test below f;' - 500 is critical. This requirement applies only where s is above 500 psi (AC1 318) 1.3 in 1000
3.00s
1 in 100 Variable
2400
2800
3200
3600
4000
4400
4800
5200
EVALUATING CONCRETE STRENGTH As the strength tests fi-om a project become available, continuous evaluation of the data is desirable. Updated determination of the average strength and standard deviation will permit an evaluation of how well the actual project values compare with values used at the beginning of the project. An understanding of the percentage of tests falling w i t h each zone, under the normal distribution curve illustrated in Fiy. 3, will aid in this evaluation. The approximate percentage of tests falling below the specified strength can be calculated anytime after test data become available using Eq. (7) and 'I('abie2 as fol-
Fig. 8-Approximate desired relationship between required average strength and average strength
Where
X
f
s
lows
p = -
x-f,'
(7)
Copyright American Concrete Institute S Provided by IHS under license with ACI No reproduction or networking permitted without license from IHS
When the probability factor has been calculated fi-om actual project data, the approximate percentage of low tests can be determined using 'Table I as follows. Find the probability factor closest to the calculated value in the column labeled "Required average strength" of Tnbk 2. The corresponding percentage can be read fi-om the columns labeled "Percentage of low tests." HQ/9980755100 Licensee=Aramco
Not for Resale, 01/22/2006 07:56:58 MST
--`````,```,````,``````````,,`-`-`,,`,,`,`,,`---
CONTROL OF CONCRETE STRENGTH At the beginning of a concreting operation, the strength level of the concrete being produced is based upon the calculation of the required average strength. This is a hypothetical production strength. It assumes that the variables affecting the strength of concrete will be the same in the fbture as they have been in the past. As the f r s t test data become available, the required average strength is replaced by the actual value-the project average strength. If the standard deviation fi-om the project is approximately equal to the value used in the calculation of the required average strength, the project average strength should be maintained close to the required average strength. If the project average strength is below the required average strength, the percentage of tests below the specified strength will be greater than the acceptable value and steps must be taken to increase the strength of the concrete. The strength of the concrete must also be increased if the standard deviation of the project is greater than the assumed standard deviation used in the determination of the required average strength. If the project standard deviation increases, the average strength of the concrete must be increased. An illustration of the ideal relationship between these values is shown in Fiy. 8.
+
Stren th of concrete is not critica!f(ACI 214, AC1 318) 1 in 10 Average of three consecu- 1 Test: 9 in 100 tive tests does not fall be- Average of three low f (AC1 3 18) : tests: 1 in 100 Strength of concrete is of minor consequence in desien 1 in 5
1.28s
1.34s
0.8%
214.3R-7
Table 4 Control standards for evaluating performance of testing program from project using the within-test coefficient of variation
Field control testina Controll standards for within-test coefficient of variation Excellent Below 1 .5
I I
I I
I I
~..
variability in concrete strength. The first category lists factors, including strength, that affect the properties of the concrete. The second category lists possible variations caused by testing procedures. Variations in testing may have a significant effect on the apparent strength of the concrete and, consequently, on the evaluation of the actual strength. It is possible to separate some of the effects of these two major sources of variability. A test consists of all the test cylinders made from the same sample of concrete, cured under the same conditions, and tested at the same age. It is assumed that two or more test cylinders made from the same sample of concrete and tested at the same age should have the same strength. Variations in the strength of these cylinders occur partly because of differences in the testing procedures. Differences in strength between two or more test cylinders made from the same sample of concrete are called within-test variations. The within-test standard deviation is calculated from the following equation
The range is the difference between the highest and lowest strengths of the cylinders making up a test. The factor d2 is based on the number of cylinders used to calculate the range. If two cylinders are used in a test, d2 equals 1.128. If three cylinders are used in a test, the chance for variability is greater and the factor is larger than 1.693. The range calculated from four cylinders per test would use d2 of 2.059. To complete the within-test evaluation, calculate the within-test coefficient of variation, which is the withintest standard deviation expressed as a percentage of the average strength. The within-test coefficient of variation can be calculated using Eq. (9)
(9) W h e r e V,, = within-test coefficient of variation expressed as a percent, and X = average strength for the class of concrete, psi. Standards for evaluating the within-test coefficient of variation are given in Tasic: J. If the within-test coefficient of variation is high (fair or poor), testing may be the reason for poor test results rather than the concrete quality.
CONTROL CHARTS
An important part of the evaluation process is the regular plotting of values for a visual picture of concrete performance. Three simplified charts prepared specifically for concrete control are illustrated in Fig. 9. While these do not contain all the features of formal control charts, they can be useful to engineers, architects, and ready-mix plant superintendents. On chart A, (Fig. 9) the compressive strengths for each of the two cylinders that make up a test are plot-
~l
l
0
'
Average
O !
I w o cylinders
A/--
".
--140
2000
S p e c i l i e d r l r e n g l h 1, Required average s l r e n g l h I E
pr
N
--
280
E cl u
.CHART
C
E
6
30
Copyright American Concrete Institute Provided by IHS under license with ACI No reproduction or networking permitted without license from IHS
--`````,```,````,``````````,,`-`-`,,`,,`,`,,`---
214.3R-8
ted as circles. The point representing the average of the two cylinders is the point to which the solid line is drawn. The same chart may be used if three or more cylinders make up a test. The dotted line, representing the required average strength, is determined from Eq . .~, kt.;, or T.>hl.- 2. If the volume of concrete produced re,ci~....i~: quires more than one test per day, the average of all tests on a given day can be plotted for that day. Tests can also be plotted in chronological order rather than by days. The individual tests will be used separately in the calculation of running averages. Chart B of Fig. 9 is the moving average of the previous five test averages from Chart A. Each time a new test average is added to the chart, a new average of five tests is calculated, using the new average along with the previous four. The number of tests used to calculate moving average can be varied to suit each job. This
This report was submitted to letter ballot of the committee and was approved in accordance with AC1 balloting procedures.
Copyright American Concrete Institute Provided by IHS under license with ACI No reproduction or networking permitted without license from IHS
--`````,```,````,``````````,,`-`-`,,`,,`,`,,`---
chart is valuable in indicating trends and will show the influence of seasonal changes, changes in materials, etc. Chart C (Fig. 9) is a moving average for the range, where the average range of the previous ten groups of companion cylinders is plotted each day. Fig. 3 shows Charts A, B, and C for 46 tests. To be fully effective, charts should be maintained throughout the entire job. Control charts can be updated and plotted by computer each time new test data are submitted for analysis. For a more rigorous treatment of the use of these concepts in the evaluation of the strength, consult the Recommended Practice for Evaluation of Strength Test Results of Concrete (AC1 214-77) (Reaffirmed 1983).