Alasdair’s Engineering Pages © A. N. Beal 2014                www.anbeal.co.uk

Alasdair’s
Engineering Pages

www.anbeal.co.uk

The Structural Engineer 87 (10)19 May 2009

Viewpoint

Concrete strength testing - are the code writers getting it right?

Alasdair Beal proffers his views on concrete strength testing and the Eurocodes

Paul Toplis’s ‘Viewpoint’ [1] is timely in drawing attention to the problems specifications for concrete strength testing can create for contractors and suppliers. It is true that some of these are caused by inexperienced engineers (‘designer/specifiers’ in Eurospeak) specifying unrealistic sampling rates and misunderstanding BS EN 206/ BS 8500 compliance rules. However the underlying problems have been created not by these engineers but by code writers and the rules they have written for the engineers.

Thus I agree that a concrete slab specification which calls for a sampling rate of 1/2 and two 28-day cube tests on every sample, no less than 40, 28-day cube tests for a concrete volume of 240m³, is completely unrealistic. However the requirement for two 28-day cube tests on every sample was not dreamed up by the project engineer - this comes from BS EN 206. Also, although the engineer should have known that a 1/2 sampling rate is not needed for concrete in a slab, it does not help when code writers remove all guidance on the subject from codes.

Given that a single cube test to measure the strength of a batch of concrete has been regarded as acceptable for a very long time, it is not clear why the BS EN 206 committee decided that we now need double this number of tests. However until BS EN 206 is changed, engineers are stuck with it - and contractors are stuck with having to make twice as many test cubes.

Thus the problem of engineers specifying excessive numbers of cube tests could be solved by a simple amendment to BS EN 206. However Mr Toplis’s comment that results from the BS EN 206/ BS 8500 compliance rules ‘may look a little counter-intuitive at first sight’ is a diplomatic description of a set of rules which produce results that may make no sense at all.

BS 8500 requires tests to be assessed in non-overlapping groups of six; the mean of each group must exceed the characteristic strength by at least 2N/mm² and individual results must not fall more than 4N/mm² below it. If the group fails either test, the entire volume of concrete it represents must be rejected.

In the table below, the BS EN 206/ BS 8500 rules are applied to a set of 18 results (specified characteristic cube strength 40N/mm²). The table also shows how changing the starting point in the data can affect the analysis.

The rows all contain the same set of test results but each starts from a different point in the sequence. As can be seen, in five cases all of the concrete would pass but in one case two groups of results would fail, leading to rejection of two thirds of the concrete. In the latter case, a batch of concrete with a strength of 37N/mm² would be accepted, yet batches with strengths as high as 46 and 48N/mm² would be rejected.

Thus the concrete supplier has a problem: depending the order he delivers the batches, he may find that the concrete is all accepted, or else two thirds of it may be rejected, even though he has supplied identical concrete in every case.

The customer has a problem too. Instead of no more than 5% of test results failing, 22% (4 out of 18) are below the specified characteristic strength. The characteristic strength calculated from the mean and standard deviation of the results is only 35.5N/mm², compared with 40N/mm² specified. The concrete has well below the specified characteristic strength, yet in five scenarios out of six the test results would have passed the BS EN 206 compliance criteria.

In this case, the BS EN 206 compliance rules mean that:

(i) the supplier faces uncertainty, with acceptance or rejection of the concrete depending on what order the batches are delivered in;

(ii)  the engineer is forced to make decisions which appear completely nonsensical: he may end up accepting a batch of concrete which has less than the specified strength, while at the same time other batches which are well above the specified value are rejected;

(iii) although the concrete characteristic strength is only 89% of the specified value, it has a 5/6 chance of being accepted.
















BS EN 206 states that its criteria give only a 1% chance of conforming concrete being rejected, which may well be true. However no figures are given for the risk of substandard concrete being accepted, which appears to be much higher. Concrete which passes the BS EN 206 test criteria may not comply with the specified characteristic strength.

This type of problems with strength test compliance rules is not new. BS EN 206 brings back ideas which were fashionable in the UK in the 1960s and 1970s, when CP 114 and CP 110 included similar statistically-based rules based on groups of test results. In 1981 I published an analysis of these methods [2] which found that they contained fundamental logical errors. Variations in concrete quality are quantised, as it is delivered in batches. Within-batch variation is low, so rather than being a random sample from an infinite volume, each test result gives (within close limits) the strength of a whole batch of concrete (typically 6m³). The statistical analysis used to assess the performance of compliance rules is not valid at normal sampling rates for structural concrete (between 1/2 and 1/6). For the detailed analysis readers are referred to the published paper and discussion. The BS EN 206 rules suffer from the same flaws as the old CP 114 and CP 110 rules.

The CP 114 and CP110 compliance rules were replaced by BS 5328, which introduced simpler rules: the mean of four results should be at least 3N/mm² above the characteristic strength and individual results should be no more than 3N/mm² below it. However in practice the ‘mean of 4’ criterion was normally treated as a ‘warning light’ and only individual batches which fell below the ‘-3N/mm²’ limit were rejected. Applied in this way the rules avoided the worst practical problems and reduced risks to both the customer (by defining a clear minimum strength) and the supplier (by reducing the volume of concrete at risk in a decision). It is unfortunate that BS EN 206 has abandoned this practical approach.

The current BS EN 206 rules are impractical, illogical, give unpredictable results and put too much concrete at risk on single decision. They also fail to ensure that the concrete supplied actually complies with the specified ‘characteristic strength’. (The inadequacy of the compliance rules undermines the calculations of partial factors and failure probabilities in Eurocode 2.)

The best way of checking concrete strength is also the simplest: a straightforward ‘minimum strength’ requirement for test results is logical, practical and avoids all the problems associated with criteria based on groups of tests. However if this idea would be a bit too radical for the codewriters, they could reduce the problems in the present BS EN 206 rules to a tolerable level by amending them as follows:

*  balance risks more fairly between suppliers and customers by changing the criteria from ‘group mean +2N/mm²/ individual batch -4N/mm²’ to ‘group mean +3N/mm²/ individual batch -3N/mm²’;

*  reduce the volume of concrete at risk on a decision by changing from groups of six to groups of four and treat the ‘group mean’ limit as a warning rather grounds for rejecting the entire volume of concrete;

*  when a test falls below the limit on individual results, adjacent batches should be investigated but only rejected if they fall below the minimum strength;

*  encourage the use of 7 day cube tests to give early warning of quality problems and allow early action to minimise consequences.

References

1. Toplis, R: ‘Concrete compressive strength testing - is the industry getting it right?’, The Structural Engineer, 86/ 23/24, 2 December 2008, p. 43.

2. Beal, A.N.: ‘Concrete Cube Strengths - What Use are Statistics?’, ICE Proc. Part 2, December 1981, pp. 1037-1048, discussion, June 1982, pp. 515-532.

TSE2009ConcreteStrengths.pdf

The original copy of this paper is available from

www.istructe.org/thestructuralengineer

No.

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

Strength

41

52

37

49

47

51

53

36

37

42

44

48

43

41

40

46

44

39

Mean/pass


[

46

pa

ss

]

[

43

pa

ss

]


[

42

pa

ss

]


Strength

52

37

49

47

51

53

36

37

42

44

48

43

41

40

46

44

39

41

Mean/pass


[

48

pa

ss

]


[

42

fail


]


[

42

fail

]


Strength

37

49

47

51

53

36

37

42

44

48

43

41

40

46

44

39

41

52

Mean/pass


[

45

pa

ss

]


[

42

pa

ss

]


[

44

pa

ss

]

Strength

49

47

51

53

36

37

42

44

48

43

41

40

46

44

39

41

52

37

Mean/pass


[

45

pa

ss

]


[

43

pa

ss

]


[

43

pa

ss

]

Strength

47

51

53

36

37

42

44

48

43

41

40

46

44

39

41

52

37

49

Mean/pass


[

44

pa

ss

]


[

44

pa

ss

]


[

44

pa

ss

]

Strength

51

53

36

37

42

44

48

43

41

40

46

44

39

41

52

37

49

47

Mean/pass


[

44

pa

ss

}


[

44

pa

ss

]


[

44

pa

ss

]

Table 1: BS EN 206/ BS 8500 rules are applied to a set of 18 results