Construction and Building Materials 15 2001.

81 92

Reliability of partially-destructive tests to assess the strength of concrete on site

J.H. Bungey U , M.N. Soutsos
Uni ersity of Li erpool, Department of Ci il Engineeering, Brownlow Street, Li erpool L69 3GQ, UK

Abstract The range of available tests for assessing the strength of insitu concrete based on measurements of surface zone properties is examined, together with developments in supporting documentation. Attention is concentrated upon a number of recent research programmes, including work undertaken as part of the European Concrete Building Project. These focus primarily upon pull-out and pull-off techniques, and encompass applications to early age strength assessment, lightweight and high strength concretes, and testing of repairs. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Concrete; Early-age; In-place; Near-surface; Partially-destructive, Strength; Testing; Variability

1. Introduction There has been recognition of the need to test in-place concrete to estimate strength for more than 70 years, but it is in the last 30 years that the most signicant developments of commercially available systems have occurred. Dozens of techniques have been proposed in an attempt to overcome the problems of damage and time associated with cutting, preparation and testing of cores. Whilst acceptance testing of concrete has traditionally involved compression testing of cylinders or cubes, these may not adequately reect the compaction or curing received by the insitu concrete and, of necessity, relate only to a small proportion of the concrete placed. In-place testing may be used to overcome some of these difculties. In the majority of current applications, however, testing is for one or two reasons: Evaluation of an existing structure new or old..
U

Monitoring strength development during new construction. Use of in-place tests for acceptance is, at present, limited but increased awareness of the importance of the surface zone in terms of durability has stimulated interest. A number of test methods have become established, and standardised either in Europe or North America, which involve the measurement of a surface zone property which can then be related to strength by means of an appropriate correlation. They typically cause localised surface damage which, although less than that associated with cores, must be considered at the planning stage. Consequently the term partiallydestructive is sometimes used to describe these methods. Research at the University of Liverpool has focused over recent years upon the interpretation of results provided by a number of these methods, including correlation procedures, and taking account of withinmember strength variations. This work has encompassed both laboratory and eld studies and has considered the suitability of techniques for:

Corresponding author. Tel.: q44-0151-794-5238; fax: q44-0151794-5218. E-mail address: bungey@liv.ac.uk J.H. Bungey..

0950-0618r01r$ - see front matter 2001 Elsevier Science Ltd. All rights reserved. PII: S 0 9 5 0 - 0 6 1 8 0 0 . 0 0 0 5 7 - X

82

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

Early age testing. Use with normal concretes with a range of aggregate types. Use with Lightweight and High Strength concretes. Testing repairs. The scope of these investigations and key ndings are considered in this paper, together with future prospects for this group of techniques.

means of a hand-held mechanically operated device, although electronic digital-recording versions are available w3x. Energy levels are relatively low and this method is very sensitive to a large number of factors which effectively limits its use to comparative applications where it is a valuable preliminary to other forms of testing. 2.2. Probe penetration resistance Measurements are made of the depth of penetration of a metal rod or pin that is forced into the surface of the hardened concrete by a driver unit. This is available in two basic forms: Pin-penetration test in which a 3.56-mm diameter hardened steel pin is driven by a spring-loaded device to create a hole up to 7.6-mm deep in the mortar. This is not recommended w2x for compressive strengths above 28 MPa. Probe-penetration test, commonly known as the Windsor Probe test in which a larger rod 6.35-mm diameter for normal weight concrete. is red from a specially designed gun using a smokeless powder charge producing a constant energy level. Penetration may be up to approximately 40 mm depending upon the strength of the concrete, and correlations to compressive strength are known to be inuenced by aggregate hardness. Measurements are usually based on the exposed probe length.

2. Available tests and documentation Principal test methods are described and discussed in detail by the principal author w1x and the American Concrete Institute w2x together with other related techniques. Surface zone insitu strength tests which are standardised in the USA and UK, are listed in Table 1 together with an indication of key features, whilst those forming the basis of this investigation are briey detailed below. Standards for many of these techniques also exist in other countries. It should be noted that break-off methods have received little attention in the UK and are not considered in this paper. 2.1. Rebound hammer This simple test is probably the most widely known and most commonly used of all the test methods. It involves an assessment of localised surface hardness by
Table 1 Standardised test methods Method Standards ASTM Rebound hammer C805

Principal features BS 1881 202 Existing concrete, best used comparatively. Existing concrete, pin or probe. Strength correlation specic to aggregate.

Probe penetration

C803

207

Pull-out Internal Fracture Cast-in Drilled

C900 C900

207 207 207

Existing concrete, high variability. New construction Lok-test.-general strength correlation possible. Existing concrete Capo-test. care with drilling and underreaming. Existing concrete, surface or partially cored, care with bonding, strength correlation specic to mix. New construction formed. or existing concrete drilled., strength correlation specic to mix and test type.

Pull-off

207

Break-off

C1150

207

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

83

Fig. 1. Probe penetration test. Fig. 2. Internal fracture test.

Theoretical analysis is complex and relates to the absorption of the initial kinetic energy by the concrete. No rigorous studies have been reported, but it is generally accepted that most of the energy is absorbed by a cone-shaped fracture zone as illustrated in Fig. 1. There may be a risk of cracking slender members and safety regulations may apply. 2.3. Pull-out Measurements are made of the force required to pull an embedded metal insert with an enlarged head from a concrete surface. The pulling load is applied by a tension jack bearing against the concrete surface through a reaction ring concentric with the insert. Failure involves the fracture, and often removal, of an approximately cone-shaped portion of concrete. A common feature of pull-out tests is that correlation with compressive strength is relatively insensitive to mixture characteristics such as cement and aggregate type except lightweight., size and proportions. They are thus, particularly useful when assessing a concrete of unknown composition. Several forms of this type of test have been developed: Internal fracture test developed in the UK in which a 6-mm diameter expanding wedge anchor bolt is inserted to a depth of 20 mm in a predrilled hole as shown in Fig. 2. A torquemeter device may in this case provide the pulling force. The method was developed specically for testing existing slender, prestressed concrete elements and although the equipment is inexpensive and relatively quick to use, variability is high. Lok-test, which has emerged as the most commonly used technique involving an insert placed in the concrete surface during construction. This may be xed to formwork or a otation cup and the head will typically be located 25 mm below the surface. The basic conguration is shown in Fig. 3. Preplanning is obviously required, thus, restricting the method to new construction. Many studies have been undertaken to analyse theoretically the mechanisms associated with this test, but there is no rm

agreement concerning that governing the ultimate failure load. It is generally recognised however, that well-dened experimental correlations can be established with compressive strength. This method is particularly suitable for strength development monitoring. Capo-test has been developed as a version of the Lok-Test that can be applied to existing concrete. Drilling and under-reaming operations are required to provide a groove into which a compressed steel ring can be expanded to provide a similar test conguration, see Fig. 4. The load is applied by means of a similar jack, but the test is obviously slower to perform. It is claimed however, w4x, that identical correlations with compressive strength may be utilised.

2.4. Pull-off This measures the direct tensile force required to pull a metal disk, together with a layer of concrete, from the surface to which it has been bonded. Where it is desired to avoid surface effects such as carbonation, the failure may be caused at some depth below the surface by partial coring, see Fig. 5. This method is particularly suitable where a tensile strength value is required, such as testing bonding of repairs, but may be

Fig. 3. Pull-out Lok. test.

84

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

3. Strength development monitoring Two major programmes involving eld testing have been undertaken to assess the suitability of different test methods for early age testing. 3.1. Cooling tower study This was aimed primarily at cooling tower construction with site work at Drax in North Yorkshire, supported by extensive laboratory work. The objective was to enable formwork to be removed with condence, even in cold weather, to enable construction schedules to be maintained. These studies demonstrated that surface hardness tests are unreliable at early ages, whilst ultrasonic pulse velocity measurements can yield good strength estimates, but usage is usually limited by the need for access to two opposite faces for reliable measurements. Penetration resistance Windsor Probe. is quick and suitable for large members such as slabs, but was shown to be unreliable at low strength values. Internal fracture tests are similarly unsuitable at early ages because of their high variability. It was concluded w5x that pull-out testing, maturity testing and temperature-matched curing are the most reliable and practicable techniques for use at low strength levels, although it should be noted that pull-off tests were not considered in this programme. Pull-out Lok. tests were conducted through cut-out panels in the timber shutters on 175-mm thick wall panels at ages commencing at 15 h . On site, tests were performed in-groups of three within 400 = 250-mm removable formwork panels. This dimension was constrained by formwork design considerations and limited the clear spacing between inserts to 165 mm. This is less than the minimum recommended by standards, but trials indicated that at the low strength levels associated with such tests no signicant inuence was detectable. In the laboratory, wall panels 1.5 m high and 2.0 m long with the same thickness, and identical reinforcement were fabricated. In this case pull-out tests were made through individual circular cut-out panels at each test point. Correlation tests were made on 225-mm cubes to give adequate edge distances, in conjunction with 150mm cubes for crushing at strengths as low as 1 MPa. Tests were performed on all six faces of the 225-mm cubes, and analysis showed that mean values were virtually identical to those obtained if the top and bottom surface tests were discarded. Variability within the group of results was however, greater than when just the four side faces were considered. Correlations were achieved by testing the gravel aggregate mixtures at varying ages, and by varying the mixture designs. Results of these tests are shown in Fig. 6a in which each correlation point represents the average of six

Fig. 4. Pull-out Capo. test.

correlated empirically to compressive strength for a particular mix. Many forms of this test are available commercially w3x but that most widely used in the UK is the Limpet system developed at Queens University, Belfast. 2.5. Temperature measurement methods Although outside of the scope of this paper, attention must be drawn to the maturity method ASTM C 1074. in which insitu strength may be estimated on the basis of a time and temperature maturity function for a specic concrete mixture. This is particularly suitable for strength estimation at early ages and most commonly utilises a temperature sensor located at an appropriate position within the pour, although chemically based inserts COMA-meter. are also available. In the UK, attention has also been paid to temperature matched curing BS 1881 pt 130., in which cubes are stored in a water-tank whose temperature is controlled by a sensor in the pour to ensure an identical temperature history to the insitu concrete. Both of these methods suffer potential practical disadvantages including vulnerability to power supply, failure or vandalism w5x, but may be particularly useful when used in combination with other tests, the pull-out method in particular.

Fig. 5. Pull-off test.

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

85

surprising to note that the greatest discrepancies, which cannot be easily explained, occur with the laboratory wall panels, whilst on site no strengths of less than 10 MPa were encountered despite the cold weather. A key feature emerging from the insitu temperature measurements was the extent of the in-place differentials across the height of the wall. The peak temperature was found to occur at approximately mid-height of the 1.5-m high lift at approximately 12 15 h after casting. This was typically 20 C higher than that experienced 100 mm below the top surface of the pour, and will lead to signicant early-age strength differences. 3.2. Cardington study

Fig. 6. a. Laboratory pull-out strength correlation. b. Site and laboratory wall pour pull-out results.

Lok-tests and three cube compressive strengths. For strengths up to 10 MPa, all points lie within 1 MPa of the mean line, but at higher strengths the sensitivity of the pull-out force to changes in cube strength is reduced and scatter increases. A further feature of concern was that the time taken to carry out a test at very low strengths might be signicantly less than that specied by standards. This was addressed by comparative tests on 225-mm cubes which indicated that neither magnitude of force nor variability will be affected provided the load rate is such that the test exceeds 20 s. Wall-pour results shown in Fig. 6b represent the average of six Lok-tests together with cube strength values estimated from maturities computed from appropriately-located temperature sensors using a temperature-time factor with y11 C datum w2x. The generally close agreement between both laboratory and site wall pour results and correlation tests, especially at very low strength levels, can be seen and is most encouraging given the uncertainties introduced by the use of maturities in determining insitu strengths. It is

This recent study has been based on the European concrete building project at the Cardington laboratories of the building research establishment in the UK. This project involved the construction of a seven-storey, insitu reinforced at-slab building frame, Fig. 7, utilising a range of construction methods with the aim of improving speed, reducing costs and improving the quality of such construction. See http:rrwww.bre.co.ukrbrercardingtonrcardlab1.html for further details on this project. The structure provided the basis for a number of construction-phase research projects including that undertaken jointly by the University of Liverpool and Queens University, Belfast w6x. In this project, pull-out testing was extended to include the capo-test and the pull-off test was also considered, supported by maturity measurements and temperature-matched curing to assist insitu strength estimates. Air-cured and water-cured cubes were also available, with results for early ages in addition to 28 days to support the aim of seeking to estimate 28-day strengths from early age insitu tests, as well as enabling comparisons of insitu strengths with cubes experiencing different curing regimes. A range of concrete mixtures incorporating different aggregate types gravel and limestone. and admixtures plasticiser and superplasticiser . with nominal cube compressive strengths of 37 and 85 MPa were used. The six different mixtures covered in this study are detailed in Table 2. Some adjustments were made during construction by the concrete supplier to maintain target mean strengths C37N-10 and C37N-11., as is common practice during construction of a large project. Tests were performed on columns at different heights top, middle and bottom., and on slabs, both adjacent to columns and in mid-bay top and soft.. The selected test methods were each used at similar locations to permit comparison of their practicality on site; for example speed, relative cost, disruption and accuracy,

86

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

Fig. 7. Cardington insitu concrete building.

and to assist determination of an optimum balance between insitu and cube testing. The insitu tests using the pull-off, lok and capo
Table 2 Concrete details

methods, were undertaken at 1 day or as soon as practicable ., 3 days, 7 days and 28 days after casting. Temperature measurements were available from the

Concrete types and dry batch weights per m3 . C85MS Microsilica Location Portland cement kg. Sand kg. 5 20 mm coarse aggregate kg.. Limestone Gravel Admixtures ml. Plasticiser Superplasticiser Cement replacements kg. Microsilica Metakaolin Free WrC Target Slump mm. Flow mm. Columns 400 732 C85MK Metakaolin Columns 400 717 C37N-10 Normal Columns 380 755 C37N-11 Normal Columns & Slab 355 785 C37P Plasticised Slab 330 815 C37F Flowing Slab 335 810

1170

1146 1025 1025 1010 1010

1232 8800

1232 8800

990 5000

40 40 0.25 0.32 0.50 0.53 0.52 0.52

100 550 600

100

100 550

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

87

Fig. 8. Lok-test inserts. a. High strength Lok-test insert. b. Flotation cup.

time of casting at several locations on columns at a depth of 25 mm below the surface, and on slabs at 25 mm above softs and 50 mm below top surfaces. Results for corresponding temperature-matched cured cubes were based on sensors at a depth of 50 mm below the tops of slabs midway between columns. High-strength Lok-test inserts were used for the highstrength mixtures, whilst standard otation cups were used to locate inserts on the top surfaces of slabs Fig. 8.. 3.3. Lok-test strength correlations Historically, most strength relationships have been assumed to be straight lines, and ordinary least-squares analysis has been used to estimate the corresponding slopes and intercepts, as shown in Fig. 9. Each point represents the average of four no. pull-out tests on the side faces of 200-mm cubes and three No. 100-mm cube compressive strength values. Results for individual groups of similar mixtures show small scatter about their respective correlation lines but the intercepts and slopes vary considerably for each individual concrete mixture. The combined correlation for all mixtures is however, surprisingly very close to the manufacturers correlation w4x. This is consistent with previous assertions that a generalised correlation can be used, but that condence limits can be improved by specic correlation for the mixture concerned w7x. The manufacturer does recommend that, in order to improve the accuracy of the strength correlation, the range of concrete strengths considered should be as wide as possible. This, for commercial buildings, which are unlikely to use so many different concretes, can be achieved by testing at various ages.

The combined strength correlation crosses the manufacturers recommended correlation at a force of 10.5 kN but because of its different slope, results in an estimated in-situ strength which is greater by 4.8 MPa at a force of 37 kN. Beyond this force the manufacturers recommended bi-linear correlation has the same slope as the Cardington correlation but the compressive strength intercept is less by 4.8 MPa. It was found that grouping all the results together to obtain one strength correlation for all the mixtures that were tested improved the condence in the strength prediction; the 95% condence interval was "4 MPa for a concrete strength of 43 MPa. This single strength correlation, based on all the results, was therefore, used for estimating the in-situ strength from measured lok test values.

Fig. 9. Lok-test strength correlations.

88

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

Fig. 10. Lok and capo-test strength correlations.

Fig. 11. Pull-off test strength correlations.

3.4. Capo-test strength correlations The possibility of using the lok-test correlation to interpret the capo-test results has been investigated. Correlations were developed in the same manner as for the lok-tests and results have conrmed that one strength correlation can be used for both the lok and capo-tests Fig. 10., even for ages as low as 1 day. The slight deviations of the two individual correlations appear to be due to the inclusion of high-strength concrete mixes for determining the strength correlation for the lok-test. Only normal-strength concretes were tested with the capo-test because of the limitations of the equipment mentioned below. 3.5. Pull-off strength correlations The ordinary least-squares method has again been used to determine the strength correlations which were obtained from six no. side face measurements on 150mm cubes. These are shown graphically in Fig. 11. It is apparent that the accuracy of these strength correlations has been affected by the limited number of sets of data, three for each mix, which was due to the difculties encountered when testing concrete at one day. It was found that the epoxy-bonding compound did not harden quickly enough to allow one-day testing. It is clear that a specic correlation is needed for each mix but if this is available, then a reasonable level of strength prediction can be achieved. 3.6. Insitu measurements Insitu lok-test strength predictions for slabs are compared with temperature-matched cube results in Fig. 12 based on the average of four no. lok-tests at a particu-

lar location together with between two and four no. compressive-strength cubes. It should be noted that lok-test results are limited to those available from a location adjacent to the controlling sensor but include results for both top and soft of the slab, thus, signicantly increasing variability. It was noted that insitu strengths, especially at the soft, were generally found to be greater in the slab region adjacent to columns than at mid-bay. A much closer relationship is seen in Fig. 13 between insitu predictions from lok-test and maturity results using a similar function to that used in the cooling tower study. which serves to conrm the reliability of the lok-test values. Comparisons of insitu and air-cured cube strength results in Fig. 14 however, clearly indicate the inadequacies of reliance on air-cured cubes at early ages, at both low- and high-strength values.

Fig. 12. Insitu strength Lok-test. vs. temperature-matched cube results.

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

89

Fig. 15. Lok-test strength correlations for high-strength concretes. Fig. 13. Insitu strength Lok-test. vs. insitu strength maturity..

4. Testing mature concrete 4.1. Normal concretes The applications of the range of test methods covered by this paper to normal strength concretes with natural aggregates have been well documented, and reliabilities and limitations established w1,7x, based on extensive work at Liverpool and elsewhere. 4.2. High strength concrete Published results for partially-destructive tests applied to high-strength concrete are very limited, although it has been suggested that probe penetration, internal fracture and break-off methods are unsuitable. Thus, pull-out and pull-off are the only realistic approaches, and both have been assessed, to a limited extent, in the Cardington project described above. The reliability of the lok-test has been conrmed up to 110 MPa but the capo-test is limited to an upper value of

approximately 70 MPa due to the characteristics of the pull-bolt system. Results suggest that the differences in mixture components and composition have no signicant effect upon pull-out forcercompressive-strength correlations. Further studies at Liverpool have conrmed the absence of inuence due to aggregate type graniterlimestone . over the whole of the strength range, and this is conrmed by Price and Hynes w8x to also include gravel aggregates. This is illustrated in Fig. 15. Price and Hynes suggest however, that the pull-out test is less sensitive at high-strength levels and attribute this to possible changes of failure mechanism linked to high paste content, good paste-aggregate bond, and aggregate fracture effects. Pull-off testing requires correlations for the specic mixture, with the maximum strength again limited by equipment capacity to approximately 100 MPa. 4.3. Lightweight concretes An extensive laboratory study utilising large scale 2.2 m= 0.5 m= 0.3 m. beam specimens together with

Fig. 14. Insitu strength Lok-test. vs. air-cured cube strength.

Fig. 16. Pull-out test strength correlations for lightweight concretes w9x.

90

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

accuracy of strength estimation was generally comparable or improved. Insitu variability w10x was however, found in some aspects to be higher than expected with normal concretes, with surface zone strengths typically 20% less than interior strengths, compared with 5 10% expected for gravel concrete. Different relationships between insitu strengths and standard 28-day cube strength were also noted. It is thus, particularly important that these issues are considered during planning and interpretation of testing on lightweight concrete. 4.4. Testing repairs This is emerging as a key application of Pull-off testing, utilising partial-coring into the substrate to permit an assessment of the tensile bond-strength between substrate and repair. This is dealt with by a CIRIA Report w11x and a minimum bond-strength may be specied as part of acceptance criteria w12x. A key issue that has emerged from studies at Liverpool University w13x is that of disk stiffness, as represented by proportions and material, relative to that of the material to which the disk is attached. This is illustrated by Fig. 18 in which it can be seen that a greater ratio of disk thickness to diameter is needed when the disk is made of aluminium rather than steel. To achieve comparable measured pull-off strengths, a minimum depth of partial-coring of 20 mm has also been recommended for commonly used 50 mm diameter aluminium disks to ensure uniformity of stress distributions at the base of the cored portion, although this may be reduced for steel disks. These results were obtained from extensive laboratory experimental studies, supported by nite-element modelling. It is thus, particularly important that only standardised equipment is used or alternatively, that the conguration is specied together with the minimum acceptable test value. A laboratory-based study in the USA w14x has also conrmed the suitability of the method for assessing the bonding of highway overlays, and the effects of freezing, thawing cycling and de-icing salts.

Fig. 17. Pull-off test strength correlations for lightweight concretes w9x.

cores and cubes has demonstrated that all methods considered can be successfully applied to lightweight concretes w9x. Correlations with strength differ from those for dense concretes in all cases and for some, tests are further affected by the nature of the aggregate and the type of ne material sand or lightweight. used. Windsor probe results using a special larger diameter probe. show increasing scatter with strength level, whilst internal fracture tests are heavily-dependent on loading method torquemeter or direct-pull. and are unsuitable for very soft aggregates. Lok-test results Fig. 16., show the closest agreement for different types of lightweight aggregate. Surface pull-off correlations are shown in Fig. 17, with clear differences according to aggregate type and it was found that correlations for partiallycored tests were signicantly different in all cases. Although a specic correlation is needed for any given situation, it was found that variability of results was in many cases less than to be expected with concrete made from natural dense aggregates, and the

5. Discussion The results presented have conrmed that pull-out and pull-off tests can be successfully applied to concrete in structures to yield estimates of insitu strength. The pull-out method in particular offers advantages of being able to utilise a general strength correlation as provided by the manufacturer. Concrete mixtures used on site may not be nalised until just before the start of construction, thus, hampering the development of specic correlations. Similarly, with existing concrete, precise details of the composition may be unknown and

Fig. 18. Effects of pull-off disk characteristics w13x.

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92

91

expensive, disruptive coring may be the only effective means of producing a specic correlation. For new construction, the lok-test has been shown to be easy-to-use and reliable over a very wide range of strengths. If it is possible to develop a specic strength correlation, possibly during construction, this may be used to improve the accuracy with which insitu strength estimates may be made. For lightweight concretes a general correlation based on Fig. 16 can be used, noting that this is different to the manufacturers correlation which relates only to concretes made with dense natural aggregates. When testing is required on existing structures windsor probe, pull-off and capo tests may be considered. Although simple to use, the windsor probe requires specic correlation for the aggregate and doubts have been cast upon reliability both at low strength levels and above 50 MPa. Pull-off tests require surface preparation and careful bonding, but are otherwise quick and straightforward to use. Specic correlations must be developed for the concrete under investigation, but the work at Cardington has conrmed that concrete with strengths up to 100 MPa can be successfully tested with strength estimation accuracies generally of the order of " 10% based on the average of six no. tests. This value is comparable to that associated with pull-out tests four no.. using specic correlations. Capo-testing is not simple and requires considerable skill to successfully perform the coring and under-reaming operations. Nevertheless, it is most encouraging that work reported here conrms the ability to use the same general correlation as the lok-test with similar likely accuracies "20% based on four no. results .. This method thus, offers the signicant benets noted above associated with the ability to avoid the development of a specic strength correlation. It can also be used as a back up to the lok-tests, if required, but the need to use a covermeter to avoid reinforcement should be noted. The application of statistics to assess the reliability of insitu strength estimates is fraught with difculties. This issue has been considered at length by Carino and others in the USA with ndings summarised by ACI 228 w2x. The subject has also been considered by RILEM Committee TC 126-in place testing. Unfortunately, there are seldom enough in-place test results to establish statistical condence, whilst the number of strength levels available when preparing a strength correlation will have a large effect upon the uncertainty of estimated strength. ACI 228 w2x recommend that six to nine strength levels should be used in performing correlations, providing a balance between precision and cost. Work at Cardington has highlighted the difculties of achieving this in practice, even on a carefully-planned project. Where it is necessary to develop correlations from cores taken from an existing structure, the difculties escalate. Few reports exist of

reliable and systematic comparisons of in-place tests with cores taken from structures w4x, thus, it has to be recognised that commonly quoted accuracies are effectively based on condence limits typically 95%. of laboratory correlations. It may well be that it would be more realistic to quote lower-bound estimates of insitu strength. The work described above has further highlighted the signicance of insitu strength variations due to location within an element. This effect is especially marked at early ages due to temperature differentials, and must be considered when establishing test locations. Care must be taken when using large correlation specimens eg. 200-mm cubes. to ensure that temperature histories match those of specimens used for crushing eg. 100-mm cubes.. The use of early age tests to predict later age strength is potentially a valuable application of testing. Analysis of results from the Cardington project continues, but preliminary ndings suggest that predictions based on lok-tests at 3 days may be sufciently accurate to identify situations requiring remedial action. If combined with maturity measurements, condence in later age predictions can be increased. It is clear from the Authors experience that simplicity is the key to increased usage of in-place testing. Combining methods can be valuable, especially at early ages eg. maturity and pull-out. but can seldom be justied except on major projects. Bickley w15x has demonstrated the economic benets of in-place testing during new construction, and experience both at Drax and Cardington has conrmed the value of testing in conjunction with formwork renewal and speed of construction. It is in this area that the greatest scope for increased usage lies at present, since in most parts of the world there is little will to move away from cube or cylinder testing for acceptance purposes.

6. Conclusions It has been demonstrated that partially-destructive tests provide a viable approach to estimation of insitu concrete strength for a range of concrete types and circumstances. Cast-in pull-out tests have been shown to be particularly suitable, both in terms of ease-of-test and reliability, and for early-age testing. Pull-out and pull-off tests may also be reliably applied both to lightweight and high-strength concretes in structures. The pull-off test, with partial coring, is valuable in assessing the bonding of repairs but results may be inuenced by the test conguration. Development of a reliable correlation with compressive strength is a critical aspect inuencing reliability of insitu strength predictions, which are difcult to dene statistically. Use of results of inplace testing must fur-

92

J.H. Bungey, M.N. Soutsos r Construction and Building Materials 15 (2001) 81 92 strength assessment of concrete-the European concrete frame building project. In: Uomoto T, editor. Proceedings of the International Conference on N.D.T. in Civil Engineering. Elsevier, 2000: 583 597. Bungey, JH. Partially destructive methods of assessing the insitu strength of concrete. Proceedings of the Institution of Civil Engineers-Structures and Buildings, Volume 99, 1993: 64 7. Price WF, Hynes JP. Insitu strength testing of high strength concrete. Mag. Conc. Res. 1996;48176.:189 97. Bungey, JH, Madandoust, R. Evaluation of non-destructive strength testing of lightweight concrete. Proceedings of the Institution of Civil Engineers-Structures and Buildings-Vol. 104, 1994: 275 83. Bungey JH, Madandoust R. Strength variations in lightweight concrete beams. Cem. Concr. Compos. 1994;16:49 55. McLeish A. Standard tests for repair materials and coatings for concrete-part 1: pull-off tests. London: CIRIA, 1993:41. Patch repair of reinforced concrete. Technical Report No.38. Slough: Concrete Society, 1991:87. Bungey JH, Madandoust R. Factors inuencing pull-off tests on concrete. Mag. Conc. Res. 1992;44158.:21 30. Lee S, Frantz GC, Stephens JE. Application of pull-off test to assess the durability of bond between new and old concrete subjected to deicer salts. A.C.I. 1997;SP 168-13:267 94. Bickley, JA, Hindo, KR. How to build faster for less-the role of in-place testing in fast track construction. A.C.I. Spring Convention, 1994. J.H. Bungey holds a MSc degree in concrete structures from Imperial College, London and a PhD from Liverpool University where he is currently Professor and Head of the Civil Engineering Department. His research interests encompass concrete materials, design and performance including nondestructive testing. He has authored numerous papers and books on these topics. ACI Member M.N. Soutsos received his BEng and PhD degrees from the University College London. He is currently a lecturer in the Department of Civil Engineering, The University of Liverpool. His research interests include: cement replacement materials, thermal stresses in concrete, nondestructive testing, as well as mix design and rheological properties of high strength concrete.

thermore, take account of insitu strength variations including surfacerinterior effects. Results conrm that the drilled version of the pull-out method capo-test. produces results which may be regarded as comparable to the cast-in version lok-test. and this may be a worthwhile method to use when assessing unknown existing concretes. There is considerable scope for future development and usage related to new construction, where test simplicity and reliability are the key issues. Long-term strength predictions from early age partially destructive tests alone are not yet, however, regarded as sufciently accurate for acceptance purposes.

w7x

w8x w9x

w10x w11x

Acknowledgements Thanks are due to Professor A.E. Long and Dr. G.D. Henderson of the Queens University of Belfast, for their collaboration relating to the Cardington project, and to Dr. R. Madandoust of Gilan University, Iran, for his contributions to several aspects of the work reported in this paper. References
w1x Bungey JH, Millard SG. Testing concrete in structures-third edition. Blackie, 1996:286. w2x American Concrete Institute. In-place methods to determine concrete strength, ACI-228, 1.R, 1995. w3x Bungey JH. Testing concrete in structures-a guide to available equipment. London: CIRIA, 1992:87. w4x Petersen, CG. Lok-test and capo-test pull-out testing-twenty years experience. In: Bungey JH, editor. Proceedings of the International Conference on N.D.T. in Civil Engineering. British Institute N.D.T., 1997:77 96. w5x Bungey JH. In: Boogaard J, Van-Dijk GM, editors. Monitoring concrete early age strength development. Non-destructive testing. Amsterdam: Elsevier, 1989:1243 48. w6x Soutsos MN, Bungey JH, Long AE, Henderson GD. Insitu

w12x w13x w14x

w15x