Serge Borel, Michel Bustamante, Luigi Gianeselli, Laboratoire Central des Ponts et Chaussees Introduction Tests on piles are generally performed to check that their structural integrity is satisfactory and that their real capacity is in reasonable accordance with the forecast load-settlement curve of the anticipated working load. For the engineer the load test is considered satisfactory when it has been possible to determine:
lthe ultimate resistance Q ulthe pile settlement under working load Q wlthe shaft Q sand toe Q presistances under Q uIn current practice unfortunately, it is rare that all of these loads, Q u, Qw, Q s, Q pare available because the load test has not been taken to failure and because of lack of instrumentation. To define the ultimate resistance Q uand possibly the shaft resistance Q s, authors such as Van der Veen (1953), Brinch Hansen (1963), Chin (1970), Davisson (1972), Mazurkiewicz (1972), Fuller & Hoy (1977) and others, have proposed methods based on mathematical or graphical approaches. A comprehensive comparative study of the nine methods most commonly used in practice has been carried out by Fellenius (1980).
Because ultimate resistance Q uremains the key value in most piling problems, it is worthwhile to recall that various definitions of this concept are given by different authors and codes. For example:
Qu, F plunging failure, observed sometimes insitu and corresponding to a rapid increase of settlement without an increase in load (Figure 1) Qu, asym asymptotic ultimate load derived mathematically or graphically (Figure 2) from the measured load settlement curve Qu, conv conventional ultimate load defined by many codes as the load causing a gross settlement equal to 10% of the equivalent pile diameter D (Figure 2), or the load at which the penetration creep rate reaches a given value.
The failure derived by the Chin method falls into the mathematicalasymptotic category and has been defined further as Q u, CHIN . This paper examines the performance of Chin's recommendations on load tests with limited settlement and the ability of his method in predicting the proportion of shaft load.
The Chin method The Chin method is recognised as one of the simplest methods and has been adopted by engineers in many countries. Primarily developed for footings and precast floating piles, Chin observed that in many cases the plot of the settlement, s, versus the ratio s/Q gives a linear relationship when the pile approaches failure. The response of the pile-soil system can then be considered in terms of hyperbolic shape. It has been expressed by Chin (1970, 1972) in the form :
s / Q = C + m s  Using different notation, equation  becomes:
s (ds )sQ = dQ Q=0 +Qasympt where for both equations :
Qthe applied load on the pile head sthe pile head gross settlement mthe slope of the linear plot corresponding to the inverse of the asymptotic value of the ultimate resistance Q asym .Ca constant corresponding to the initial slope (ds )of the load settlement curve plotted in linear coordinates dQ Q=0 .In a system of hyperbolic coordinates s versus s/Q, a given typical relationship s-Q obtained from a routine pile test (Figure 3a) system becomes generally:
lin most cases and according to the evidence presented by Chin (1970) a single straight line as shown on Figure 3b.
lfor other cases, two straight lines plot with an A and B part as illustrated on Figure 3c.
Chin's final proposal (1978) allowing the detection of broken piles from a particular plot will not be addressed here.
The graph in Figure 3b is supposed to model the behaviour of a pile supporting the load mainly through shaft friction or purely in end bearing. Because the linearity is not always explicit for the first points of the relationship, Chin (1972) recommended that these are rejected when determining C and m.
The bilinear plot with its two straight lines which intersect (Figure 3c) is supposed to correspond to piles supported by combined shaft friction and end bearing. Because toe resistance needs higher settlement to be mobilised than shaft resistance, it has been suggested that the first part (A) would represent shaft resistance while the second part (B) would represent total load. The validity of these assumptions will be shown later.
The advantages presented a priori by the method are undeniable for common practice:
lthe method is not linked with any soil parameters;
lit gives the possibility to analyse the results of either CRP or ML tests and for piles subject to compression or tension loading;
lit allows a continuous check while the test is being carried out;
lthe interpretation can be carried out quite simply using at most a pocket calculator;
lthe definition of characteristic loads Q u, Q s, Q pis based on mathematical rules;
lbecause of the linearity of the s vs s/Q plot, it is easy to determine any load associated to any given gross settlement. At a settlement equal to 10% of the pile diameter one can obtain Q u, CHIN 10% which is an estimate of Q u, conv according to the 10% criteria.
However, Fellenius (1980), Fleming (1992) and England (1999) have shown that considerable drawbacks can complicate the interpretation and distort the performance of the method. They are:
lthe necessity to collect data at sufficiently large gross movements, s, to avoid deducing false ultimate resistance;
la marked sensitivity to the chosen test procedure demanding constant time increments as well as a sufficient number of load steps;
lthe validity of the method is jeopardised by the fact that the real response of the pile/soil system is not exactly a hyperbolic form. In particular, elastic shortening of the pile is a major cause of erroneous analysis, especially in the first part (A) of the curve, where elastic shortening can be a large proportion of the pile head movement.
Before commencing a more comprehensive study, a strict application of the Chin method has been made on one pile tested by the authors at a site in Mittersheim (France) in 2001.
The pile was a 480mm diameter CFA pile, 7m long and bored through marly clay (26% of CaCO 3). At the pile toe level, the soil is characterised by a CPT cone resistance q c= 5MPa and a PMT limit pressure p l= 1.8MPa. The water table is located at 4m depth.
The pile was instrumented with two strings of removable extensometers, which made it possible to measure the mobilisation of shaft and toe resistance during the loading stages. It was loaded according to a maintained load test (MLT) procedure, with 13 steps of 100kN, each maintained for 60 minutes. The last load Q max =1.3MN was applied during 15 minutes only because of the rapid increase of settlement which was observed, indicating plunging failure under that load.
Figure 4 shows the typical load settlement curve, plotted using the s values measured at the end of each load step. The load settlement curve is also plotted in the hyperbolic coordinates s vs s/Q.
When enlarged, the relationship s vs s/Q indicates clearly A and B parts as shown on Figure 5 which focuses on the first loading steps.
A linear regression made using the seven last points before failure (600kN to 1,200kN) has resulted in:
C = (ds )= 2.36 10 -3 mm/kN and Q u, CHIN = Q asym = 1 = 1.315 MN dQ Q =0 mApplying the hyperbolic Chin formula to a gross settlement limited to 48mm (10% of the equivalent diameter) gives Q u, CHIN 10% = 1.235MN.
The instrumentation of the pile (Figures 6a-c) makes it possible to compare the shaft resistance determined according to Chin's criterion on the first part of the curve Q s, CHIN 10% = 715kN with the measured value Q s, u 10% = 860kN. For the considered case history the agreement can be considered as quite good in geotechnical terms, as the difference between the calculated and measured values is only 17%.
Following these first encouraging results the performance of the method has been investigated for a greater number of pile tests carried out by the authors. The following items were investigated:
lsensitivity of the method as a function of the available field readings when determining the ultimate capacity;
lperformance of the method when estimating shaft and toe resistance.
The reviewed case histories A total of 50 loading tests from around Europe, carried out by the authors since 1990, have been examined. All tests were run under a MLT procedure up to a head settlement greater than 10% of the pile diameter and without intermediary unloading and reloading cycle. Each load step was maintained for a period of generally 60 minutes.
All of the readings were digital. The load was always checked with a load cell and the settlements were measured with the help of linear potentiometers linked to fixed reference beams. The piles were all instrumented using removable extensometers (Bustamante et Doix, 1991), which made it possible to measure the shaft and the toe resistances.
The 50 tests have included various pile techniques and covered an extensive range of soils. Four piles were loaded in tension. The loads applied on piles ranged from 675kN to 10,000kN, the pile lengths from 5m to 45m, and the pile diameters from 200mm to 1,000mm. Note that due to substantial elastic shortening, micropiles with free (ungrouted) length have not been considered in this study.
The following piles have been examined: CFA (12 tests), bored (six tests), impact driven steel and concrete (17 tests), vibratory driven (five tests) and screwed displacement auger (10 tests). The piles were installed in clay (20 tests), sand (15 tests), calcareous soils (eight tests) and mixed soils (seven cases).
Ultimate resistance prediction To assess the performance of the Chin method in determining the ultimate resistance, two different approaches have been followed:
ltaking into account the complete load settlement curve;
lconsidering a partial load settlement curve obtained by removing gradually the last points, as if the test was stopped at a lower loading step.
Determining the ultimate resistance from the complete curves Each test has been plotted in hyperbolic coordinates, to identify the second part of the load settlement curve (part B), from which the Chin parameters have been calculated. The Chin ultimate resistance according to the 10% criteria Q u, CHIN 10% has been compared to measured ultimate resistance. As shown on Figure 7, the Chin method gives a very satisfactory estimation of the measured ultimate resistance, except in five cases, where the Chin resistance exceeds the measured one by more than 25%. These are cases where a plunging failure has been observed, which is obviously not anticipated by the Chin method.
Determining the ultimate resistance from partial load-settlement curves To assess the sensitivity of the Chin method in function of the magnitude of available head settlement and the number of loading steps, the load settlement curves have been progressively altered by removing, one after another, the last points.
Figure 8 shows the example of the pile tested at the Mittersheim site. The last two points have been removed to obtain a shortened load settlement curve, consisting of the first 11 loading steps, of the 13 actually taken. The maximum settlement on the curve s = 16mm is only 3.3% of the pile diameter.
The Chin method applied to the first 11 loading steps make it possible to extrapolate the ultimate resistance according to the 10% criteria: Q u, CHIN 10% = 1,153kN, which is 92% of the measured ultimate resistance.
This procedure has been repeated, suppressing one after another of the points, until the load settlement curve is made of the first 2 loading steps only, which is the minimum necessary to apply the Chin method. The results are shown on Figure 9. Note that to reduce the effects of human errors and contrary to Chin's recommendation, the first points of the load settlement curve were not rejected in a first stage.
This procedure has been repeated for all of the 50 loading tests (Figure 10). The main conclusions to be drawn from this figure are the following:
lthe reliability of the Chin prediction improves when the last measured gross settlement approaches 10% of the pile diameter;
lthe majority of the data (in black circles) lie inside a clearly defined envelope. The only cases falling outside, a total of seven out of 50, correspond to:
- plunging failure (five cases in red squares), a case not anticipated by the Chin method;
- tension test on driven sheet piles in dense coarse gravels (two cases in blue triangles) with an atypically shaped curve and only partially mobilised shaft friction (Borel et al, 2002);
laccording to a detailed study neither the installation technique nor the soil type appear to have a noticeable effect on the results;
las soon as the head settlement exceeds 5% of the pile diameter, the ultimate resistance predicted by the Chin is in the range of 75% to 110% of the measured resistance.
The last conclusion is of practical interest for piles, which have not been loaded to failure on site.
A different approach was made in a second stage in order to avoid the influence of the first part of the load settlement curve, which is representative of the shaft friction only (part A). The two first points and the settlements lower than 0.5% of the pile diameter were removed from the curves.
The new envelope obtained is shown on Figure 11. Its upper and lower limits shown in green are expressed by equation:
predicted ultimate resistance head settlement s ]measured ultimate resistance =1[0.35+0.4x log (pile diameter D )  This procedure enabled the ultimate resistance to be predicted to an accuracy of -17% for pile head settlement exceeding 5% of the pile diameter, and -25% for gross settlement greater than 3% of the pile diameter.
The new envelope is not as good as the envelope obtained by Kaniraj (1988), who applied the Chin method to 24 pile tests and concluded that the error between predicted and actual load is only 10% for a relative pile head settlement s/D > 2.5%.
Determining shaft and toe resistances According to Chin, a pile acting in combined friction and end-bearing can be represented by a plot with two straight lines A and B. However, such a plot does not always appear clearly in current practice because of distorting factors. Out of the reviewed 50 piles the following typical plots were identified:
l36 cases in good agreement with Chin's hypothesis showing A and B parts (Figure 5).
lsix cases with a S shape curve (Figure 12a);
lseven cases displaying an inverted A part (Figure 12b); these involved piles installed in stiff clay with a measured toe resistance that was less than 25% of the ultimate resistance.
Thus, it was possible to determine the shaft resistance in 43 cases, i. e. 86% of the total. The histogram of predicted versus measured shaft resistance using the 10% criteria is shown on Figure 13. Due to the cases for which the predicted value greatly exceeds the measured ones, the average is 1.07 and the standard deviation is 0.55, even though most of the predictions are in the range of 60% to 120%.
Figure 14 shows the ratio of the predicted/measured shaft friction versus the percentage of the load carried by shaft to allow better identification of the cases for which the Chin method gives poor results.
The main conclusions are as follows:
lthe predicted shaft friction is in the range of 60% to 120% of the measured value when the shaft resistance is greater than 55%;
lon the contrary, the Chin method significantly overpredicts the shaft resistance where the load is mainly resisted by the toe.
For the three cases giving the worst estimation, a geotechnical engineer would probably reject the prediction according to the Chin method because of the inconsistency between the soil investigation and the predicted shaft and toe resistance. These piles are :
-A screw pile (L = 7.6 m D = 510/710 mm) installed through a 5m thick layer of peat to refusal against a layer of dense gravelly clay (CTRL 310, UK);
-A steel HP 360x370 pile (L= 9.80 m) driven through clayey sand and schistic marls (Esch, Luxembourg), - An open tube (L = 21 m) driven through chalk below the water table (Scardon, France).
Another approach, summarised on Figure 15, shows that the absolute error made in the predicted shaft resistance can exceed 30% of the ultimate resistance, quite independently of the percentage of load resisted by the shaft.
Once the shaft and total capacities are estimated using the Chin method, it is possible to deduce the toe resistance by subtracting the shaft resistance from the ultimate resistance. The error on the toe resistance is the same as the error on shaft resistance plotted on Figure 15.
Figure 16 shows that the increase of the number of loading steps does not improve significantly the accuracy of the shaft resistance prediction.
Conclusions Chin's method has been applied to 50 full-scale pile tests in order to assess its performance. The selected case histories are representative of the most common installation techniques used and soil categories encountered. All of the piles were instrumented using removable extensometers, which make it possible to measure the distribution of load between the shaft and the toe.
The Chin method makes it possible to predict the ultimate resistance, according to the 10% criteria, even if the head settlement did not reach 10% of the pile diameter. The estimation improves when the head settlement approaches 10% of pile diameter. The ultimate resistances were predicted in the range of:
l-25% as soon as the head settlement exceeds 3% of the pile diameter;
l-17% for a head settlement greater than 5% of the pile diameter.
The method appears to be of generally poor performance in predicting the distribution of load between the shaft and toe:
lthe Chin method overpredicts significantly the shaft resistance when the load is mainly resisted by the toe;
lthe shaft resistance is estimated in the range of 60% to 120% of the measured value when the shaft resistance is greater than 55%;
l the absolute error made on predicted shaft and toe resistances can exceed 30% of the ultimate resistance, quite independently of the percentage of load carried by the shaft.
Direct measurement of the shaft and toe resistances are still of tremendous value even for routine tests as they provide a much better understanding of pile behaviour.
References Borel S, Bustamante M, Gianeselli L (2002) Comparative field studies of the bearing capacity of vibratory and impact driven sheet piles, Proceedings International Conference on Vibratory Pile Driving and Deep Soil Compaction TransVib 2002, Louvainla-Neuve, 167-174.
Bustamante M, Doix B (1991) A new model of LPC removable extensometer, Proc. 4th International Deep Foundation Institute Conference, pp 475-480.
Bustamante M, Gianeselli L (1980-1999) LCPC internal accounts and authors publications covering 30 reviewed case histories.
Chin FK (1970) Estimation of the ultimate load of piles from tests not carried to failure, Proceedings of the Second Southeast Asian Conference on Soil Engineering, pp83-91.
Chin FK (1972) The inverse slope as a prediction of ultimate bearing capacity of piles, Proceedings 3rd Southeast Asian Conference on Soil Engineering, pp83-91.
Chin FK (1978) Diagnosis of pile condition, Geotechnical Engineering, vol 9, pp85-104.
Chin FK, Vail AJ (1973) Behaviour of piles in Alluvium, Proceedings, 8th International Conference on Soil Mechanics and Foundation Engineering, vol 2.1, pp47-52.
England M (1999) A pile behaviour model, Thesis submitted to Imperial College, 305 pages.
Fellenius BH (1980) The analysis of results from routine pile load tests, Ground Engineering, vol13, no 6, pp 19-31.
Fleming WGK (1992) A new method for single pile settlement prediction and analysis, Geotechnique no 42, vol3, pp411-425.
Kaniraj Shenbaga R (1998) Interpretation of pile acceptance criteria from deficient data, ASCE Journal of Geotechnical and Geoenvironmental Engineering, vol. 124, no10, pp 1035-1040.