Assessment of Several Interpretation Criteria for Tensile Axial Load Tests on Deep Foundations

Abstract

Axial tensile load tests on deep foundations are increasingly significant because of the expansion in their usage for a myriad of projects requiring support of tensile loads as well as for being a less costly alternative to compression load tests. A variety of interpretation criteria for compression load tests is available, which are often also used for tensile tests, but a few have been proposed specifically for tension load tests. In this study, the performance of seven of the available widely used interpretation criteria was evaluated using a data set of 80 tensile load tests. These criteria are (i) original Davisson, (ii) AASHTO, (iii) New York City building code, (iv) French building code, (v) 5% diameter, (vi) 10% diameter, and (vii) New York University (NYU) compression criteria. Their performance was evaluated for applicability, accuracy, precision, and the diameter and length effects. The AASHTO criterion exhibited the best accuracy and a moderate precision, which is promising. The study also presents efforts to identify a better performing criterion, which yields the introduction of the NYU tension criterion where capacity is defined as that corresponding to the smallest displacement of (i) half the elastic elongation of the pile plus 0.25 in.; (ii) pile head displacement corresponding to the first incidence of pull-out or strain-softening resulting in loss of more than 5% of capacity; and (iii) pile head displacement of 0.5 in., unless modified by the structural engineer of record. The NYU tension criterion achieved a similar accuracy to the AASHTO criterion and a better precision, which is optimal.

Keywords

infrastructure geology and geoenvironmental engineering foundations of bridges and other structures design drilled shafts in situ testing piles

Deep foundations have been increasingly used to support infrastructure projects and high-rise buildings, among other structures. These foundations are commonly subjected to compressive loading where they are expected to transfer the building load to deeper, stronger soil layers. However, deep foundations are occasionally used to resist other types of loads such as lateral or tensile loads. These loading conditions generally develop in tall structures, buoyant structures, or structures built on swelling clays ( 1 ). Therefore, understanding the behavior of piles under tension loading is increasingly becoming an important topic in deep foundation design, with pile load testing being the approach most often used in practice to verify the expected behavior of deep foundations under loading.

Load tests are frequently used to confirm design assumptions, assess soil strength, enhance reliability, evaluate the suitability of construction methods, and reduce costs. Load testing takes different forms ranging from static load testing to high rate of loading tests such as dynamic, Statnamic, and high-strain dynamic (Apple) load testing. To date, static load testing remains the gold standard as it is the least complex, requires the least assumptions to interpret the results, and does not require specialized equipment.

Compressive, tensile, and lateral loading are all standard techniques applied in static load testing. Each method serves a different purpose and measures different aspects of the pile’s performance, including compressive capacity, tensile strength, or resistance to lateral loading. Whereas compressive load testing has the widest acceptance and recognition, tensile load testing has a lower cost compared with compressive testing. Tensile tests require less complex loading as their reaction frames can bear on the ground. For this reason, some geotechnical engineers opt for conducting a tensile load test even if the pile will be carrying a compressive load, where the compressive capacity is extrapolated from the results of the tensile load test. Tensile load tests are also used on occasion to isolate skin friction from end bearing.

Interpreting load test results can vary substantially among engineers and is rather subjective. Although engineers can agree on defining the capacity when the pile fails, load tests generally do not lend themselves to immediate interpretation of capacity or nominal resistance ( 2 ). Therefore, a reliable interpretation criterion is needed to ensure the accuracy of defining the pile capacity whether in tension or compression.

In the past 50 years, a plethora of interpretation criteria has been proposed to aid with interpreting the capacity from individual load tests. Some were developed for specific pile types, whereas others were intended for certain pile sizes. The cross-application of the various criteria is not generally recommended, as the performance of a specific criterion is not necessarily proven for other pile types. However, little research is available for interpretation of tension load tests. In fact, a limited number of interpretation criteria were originally intended for tension load tests on deep foundations.

In this study, the performance of seven of the most widely used criteria was evaluated using a data set of 80 tensile load tests compiled from the Federal Highway Administration’s (FHWA) Deep Foundation Load Test Database (DFLTD) v.2 ( 3 ) and supplemented by an extensive literature search. The data set ranged in diameter and length and included five different pile types: (i) open-ended steel pipe piles, (ii) closed steel pipe piles, (iii) steel H-piles, (iv) drilled shafts, and (v) square concrete piles.

This study aims to present a side-by-side comparison of the performance of the available criteria for applicability, accuracy, precision, as well as diameter and length effects, which hopefully could provide practicing engineers with insights on how each criterion performs and which criterion would be best to use.

Load Test Data Set

A data set of tension load tests was compiled for this study from two sources. The first was the FHWA’s DFLTD v.2 that has been migrated to a relational database management system by Machairas et al. ( 4 ). The database contains a considerable amount of load tests of various types (tension, compression, and lateral) and loading techniques (static, dynamic, cyclic, and Statnamic) that were conducted on different pile types that range in diameter and length. However, DFLTD v.2 includes only 60 tension load tests out of nearly 1,200 static load tests, which was deemed too few for the study. To increase the size of the data set, the authors scoured the available literature and managed to collect an additional 48 case histories. In most of these cases, the load test data were only available as a plot; therefore, an online digitizer was used to convert these plots into a numerical format, increasing the total available load tests to 108.

Many of the available load tests were proof load tests that were conducted up to a multiple of the design load; thus, they did not meet the requirements of any interpretation criterion. To ensure the integrity of the analysis, piles that did not cross any criterion were excluded from the analysis, which resulted in the exclusion of 21 load tests, leaving 85 cases. Finally, piles smaller than 10 in. in diameter were also excluded as they were deemed inconsistent with current piling practice. This resulted in the exclusion of five more piles, which left 80 cases included in the analysis ranging between 10 and 90 in. (0.25–2.28 m) in diameter and 18 and 240 ft (5.5–62 m) in length. The distribution of the diameter and length for the data set is presented in Figure 1.

Available Interpretation Criteria

Unlike compression tests, interpretation criteria for tension load tests are scarce. To find tension-specific criteria, the authors conducted a thorough literature review, surveyed practicing geotechnical firms, and searched the building code for several countries. This yielded a handful of criteria that range in complexity and are in many cases commonly used for interpreting tension and compression load tests. The description of these methods is presented next.

Original Davisson

The original Davisson criterion gained wide popularity for different types of foundations since its development in 1972 ( 5 ), despite some long-standing concerns about its formulation ( 6 ). It was originally developed for small end-bearing piles in which the capacity is defined as the load corresponding to a settlement equal to the elastic compression (extension) of a column (PL/AE) plus the sum of settlements required to mobilize (i) end bearing (viz., quake = 10% of diameter) and (ii) skin friction (0.15 in.) expressed as the load corresponding to a settlement, Δ, given by

Δ = \frac{PL}{AE} + \frac{D}{120} + 0.15 in .

(1)

where D is the pile diameter, L is the pile length, A is the pile area, and E is the pile elastic modulus (all lengths are in inches). The original Davisson criterion can be objectively applied; however, its performance for some pile types has not been evaluated.

AASHTO

The AASHTO ( 7 ) criterion is a variation of the original Davisson criterion where the end bearing term is removed from Equation 1 as piles in tension do not experience any end bearing and rely solely on the skin friction. This criterion was originally recommended by Hannigan et al. ( 8 ), as a more faithful interpretation of Davisson’s intent. The capacity is defined as the load corresponding to a settlement, Δ, given by

Δ = \frac{PL}{AE} + 0.15 in .

(2)

New York City Building Code (NYCBC)

The NYCBC defines the allowable load as that corresponding to the lesser settlement of (i) the load corresponding to a settlement equal to the elastic shortening (elongation) of the pile (PL/AE) + 0.75 in.; or (ii) the load that causes a maximum net settlement of no more than 0.01 in. per ton of applied test load (net settlement = gross settlement from applied load minus rebound after removing the test load). Customarily, the first provision dominates in most cases, which was also the case in this study, whereas the second prevailed in 17 cases where the NYCBC criterion was applicable. It is noteworthy that the interpreted capacity of the majority of these 17 cases was less than 150 kips (75 tons), which means that the second provision corresponded to a settlement less than 0.75 in., whereas the first provision corresponds to (PL/AE) + 0.75 in.

French Building Code (FBC)

The FBC defines the capacity as that corresponding to Equation 3, which is an offset limit similar to the AASHTO criterion but with a different intercept.

Δ = \frac{PL}{AE} + 0.4 in .

(3)

10% Diameter

The 10% diameter criterion is one of the oldest interpretation criteria dating back to the 1940s ( 9 ). It defines the capacity as the load corresponding to a displacement equal to 10% of the diameter. It is one of the few criteria that is recommended by multiple building codes across the world, including Germany and Japan, which could be attributed to its simplicity and ease of application and ensuring that the pile displacement corresponds to or exceeds failure. However, the methodology typically results in capacities corresponding to excessive displacements for large diameter foundations.

5% Diameter

The 5% diameter criterion was originally developed for drilled shafts and was later adopted for driven piles ( 10 ). Similar to the 10% diameter criterion, it defines the load as that corresponding to a displacement equal to 5% of the diameter. The 5% diameter criterion was suggested by GRL Engineers as the criterion of choice for drilled shafts in tension ( 11 ).

NYU Compression

The NYU compression criterion was first developed for compressive load tests on large diameter open-ended piles ( 12 ). After showing encouraging performance against 14 criteria, it was later applied to drilled shafts ( 13 ) and other pile types ( 2 ) where it cemented its potential as a universal interpretation criterion for compressive load tests on piles.

The NYU compression criterion accommodates several advantageous elements. First, an intercept-based criterion can be used objectively, making it a viable option as an autonomous interpretation criterion that requires minimal to no human involvement, and can be applied to a data set of load tests. Second, an ideal interpretation criterion should be able to determine the nominal capacity of a pile before the load causing plunging or strain-softening. Third, with the continuous increase in pile sizes, the diameter should not be explicitly included in any criterion. Therefore, the NYCBC criterion was adopted as the basis for the NYU compression criterion where the capacity is defined as that corresponding to the smallest of the following pile head displacements:

(PL/AE) + 0.75 in. (20 mm), where P is the load, L is the length of the pile, A is the area of the pile tip, and E is the pile’s elastic modulus;

displacement corresponding to the first incidence of plunging or strain-softening resulting in loss of more than 5% of capacity; and

5% of the pile diameter, unless modified by the structural engineer of record.

Analysis Methodology

The performance of the aforementioned interpretation criteria was evaluated using the available 80 load tests. The authors developed multiple Python scripts to (i) compile the load tests from the FHWA’s DFLTD; (ii) compile the collected load tests from the literature; (iii) compute the interpreted capacity for each criterion; and (iv) conduct the analysis on the output. For computing the interpreted capacity, the authors wrote a script that determines the loading portion of the load-settlement curve, then fits a curve to these points, finds the intersection between the load test and the interpretation criteria, stores these values, and generates a comprehensive plot of the load test, the available interpretation criteria, and pile information, including the diameter, length, elastic modulus, and pile type. The generated plots were manually inspected to identify any faulty load tests that require fine-tuning to the plotting script after each run, leading to the repetition of the analysis approximately 35 times. This was essential for ensuring the integrity of the data used in this study and maintaining the meaningfulness of the results.

Engineers typically define the pile capacity as the load causing failure during load testing. However, load tests do not lend themselves to immediate interpretation. For example, the load test shown in Figure 2 demonstrates that various available criteria yield different interpreted resistances. Add to this that many of the available load tests were most likely conducted on production piles where reaching failure is undesirable and expensive. This led to the absence of a “ground truth” capacity. As a solution, the authors decided to use the point of maximum curvature as a datum of comparison for the performance of all criteria. The maximum curvature point was chosen as it theoretically represents the inflection point after which the pile starts moving at a higher rate with a small load increment. Thus, the capacity at the maximum curvature point could be considered as the capacity corresponding to failure. Maximum curvature has also been the basis for several historic pile interpretation criteria such as DeBeer ( 14 ) and Vesic ( 15 ). For that purpose, the authors wrote a script that finds the point of maximum curvature for each load test, and stores and plots it. The script output was double-checked by the authors to ensure the integrity of the analysis.

Owing to the wide range of pile diameter and lengths, the interpreted capacity can vary significantly. To establish a common scale in the analysis, the capacity at the point of maximum curvature (Q_{max_c}) was used to normalize the individual interpreted capacities (Q_m/Q_{max_c}). It is, however, noteworthy that the point of maximum curvature (Q_{max_c}) cannot readily be used as a criterion, as its interpretation can be subjective. An objective assessment of Q_{max_c} was achieved in this study by using a Python package called kneebow, which calculates the point of maximum curvature of a curve, commonly referred to as the knee or the elbow of the curve, therefore, the name of the package. The code first finds the slope between the minimum and the maximum values in the data and then rotates the curve such that this slope between the minimum and maximum values is set to zero. The rotation is carried out by multiplying the data with a rotation matrix function of $θ$ , where $θ$ is the slope. Finally, kneebow determines the point of maximum curvature by simply taking the minimum, or the maximum, value of the data. Clearly, this process is not suited for routine interpretation of load test data.

Performance of the Available Criteria

The most important features that define the quality of an interpretation criterion are its unambiguous performance regardless of the test conditions and the percentage of cases it is applicable to. However, a few of the evaluated criteria were not applicable to many load tests included in this study. This could be attributed to some of the tests not being conducted to failure; however, because of the lack of piling information in the database, the authors can only surmise the reasons behind this observation. Therefore, the applicability of the evaluated criteria was investigated in Table 1. It was observed that the AASHTO, the original Davisson, and the FBC criteria were applicable to 98%, 87%, and 88% of the available cases, respectively. The NYCBC criterion and the 5% diameter criterion were applicable to approximately 60% of the cases, which is not encouraging. The 10% diameter criterion was applicable to only 21% of the cases, which is disappointing but consistent with the findings from previous studies ( 2 , 12 , 13 ). The NYU compression criterion showed a promising performance where it was applicable to 100% of the cases. This is expected because of the inclusion of the failure provision that ensures that the capacity is identified before or immediately after the onset of failure.

Table 1.

Applicability of Various Interpretation Criteria

Criterion	Overall (N = 80)	Concrete piles (N = 27)	Steel pipe piles (N = 38)	H-piles (N = 15)
Original Davisson	70	24	31	15
AASHTO	79	27	37	15
NYCBC	47	14	25	8
FBC	71	23	34	14
5% Diameter	49	15	20	14
10% Diameter	17	4	6	7
NYU compression	80	27	38	15

Note: AASHTO = American Association of State Highway and Transportation Officials; NYCBC = New York City Building Code; FBC = French Building Code; NYU = New York University.

Another important feature that characterizes the quality of an interpretation criterion is its accuracy and precision. Therefore, the accuracy and precision of the available methods were evaluated using the mean and standard deviation of the normalized interpreted capacities (Q_m/Q_{max_c}), where the optimal accuracy and precision were defined as 1.00 and 0, respectively. This assumes that (Q_{max_c}) is the “ground truth” and that an ideal interpretation criterion should define the capacity at the point of maximum curvature or shortly before reaching it. The mean and standard deviation of Q_m/Q_{max_c} of the available criteria are presented in Figure 3, as a bar chart and error bars, respectively, along with the number of cases each criterion was applicable to. It was observed that four methods overestimated the capacity by at least 20%, two methods overestimated the capacity by at least 10%, and only the AASHTO criterion underestimated the capacity by a mere 2%, which is desirable. It was also observed that all criteria exhibited a standard deviation as low as 0.26 and as high as 0.42. This shows that the AASHTO criterion has the highest accuracy (0.98) and moderate precision (0.31), whereas the standard Davisson criterion has the highest precision (0.24) and a moderate accuracy (1.10). This also suggests that using some of the available criteria is unsafe as they overestimate the capacity by a margin of 10% to 20%, which is undesirable.

In light of these observations, recommending the AASHTO criterion for interpreting tension load tests would be justifiable. However, the authors opted to explore the possibility of a new criterion that can surpass the AASHTO criterion.

Proposed Interpretation Criterion

With the exception of the 5% and 10% diameter criteria, all the other criteria use the elastic compression of a column (PL/AE) and only differ in the intercept. A notable observation is that as the intercept increases, the overestimation of the capacity by the interpretation criteria increases. This can be clearly observed when comparing the performance of the AASHTO criterion that has an intercept of 0.15 in. (∼0.004 m) and an average Q_m/Q_{max_c} of 0.98, and the NYCBC that has an intercept of 0.75 in. (∼0.02 m) and an average Q_m/Q_{max_c} of 1.22.

Trial Analyses with a Capacity Interpreted at a Fixed Pile Displacement

The authors explored the use of a fixed displacement as an interpretation criterion assuming that development of side friction is related to the magnitude of displacement. The approach introduces an error because a fixed constant displacement is only valid for an infinitely rigid pile in comparison to the soil. The values computed at various fixed displacements ranging from 0.25 to 0.55 in (0.00635–0.014 m) were selected for trial analyses as these values correspond to the quake normally required to develop skin friction in a variety of soils. The mean and standard deviation of the normalized Q_m/Q_{max_c} are plotted in Figure 4 against the fixed displacement used to obtain Q_m. The best fixed displacements were found to be in the range of 0.35 to 0.4 in. (0.009–0.010 m), suggesting that the offset value must be less than or equal to 0.4 in (0.01 m). At this fixed pile displacement, the mean and standard deviation of Q_m/Q_{max_c} were 0.99 and 0.24, respectively.

Trial Analyses with a Capacity Interpreted at a Percentage of the Elastic Elongation

Another noteworthy consideration is the usage of the full elastic compression (extension) of a column when testing piles in tension. The notion of using the elastic compression of a column is based on the suggestion that the full length of the column experiences the load when piles are tested in compression as the load transfers from the surface to the pile toe. This is rarely the case for piles in compression but grossly in error for piles in tension. Piles in tension cannot experience any end bearing, consequently assuming the full elongation of the pile is unreasonable. Therefore, the authors theorized that using half of the elastic elongation of a column plus an intercept between 0.15 and 0.75 in. (∼0.004–0.02 m) could result in a better performance in interpreting the capacity from a tension load test.

To explore this theory, twelve criteria that use half the elastic elongation of a column [1/2]·[PL/AE] and an intercept ranging between 0.15 and 0.75 in. (∼0.004–0.02 m) with 0.05 in. (0.0013 m) increments were investigated. The performance of these methods was evaluated using the same methodology where the normalized interpreted capacities (Q_m/Q_{max_c}) were used and are presented in Figure 5 next to the available criteria, along with the number of cases each criterion was applicable to.

Two trends immediately emerge. The first is that as the intercept increases, the criteria start to overestimate the interpreted capacity, which confirms the initial observation made by the authors. The second trend is that as the intercept increases, the number of cases a criterion is applicable to decreases, which is expected.

The authors also explored using other multiples of PL/AE on the basis that piles having reserve geotechnical capacity will transfer no load near the tip and most of the load transfer may occur near the top, thus one-third may better represent the distribution of load along the pile. These trials did not prove superior to a multiple of half. The best combination corresponded to a settlement of (1/3)·(PL/AE) + 0.3 in., where the mean and standard deviation of Q_m/Q_{max_c} were 0.99 and 0.24, respectively (Figure 6).

NYU Tension Criterion for Interpreting the Capacity of Tensile Load Tests

Three criteria can be used to achieve Q_m/Q_{max_c} of approximately 1 ± 0.02 and a standard deviation of 0.25 ± 0.01, including (i) a fixed pile displacement of 0.4 in. (0.01 m), (ii) capacity corresponding to a pile head displacement Δ=(PL/2AE)+0.25 in., and (ii) capacity corresponding to a pile head displacement Δ=(PL/3AE)+0.3 in.

The use of a fixed pile displacement for interpreting the uplift capacity represents a significant deviation from current piling practice. At the same time, a one-third multiple to the elastic elongation is difficult to substantiate in all cases as it requires more assumptions than simply taking the average load along the pile shaft. Thus, the authors chose the half multiple of the elastic elongation. It can also be observed that using half the elastic elongation plus 0.25 in. (∼0.006 m) exhibits similar performance to the AASHTO criterion with a mean Q_m/Q_{max_c} of 0.98, but with a lower standard deviation of 0.25, which is optimal.

Two other provisions from the NYU universal compression criterion ( 2 ) were retained. The first was the serviceability limit, which was set at 0.5 in. as long as it is not overridden by the structural engineer of record. Again, trial analyses were used to determine this value from a range of values between 0.4 and 1 in. (0.01–0.025 m). The second provision was that capacity must be interpreted before strain-softening resulting in a loss of more than 5% of capacity. Thus, the NYU tension criterion was evaluated alongside the available criteria, expressed as the capacity corresponding to the smallest of the following pile head displacements, as follows:

(PL/2AE)+0.25 in., where P is the load, L is the length of the pile, A is the area of the pile tip, and E is the pile’s elastic modulus;

pile head displacement corresponding to the first incidence of pull-out or strain-softening, resulting in loss of more than 5% of capacity; and

pile head displacement of 0.5 in., unless modified by the structural engineer of record.

A breakdown of the governing provision for the NYU tension criterion is presented in Figure 7. It is evident that most load tests are interpreted using the intercept provision. None of the tests were interpreted using the strain-softening provision, but it was deemed necessary to maintain the provision to avoid misinterpretation of the authors’ intent. Finally, a minority of tests were governed by the serviceability criterion of 0.5 in. It is noteworthy that all the tests interpreted by the 0.5-in. serviceability provision were carried out on open-ended pipe piles.

Effect of Pile Diameter and Length

The performance of an optimal capacity criterion should not be dependent on the length or diameter of the pile on which it is applied. The effect of the pile diameter and length on the performance of various interpretation criteria under consideration was investigated next by plotting the normalized interpreted capacities (Q_m/Q_{max_c}) versus the diameter and length for each criterion and categorized by pile type in Figures 8 and 9, respectively. Moreover, trendlines and their slopes were also plotted for each criterion. For the diameter effect, the slopes ranged between −0.0073D and 0.0020D, where D is the diameter of the pile, suggesting that Q_m/Q_{max_c} decreases by ∼7% or increases by ∼2% when the diameter increases by 20 in. (∼0.5 m). The NYU compression criterion showed nearly no diameter effect, whereas the 5% and 10% diameter criteria showed the largest slope, suggesting that they are highly affected by change in the diameter. This is expected as both rely solely on the pile diameter. It was also observed that the 10% diameter criterion consistently overestimated the capacity compared with Q_{max_c} and was only applicable to piles smaller than 40 in. (1 m) in diameter, which is expected as it corresponds to larger displacements for larger piles and that is undesirable. In general, the slopes were negligible with minimal diameter effect.

The proposed NYU tension criterion showed a moderate diameter effect equal to −0.0041D, suggesting that Q_m/Q_{max_c} decreases by ∼4% with a 20-in. (∼0.5 m) increase in the diameter. This is encouraging and suggests that the proposed criterion is not diameter dependent.

For the length effect, the trend line slopes were generally negligible ranging between −0.0011L and 0.0012L, where L is the pile length, suggesting that Q_m/Q_{max_c} differs by −11% to 12% when the pile length increases by a 100 ft (∼30 m). Once again, the NYU compression criterion showed the smallest slope, followed by the original Davisson criterion, and the 5% and 10% diameter criteria. On the other hand, the AASHTO criterion exhibited the highest positive slope, suggesting that it overestimates the capacity with the increase in pile length. Contrarily, the proposed NYU tension criterion exhibited one of the highest negative slopes (−0.0011L), suggesting that it underestimates the capacity as the length increases. This, although not desirable, is not alarming as the slope is generally small, and it is preferable to underestimate than to overestimate the capacity and is more conservative in the design stage. The slope effect can be reduced to −0.006L if the intercept term is changed from 0.25 in. to 0.75 in; however, this change corresponds to a Q_m/Q_{max_c} of 1.20 and a standard deviation of 0.25, which was deemed less desirable than that of the proposed criterion.

Limitations

This study inevitably has certain limitations like any study that uses statistical analysis of pile load test data. The first is the limited number of available tension load tests on piles, which resulted from conducting the analysis on a small data set for certain pile types such as H-piles. Although this number is not large by data analytics standards, it is considered adequate in the geotechnical engineering realm because of the limited availability of load tests in general and tension load tests in particular. A second limitation is the absence of information about the piling process. A third limitation is that the authors were not able to account for plugging of open-ended pipe piles nor pile setup because of the absence of this information in the available database. Plugging has been shown to affect the behavior of pipe piles ( 16 – 19 ), as well as their capacity ( 20 , 21 ), but it is not believed to affect the interpretation criterion used for a load test. A fourth limitation is that the proposed criterion, like nearly all other pile load test interpretation criteria, does not account for different stress and strain demands on the soil–foundation interface for drilled and driven foundations. A fifth limitation is that modulus degradation of piles in tension ( 22 ) has not been accounted for, in accordance with U.S. practice when applying pile interpretation criteria. These factors combine to introduce a degree of uncertainty in the analyses. However, the level of uncertainty is on a par with that of many pile design methods that have largely been developed with fewer than 100 load tests in compression. The situation for load tests in tension is less ideal, with far fewer available load tests. Finally, a variety of load testing approaches have gained in popularity in recent years including dynamic, Statnamic, and high-strain dynamic testing. The proposed criterion has only been checked for static load testing and is not indicated for any other load testing method.

Conclusions and Recommendations

The efficacy of seven available interpretation criteria for tension load tests was evaluated using a data set of 80 load tests conducted on various deep foundations including open and closed steel pipe piles, H-piles, drilled shafts, and square concrete piles. These cases were collected from the FHWA’s DFLTD v.2 and an exhaustive literature search. The piles ranged from 10 to 90 in. in diameter and from 18 to 240 ft in length. The evaluated criteria included ones that were not originally developed for tension load tests, such as the original Davisson, the NYCBC, the 5% and 10% diameter, as well as the NYU compression criteria, and included methods recommended for tension load tests such as the AASHTO and the FBC criteria. The capacity at the point of maximum curvature (Q_{max_c}) was used as the datum of comparison, as it represents the first point of failure and was used to normalize the individual interpreted capacities (Q_m/Q_{max_c}) to provide an intuitive scale of comparison. The analysis unveiled the following observations:

The AASHTO and NYU compression criteria were virtually applicable to the entire data set, whereas the original Davisson and FBC criteria were applicable to 90% of the cases, and the NYCBC and 5% diameter criteria were applicable to nearly 60% of the cases.

The 10% diameter criterion was applicable to 21% of the cases. It also corresponds to a mean and standard deviation Q_m/Q_{max_c} of 1.45 and 0.42, respectively, which is the worst among the evaluated criteria. The continued specification of the 10% diameter criterion by multiple building codes for interpreting the capacity of piles in tension is unwarranted and the authors highly recommend against it.

Apart from the AASHTO criterion, all the methods overestimated the capacity by at least 10%, which is unsafe for design purposes. The AASHTO criterion underestimated the capacity by 2% (Q_m/Q_{max_c} = 0.98), which is optimal.

The original Davisson criterion had the best standard deviation for Q_m/Q_{max_c} (0.26), whereas the AASHTO criterion exhibited a moderate precision having a standard deviation Q_m/Q_{max_c} = 0.31.

Using the full elastic compression of a pile as the foundation of an interpretation criterion for tensile load tests is implausible as piles in tension do not experience end bearing, therefore:

Twelve criteria that use half the elastic elongation of a column and an intercept ranging between 0.15 and 0.75 in. were evaluated. Using an intercept of 0.25 in. plus half the elastic elongation showed a parallel performance to the AASHTO criterion exhibiting a mean Q_m/Q_{max_c} of 0.98, yet a lower standard deviation of 0.25, which is encouraging. This criterion was termed the NYU tension criterion and is shown in Equation 4.

When investigating the effect of diameter on the Q_m/Q_{max_c}, it was found that the NYU compression criterion exhibited a minor diameter effect, which is encouraging, In comparison, the 10% diameter criterion was the most affected by any change in diameter, whereas the AASHTO criterion exhibited a negligible diameter effect, which is optimal.

For the effect of length on the Q_m/Q_{max_c}, it was observed that the AASHTO criterion had the highest length effect, whereas the NYU compression criterion had the lowest. The NYU tension criterion exhibited a high length effect, which remains generally small, and conservative compared with the other methods.

These observations support recommending one of the following criteria for interpreting the capacity of piles in tension:

The AASHTO tension criterion expressed by Equation 2 provided the most consistent performance among the available methods examined. However, the method uses an implausible assumption that the entire load is transferred to the pile tip in tension. Nevertheless, its continued use can easily be justified by its performance and past use in the AASHTO code.

The NYU tension criterion was developed specifically for interpreting tensile load tests on piles. It was applicable to all the inspected load tests, achieved ideal accuracy and precision, and was evaluated on various pile types. The NYU tension criterion defines the capacity as the load corresponding to a pile head displacement corresponding to the smallest of the following terms:

○ (PL/2AE) + 0.25 in., where P is the load, L is the length of the pile, A is the area of the pile tip, and E is the pile’s uncracked elastic modulus;

○ displacement corresponding to the first incidence of pull-out or strain-softening resulting in loss of more than 5% of tensile capacity; and

○ a serviceability criterion corresponding to 0.5 in., unless modified by the structural engineer of record.

The continued use of several popular capacity interpretation criteria, initially intended for piles in compression, cannot be justified in light of the available data. The criteria that should be abandoned for piles in tension include (i) original Davisson , (ii) NYCBC, (iii) FBC, (iv) NYU compression, (v) 10% diameter, and (iv) 5% diameter criteria.

Nomenclature

The following terms were used throughout the paper

A Pile Area

AASHTO American Association of State Highway and Transportation Officials

DFLTD Deep Foundation Load Test Database

D Pile Diameter

E Elastic Modulus

FBC French Building Code

FHWA Federal Highway Administration

L Pile Length

NYCBC New York City Building Code

NYU New York University

P Load

Q_m Interpreted Capacity

Q_max Maximum Load

Q_{max_c} “Ground truth” defined as the capacity at the point of maximum curvature or shortly before reaching it

Δ Settlement

Footnotes

Acknowledgements

The two recommended criteria represent an extension of the pioneering work of Professor Davisson; all this work builds on his legacy. The authors also gratefully acknowledge the contribution of prior research group members in paving the way for this study. In particular, Nick Machairas, PhD, now data analytics leader at Haley and Aldrich, Inc., ported the FHWA database to a relational database that was used in this work.

Author Contributions

The authors confirm their contribution to the paper as follows: study conception and design: A. Kodsy, M. G. Iskander, B. Öztürk, Y. Bazi; data collection: Y. Bazi; analysis and interpretation of results: A. Kodsy, B. Öztürk; draft manuscript preparation: A. Kodsy, M. G. Iskander. All authors reviewed the results and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Antonio Kodsy

Baturalp Öztürk

Youssef Bazi

Magued G. Iskander

References

Tomlinson

Woodward

Pile Design and Construction Practice. CRC Press, London, 2007. https://doi.org/10.4324/9780203964293.

Kodsy

Iskander

Sampurnanand Pandey

Universal Criterion for Interpreting Capacity from Load Tests on Piles. Transportation Research Record: Journal of the Transportation Research Board, 2022. 2676: 530–541.

Petek

Mitchell

Ellis

FHWA Deep Foundation Load Test Database Version 2.0 User Manual. Report No. FHWA-HRT-17-034. U.S. Department of Transportation Federal Highway Administration, McLean, VA, 2016. https://rosap.ntl.bts.gov/view/dot/35880.

Machairas

Highley

G. A.

Iskander

M. G.

Evaluation of FHWA Pile Design Method Against the FHWA Deep Foundation Load Test Database Version 2.0. Transportation Research Record: Journal of the Transportation Research Board, 2018. 2672: 268–277.

Davisson

M. T.

High Capacity Piles. Proceedings of Lecture Series on Innovations in Foundation Construction, Vol. 52, 1972, pp. 81–112.

NeSmith

W. M.

Siegel

T. C.

Shortcomings of the Davisson Offset Limit Applied to Axial Compressive Load Tests on Cast-in-Place Piles. In Contemporary Topics in Deep Foundations ( Iskander

Laefer

D. F.

Hussein

M. H.

eds.), American Society of Civil Engineers, Reston, VA, 2009, pp. 568–574. https://doi.org/10.1061/41021(335)71.

American Association of State Highway and Transportation Officials (AASHTO). AASHTO LRFD Bridge Design Specifications, Customary U.S. Units, 7th Edition, with 2015 Interim Revisions. AASHTO-LRFDUS-7-M, Washington, D.C., 2014.

Hannigan

P. J.

Goble

G. G.

Likins

G. E.

Rausche

Design and Construction of Driven Pile Foundations Volume II. Report No. FHWA-NHI-05-043. National Highway Institute (U.S.), Washington, D.C., 2006.

Terzaghi

Discussion on Progress Report of the Committee on the Bearing Value of Pile Foundation. Pile-Driving Formulas. American Society of Civil Engineers Proceedings, Vol. 68, 1942, pp. 311–323.

10.

Reese

L. C.

Isenhower

W. M.

Wang

S.-T.

Analysis and Design of Shallow and Deep Foundations. John Wiley & Sons, Inc., Hoboken, NJ, 2005.

11.

GRL Engineers. Load-Movement Measurements and Failure Load Determination. Online Report, 2016. http://web.archive.org/web/20220728143331/http://www.grlengineers.com/wp-content/uploads/2017/04/load-movement-measurements-and-failure-load-interpretation.pdf.

12.

Kodsy

Machairas

Iskander

Assessment of Several Interpreted Pile Capacity Criteria for Large-Diameter Open-Ended Piles. Geotechnical Testing Journal, Vol. 44, No. 5, 2021, Article 20200074. https://doi.org/10.1520/GTJ20200074.

13.

Kodsy

Iskander

Machairas

Assessment of Several Nominal Resistance Interpretation Criteria for Drilled Foundations. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 148, No. 2, 2022, Article 04021181. https://doi.org/10.1061/(ASCE)GT.19435606.0002733.

14.

DeBeer

E. E.

Proefondervindelijke bijdrage tot de studie van het grensdraagvermogen van zand onder funderingen opstaal (in Dutch). Tijdschrift der Openbare Werken van Belgie, Vol. 6, 1967 and Vol. 4, 5, and 6, 1968.

15.

Vesic

A. B.

Bearing Capacity of Deep Foundations in Sand. Highway Research Record 39. Highway Research Record, Washington, D.C., 1963, pp. 112–153.

16.

Paikowsky

S. G.

Whitman

R. V.

Baligh

M. M.

A New Look at the Phenomenon of Offshore Pile Plugging. Marine Geotechnology, Vol. 8, 1989, pp. 213–230.

17.

Paik

Salgado

Determination of Bearing Capacity of Open-Ended Piles in Sand. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 129, No. 1, 2003, pp. 46–57. https://doi.org/10.1061/(ASCE)1090-0241(2003)129:1(46).

18.

Kodsy

Iskander

Insights Into Plugging of Pipe Piles Based on Pile Dimensions. Applied Sciences, Vol. 12, No. 5, 2022, p. 2711. https://doi.org/10.3390/app12052711.

19.

Iskander

Behavior of Pipe Piles in Sand: Plugging and Pore-Water Pressure Generation during Installation and Loading. Springer Series in Geomechanics and Geoengineering, Berlin/Heidelberg, Germany, 2011.

20.

Rizk

Kodsy

Iskander

Machairas

Efficacy of Design Methods for Predicting the Capacity of Large-Diameter Open-Ended Piles. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 148, No. 10, 2022, p. 04022078. https://ascelibrary.org/doi/abs/10.1061/%28ASCE%29GT.1943-5606.0002824.

21.

Iskander

On the Design of Instrumented Double-Wall Model Piles Used to Investigate Plugging of Open-Ended Pipe Piles. Geotechnical Testing Journal, Vol. 34, No. 2, 2011, pp. 147–154. https://doi.org/10.1520/GTJ103096.

22.

England

On the Subject of Piles in Tension. In Full-Scale Testing and Foundation Design: Honoring Bengt H. Fellenius ( Hussein

M. H.

Massarsch

K. R.

Likins

G. E.

Holtz

R. D.

eds.), American Society of Civil Engineers, Reston, VA, 2012, pp. 680–693.