Abstract The performance of a bored piled foundation is affected by the construction quality of the individual piles comprising the piling system.
For the purposes of quality assurance, a programme of non-destructive integrity testing (NDT) is usually implemented on a specified percentage of the working piles. Traditionally, the proportion of piles to be tested is decided upon by the foundation designer based on site-specific knowledge, engineering judgement and his prior experiences of NDT. This paper shows how the use of statistics and probability theory may help influence this decision.
Introduction Bored cast insitu piling is regularly specified in the UK to support the ever-taller structures that are being built in urban locations. However the high degree of contractor workmanship, experience and supervision necessary in construction, coupled with the naturally variable soil and groundwater conditions on site, makes bored cast insitu piles particularly vulnerable to structural faults and variable in construction quality (Thorburn and Thorburn, 1977).
The subterranean nature of their formation makes direct, visual inspection of the final product impossible in most cases. Faced with uncertain quality, it is unknown whether the piles will be suitable for their intended operational function. Therefore to verify satisfactory construction and detect any defective piles that may exist, non-destructive integrity testing is usually implemented on a specified percentage of the pile population.
Blanket inspection of every pile is the most conservative yet widely adopted quality assurance (QA) strategy in the UK and internationally.
However, when this is impossible, sampling inspection (eg testing less than 100% of the piles) is conducted. Due to a lack of codified recommendations or quantitative guidance, the number of piles tested in such a case is traditionally decided upon by the foundation designer. At its best, this decision is based on detailed construction observations, site-specific knowledge and engineering judgement; at its worst, this may be little more than a token number of piles chosen purely to satisfy contractual QA obligations.
Furthermore, due to the indirect nature of their examination, few methods of integrity testing are infallible. A misclassification of the true pile quality caused by an inaccurate integrity test or an inaccurate interpretation of a test result can either lead to a defective pile going undetected (false-negative result) and being mistakenly incorporated into the foundation or a sound pile being wrongly condemned (false-positive result) and subjected to unnecessary inspection or remedial work.
Incorporating defective piles into the foundation system may adversely affect its performance, safety or durability, and depending on the redundancy offered by the piling layout and the robustness of the superstructure, result in failure with consequential remedial costs or even loss of life.
Alternatively, implementing unnecessary inspection or remedial work on sound piles may cause defects to be inadvertently formed where there were none, or adversely affect project economy by increasing construction cost and programme of the foundation works.
Disputes about the cost of remedial works are common between clients or main contractors and their piling subcontractors when further examination shows a pile with an apparent defect to be subsequently sound.
Therefore, after implementing a QA programme of integrity testing on up to 100% of the working piles, using an integrity testing system that may be less than 100% reliable at detecting defective and nondefective piles, how confident can a foundation designer be in the construction quality of the entire piling system, and thereby the final piled foundation? Also, what is the minimum proportion of piles that should be tested on a project, under certain conditions, to achieve an acceptable level of quality assurance?
This paper shows how the use of statistics and probability theory may help answer such questions. Through the use of a statistical approach, a foundation designer may be able to quantify their level of confidence in the foundation quality, in addition to deciding on an efficient QA programme of integrity testing. Previously developed statistical approaches have not taken into account the inaccuracy of the integrity testing process and the likelihood that a wrong diagnosis could be made regarding the true pile quality. This paper presents and discusses the results of analyses performed using a more detailed statistical sampling approach, which considers the reliability of the integrity testing system.
Bored piling and the occurrence of defects In general, piling contractors produce bored cast insitu piles to the best of their ability, following industry best practice to reduce the likelihood of faults. Nevertheless faults can still occur both during and after construction in a small proportion of a pile population, with the factors in Table 1 influencing the number of piles affected on a given site.
Thorburn and Thorburn (1977) and Fleming et al (1985) provide a comprehensive review of the numerous types of piling fault that can occur in bored piles along with their variety of causes.
The presence of a piling fault does not necessarily imply that a pile cannot be used for its intended purpose. Once constructed, there must be a belief that each and every pile satisfies three general design criteria:
Meet serviceability limit state requirements and support a proposed working load without settling by more than is permissible.
Meet ultimate limit state requirements and carry a specified load with an adequate factor of safety against failure.
Remain durable throughout the design life of the structure, such that the pile is not affected by time dependent activities like corrosion of the steel reinforcement or chemical attack of the concrete, which would affect its ability to continue satisfying the first two criteria.
Some faults that occur will be more serious than others, given the site specific operating conditions of a particular foundation. However, any fault that is likely to adversely affect performance, safety or durability of a pile, in either the short or long term, should be considered a defect, with a pile containing one or more defects being classified as defective.
Some information on the frequency of defective bored piles, identified through NDT inspection, can be taken from the results of surveys published in the technical literature. Davis and Dunn (1974) report 9.7% defective out of a total 717 piles tested on five projects; Fleming et al (1985) found 1.5% defective out of a total 5,000 piles tested and 1.9% defective out of a further 4,550 piles tested; Ellway (1987) reports 4.2% defective of a total 4,400 piles tested; Thasnanipan et al (1988) state 3.3% defective of a total 8,689 piles tested; Lew et al (2002) report 7% defective within a population of 380 piles tested and 1.5% defective of a total 5,000 piles tested. Preiss and Shapiro (1981) suggest that approximately 5% to 10% of the piles on a project could be defective.
Obviously, there are numerous factors influencing the quality of the piling work in each case, some of which may be noted in Table 1, as well as the specific type of integrity test used, the reliability of the conclusions drawn and the threshold definition for defective pile adopted. However the survey results highlight the fact that, in general, defective bored piles are quite a rare occurrence, nevertheless may still represent a significant number of the piles on an individual site.
Redundancy offered by the piling layout The reliability of a piled foundation (eg its ability to perform as intended under operating conditions (Davis, 1999)) is not only affected by the construction quality of each pile in the population but also the layout of the piling system. Figure 1 shows the layout of a typical piled foundation. The pile population comprises 100 piles, with five piles forming each pile group and 20 pile groups forming the piling system. Normally, a project team's requirements of a piled foundation are that it must be constructed as quickly as possible, as cheaply as possible and perform adequately in operation (Chapman and Marcetteau, 2004).
In the UK piling industry the use of single large diameter bored piles is now often favoured to support a column instead of several smaller piles in a group. In many cases the direct piling costs associated with this type of layout are greater than that for a pile group, but the substructure costs attributed to a system of single piles are generally less, due to pile caps being unnecessary. The main reason for choosing single piles is that the substructure tends to be constructed more rapidly so long as the rate of concrete supply to the site is not a constraining factor. However as shown in Table 2, the consequence of a defect in any individual pile depends on the redundancy offered by the piling layout. Larger pile groups are inherently more redundant than smaller groups or single pile groups due to the reserve capacity offered by neighbouring piles and the ability to redistribute load through the soil and pile cap (Zhang et al, 2001; Paikowsky, 2003).
Table 2 illustrates the likely consequences of a defect causing a loss of pile capacity, when single and multiple piles are favoured for the foundation layout. Some pile groups may tolerate the presence of a defective pile without being significantly adversely affected, due to a higher level of redundancy being offered. Paikowsky (2003) defines a non-redundant group as one for which failure of a single pile will adversely affect the structural column its supports, with limited or no ability of other piles supporting the same column to mitigate its effects.
Defects are particularly critical for single large diameter piles because if such a non-redundant pile is unable to support its working load this may lead to excessive settlement of the column it supports, thereby causing significant distortions in the superstructure. Thus, as the level of redundancy offered by a pile group diminishes, it becomes more crucial to assure the construction quality of every pile to avoid foundation problems.
Pile testing There are two fundamentally different uncertainties associated with foundation piles that tend to be investigated through testing:
Pile capacity: Will the pile satisfactorily withstand a specified loading?
Pile integrity: Is the pile of the correct dimensions and structural quality?
Bored pile capacity is traditionally determined through static load testing. The load test directly examines pile performance and establishes whether there is enough strength provided by the pile and enough resistance provided by the surrounding ground to support an applied load with an acceptable amount of settlement (Fleming et al, 1985).
Pile integrity is traditionally determined through non-destructive inspection. An NDT test indirectly examines pile construction quality and determines whether the insitu pile is of the necessary specification with correct dimensions, material homogeneity and free of structural faults.
The direct cost of static load testing coupled with the programme disruption caused by the test frame's presence on site makes it prohibitively expensive to test more than a token proportion of piles. As such, load tests are limited to about 1% to 2% of the working piles (Turner, 1997).
However, non-destructive testing is rapid and cheap in comparison, thereby making it feasible to inspect a large proportion, if not every pile on a given site.
Integrity testing guidance No codified recommendations are given in any UK piling specification (ICE, 1988a, 1988b, 1996) or Eurocode 7 (1997) on what should be the minimum number or percentage of piles integrity tested using non-destructive techniques. However in Ciria Report 144, which is one of the principal up-to-date references and industry guides for pile integrity testing, Turner (1997) makes reference to a statistical approach developed by Preiss and Shapiro (1979, 1981) that aims to give quantitative guidance on the level of testing necessary under certain conditions.
Preiss and Shapiro's statistical approach indicates what proportion of piles should be integrity tested to achieve confidence of at least 90% in there being less than a certain number of defective piles among both the tested and untested piles in the population. 'If 10 out of 100 piles are allowed to be defective and in the tested sample no defective piles are detected then 20 piles should be tested' [Preiss and Shapiro, 1979].
A red dashed line in Figure 2 illustrates this point. Similarly, by testing 70% of the pile population and detecting two defective piles, the probability (or risk) of there being more than six defective piles in the population is less than 10%. Therefore in this case, if the foundation can tolerate six defective piles and the engineer accepts the 10% likelihood of there being more than the tolerable number defective, then 70 out of 100 piles should be tested. In general, to achieve this 90% level of confidence in there being no more than a tolerable number of defective piles in the foundation, the number of piles that need to be integrity tested:
increases as the number of defective piles detected in the sample increases, for a given proportion of defective piles tolerable;
decreases as the tolerable proportion of defective piles increases, for a given number of defective piles detected in the tested sample.
Probabilistic and statistical techniques can be useful in dealing with problems of quality assurance in geotechnical engineering (NRC, 1995). Weltman (1980), Preiss and Shapiro (1979, 1981) and Paikowsky (2003) have all proposed statistical approaches to determine a suitable number of piles to be tested for quality assurance. However, the methodologies have assumed the physical pile test and the interpretation of the test result to be 100% reliable, without accounting for the occurrence of a falsenegative or false-positive inspection error. Clearly, this is not the case in practice as the results of non-destructive tests are notoriously difficult to interpret, making it impossible to be certain in every case that the correct conclusions have been reached.
Reliability of integrity testing The most widely specified forms of non-destructive examination in the UK are:
External: Sonic Echo (SE) and Transient Dynamic Response (TDR) Internal: Cross-hole Sonic Logging (CSL) These are known as indirect testing methods as they infer structural features of a pile from its acoustic response to the test. Weltman (1977) and Turner (1997) review the various NDT methods for assessing pile integrity along with their attributes under different operating conditions. With these methods, the indirect nature of the integrity inference means that a high degree of judgement and subjective interpretation is required when evaluating the structural quality of each pile tested.
Integrity testing reliability can be defined as the degree to which both the evidence provided by an NDT test and the interpretation of that evidence corresponds with the true state of a pile insitu. As such, an inaccurate claim of (in)sufficient pile integrity may arise through improper test application or a lack of knowledge and experience in data interpretation. Hertlein (1998) has examined the reliability of different integrity testing methods through numerous case histories.
Furthermore, a detailed discussion of the factors influencing the accuracy of internal and external techniques has been provided by Davis (1999). Some factors influencing the reliability of testing conclusions are given in Table 3.
Clearly, the appropriateness of the test method for the particular pile features being searched for is vital. SE and TDR techniques are excellent at detecting transverse cracks, which can have little engineering significance, so may thereby yield many false-positive results [Lew et al, 2002]. TDR tests tend to be more detailed than SE tests due to the instrumented hammer used to strike the pile head recording more quantitative data, and the ability to analyse the results in both the time and frequency domain.
On the other hand, CSL techniques are better at detecting real defects, however they are considerably more expensive and time consuming and so tend to be reserved for large-diameter piles supporting expensive buildings, where the consequence of defects occurring are more severe.
When interpreting an integrity test result, the threshold beyond which a pile is classified as defective is particularly crucial. Setting a tight threshold may produce many false-positive results, whilst a generous threshold could provide numerous false-negative results, that is, defective piles slipping through the screening process. It is not uncommon for NDT contractors, which are normally subcontractors to the piling contractors, to come under commercial pressure to justify thresholds that are seen as too tight (excessively conservative). For this reason, industry-wide standardisation of test methods and threshold criteria based on quantitative assessments is to be encouraged.
Although there is evidence to suggest that NDT methods can on occasion, be incorrectly specified, misused or misinterpreted, they remain a valuable tool for quality assurance of bored cast insitu piles once they are built (Hertlein, 1998). They are particularly useful when coupled with accurate piling construction records for highlighting piles that are more likely to be defective.
Background to statistical analysis By inspecting every pile on site using an integrity test that is perfectly reliable, a foundation designer can be certain of detecting all the defective piles that may exist and, after their repair, be absolutely confident in the structural quality of the foundation. However this is an idealised situation, as although it is possible and often favoured in practice to implement a 100% testing strategy, the integrity testing conclusions are rarely 100% reliable as previously discussed. Therefore a more detailed statistical methodology may be developed by considering the reliability of the integrity testing system used (Cameron et al, 2002).
The sequence of events associated with pile construction and integrity testing, accounting for the possibility of an inspection error, are modelled in Figure 3. Foundation quality is expressed by the number of defective piles remaining, and thus incorporated into the final piling system, after the implementation of a programme of integrity testing and subsequent repair.
As shown in Table 2, it may be possible for a foundation to tolerate a number of defective piles without being significantly adversely affected, due to the redundancy offered by the piling layout. Therefore, to assure foundation quality through a programme of non-destructive testing, at least all but a tolerable number of defective piles must be detected. The degree of confidence in satisfying this QA criterion (level of quality assurance achieved) can be quantified through the use of a statistical procedure. Referring to Figure 3, the statistical analysis answers the question: What is the probability of detecting at least all but a tolerable number of defective piles amongst the (nDt) apparently defective piles detected and subsequently repaired within a sample of (n) piles tested at random using an integrity test with a defective and non-defective detection reliability of 1-Efn and 1-Efp respectively, given that a total of ND defective piles are assumed to initially exist within a population of (N) piles prior to the programme of inspection and repair?
This degree of confidence is given symbolically as: Pr[dd (ND - TD)].
Discussion of results The results of statistical analyses, for a range of likely practical values, are given in Figure 4. Each graph shows the number of piles tested versus degree of confidence, within a population of 100 piles for a given number of defective piles in the population, testing reliability and tolerable number of defective piles.
Assume that the pile population contains 5% defectives. If the foundation can tolerate up to four defective piles (TD = 4) and the integrity testing system used is 100% reliable at detecting defective piles, then a totally competent testing specialist would be 90% confident in detecting at least one defective by testing 37 piles. However if the test is only 60% reliable then the same testing specialist would be only 71% confident. Furthermore, to achieve the same 90% level of confidence, with a 60% reliable test, 62 piles should be inspected. Thus, to achieve the same level of quality assurance with the less reliable integrity test, almost twice as many piles should be tested. Therefore, the results show that:
for a given number of piles tested, a greater degree of quality assurance will be achieved with a more reliable integrity testing system used;
a more reliable integrity testing system requires a smaller sample of piles to be tested to achieve a pre-determined level of quality assurance.
A programme of integrity testing is effective if the QA criteria are satisfied with an acceptable level of confidence and efficient if these objectives are accomplished with as little effort as possible.
If it is assumed that the cost and time required to test a pile is the same for each type of integrity test, then to achieve a predetermined level of quality assurance, the more reliable the test is the more effective and efficient the testing programme due to the smaller sample of piles to be tested.
However in reality, the cost and time to test a pile varies with the different NDT methods. For example, testing a pile using a SE or TDR method may only take a few minutes (10 to 20 piles tested per hour with good access (FPS, 1999)) and cost typically £3 to £5 per pile in addition to a site mobilisation fee and minimum testing charge, while a CSL test may take up to an hour to perform and cost hundreds of pounds depending on the length of the pile and the number of pre-installed access tubes.
Accounting for such factors in achieving a predetermined level of quality assurance, increasing the reliability of the integrity testing system improves the testing effectiveness, as a smaller sample of piles must be tested. However the testing efficiency does not necessarily improve due to the increased time and cost of testing each pile with a more reliable test method.
Most foundation designers like to believe that they are achieving high levels of confidence during integrity testing. However, even when a NDT programme is carried out on an entire pile population, a foundation designer can rarely guarantee that there will be absolutely no defective piles incorporated into the final foundation.
In fact, the results show that even if blanket inspection of every pile is performed, in the hope of achieving total quality (ie TD = 0), the inaccuracy in the integrity testing system only allows for low levels of confidence to be achieved. For example, assume that the population contains 5% defectives and the integrity testing system used is 80% reliable. If a programme of 100% inspection is carried out under these conditions, the likelihood of detecting all the defective piles in the tested sample is only 32%. Thus, on the basis of these data, foundation designers need to allow for either accepting:
lower levels of confidence in assuring total quality of the foundation; that the piled foundation should tolerate a number of undetected defective piles.
Furthermore, for a given level of confidence, the more defective piles that can be tolerated, then the less integrity testing is required. Thus, the amount of in built redundancy offered by the piling system (tolerable proportion defective) may influence the decision as to how many piles should be integrity tested.
It is the general recommendation of many (Turner, 1997; Paikowsky, 2003) that every pile on a given site should be post construction tested for defects. However, in practice it is often difficult to test 100% of the working piles due a number of factors including cost, time and access.
As foundation construction is usually on a project's critical path, any delays that may occur during pile testing can adversely affect the progress of all follow-on activities, the direct cost of which may exceed the value of the entire piling contract (Chapman and Marcetteau, 2004).
As such, there is a significant advantage in terms of cost and programme if some piles could be omitted from testing.
Consider a programme of inspection carried out on a population of 100 piles, assumed to contain 5% defectives, using an integrity test that is 80% reliable. Under these conditions, Figure 5 shows the increase in confidence (quality assurance) gained for every additional 10 piles integrity tested. It is shown that where the foundation can tolerate up to four defective piles, most benefit comes from the initial piles tested.
However, where the foundation cannot tolerate any defective piles within it, most benefit comes from testing the final piles. Therefore, the greater the reliance on each individual pile in a population, the more important it is to check the integrity of every single pile, even when performing 100% testing may be disruptive to the construction programme.
Figure 5 shows that in some cases the added value gained by the last piles tested could be judged by the foundation designer not to be worth the time and money involved in the testing. This is particularly true on large projects with a rapid construction programme, where the construction manager wants the groundwork contractor to follow on directly as the piling contractor leaves site.
Often on these projects, there is immense pressure not to integrity test the final say 20% of the piles to allow an earlier package handover.
The analysis shows that such an omission may be possible only if the tolerable proportion of defective piles is high. Hence, there could be programme advantages in selecting a more redundant foundation system.
Many engineers like to perform 100% inspection in the hope of detecting all the defective piles that may exist on a site. However, the results of statistical analyses have shown that even if every pile is tested, the inaccuracy of the integrity testing system used does not allow for high levels of confidence to be achieved.
Furthermore, in some cases testing slightly less than 100% of the piles may be a viable option, with little reduction in the level of quality assurance occurring but with potential benefit in terms of cost and programme savings. Nevertheless, it must be highlighted that the only way that this can happen is by selecting a more redundant foundation design.
Hence, the redundancy offered by supporting columns on a layout of pile groups, as opposed to single large diameter piles, may serve as a major reason for a decrease in the amount of integrity testing necessary for quality assurance of a bored pile foundation.
As in any modelling activity, assumptions and simplifications have been made to the real life pile construction and integrity testing sequence to model the problem in a statistical manner.
Notwithstanding, the use of these statistical results, supplementing experience and sound engineering judgement, may provide a more rational basis for deciding on a suitable level of integrity testing, thereby helping to improve foundation safety and construction efficiency.
The paper highlights the folly of absolute reliance on the results of integrity testing as the sole arbiter of quality in bored pile foundations. While integrity tests have a valuable role to play in the detection of defects, statistical analysis accounting for inaccuracies in the techniques, indicates that a high level of dependence should not be attributed to the results of integrity tests alone. Moreover, the best approach in ensuring quality of piled foundations is to follow the basic principles laid out by Chapman and Marcetteau (2004): good site investigation properly gathered together l careful design appropriate for the particular ground conditions as part of a coherent design process l appropriate choice of acceptable piling methods l clear specification and fair procurement of the piling contract l experienced contractor who has considered all the risks l independent supervision to verify that standards are maintained.
Adopting this doctrine should reduce the likelihood of defects occurring and maximise the likelihood that suspected piles will have already been identified prior to integrity testing. The findings of this paper reinforce these well-accepted principles.