Introduction The meeting, chaired by BGA honorary secretary Neil Smith, was presented by Dr Ken Fleming and Dr Melvin England of Cementation.
Non-linearity in soil behaviour is well known and much referenced in geotechnical literature. Specific non-linear behaviour depends on the mechanism of the event being observed. For example in the case of a pile there are usually two mechanisms - one related to shaft friction mobilisation, involving shear local to the pile shaft only, and one relating to end bearing which is characterised by bulk displacement of soil.
All of the related load-deformation characteristics are influenced by the time any specific load or stress is in place - behaviour commonly referred to as creep. The behaviour is, in effect, three-dimensional, with the axes being stress, strain and time. To disentangle the effect of each, one of the variables must be locked or removed from each loading event. Load control has become an important issue in pile testing. When constant loads are applied, the variation of strain with time becomes ordered and can be modelled, if the underlying behavioural mechanisms are taken into account.
England referred to the recent Rankine Lecture by Professor John Atkinson in which he concluded 'the rate at which a test is performed can lead to different (stress-strain) results' In pile tests this becomes obvious. Determining final settlement under a constant load is imperative to interpret tests in a consistent and meaningful way. At the same time the mathematics used to determine final settlement must take into account the two mechanisms that characterise pile behaviour - shaft friction and end bearing.
The effects of holding loads constant is easily demonstrated by comparing a test which has been carried out under manual control and one which has been carried out under computer control.
Figure 1 is an example of the displacement-time recorded using manual load control. Elastic shortening of a pile is an important feature, accounting for much of the irregularity encountered in manual tests where it is difficult to maintain the load within say 1%. Automatic load control allows loads to be controlled within 0.1% (a typical record is shown in Figure 2).
Figure 3 shows a computerised testing arrangement.
The smooth displacement-time function which occurs with a controlled load depends on the mechanisms present. In most tests there is one smooth function for shaft friction and another smooth function for end bearing. The observed time displacement is an combination of these two functions. When they are mathematically separated from the overall result (Figure 2), it becomes apparent that they are similar but have different magnitude. As a result, depending on the time elapsed since load application, the distribution of load between the shaft and base may be inferred to vary within the single load application.
England concluded that to analyse foundation behaviour properly, the importance of time functions must be recognised, they must be accurately modelled and the final deformation under any given load must be determinable. Only at the final asymptotic value of time-related deformation under a constant load can it be said that the distribution of load along a pile has reached a constant state.
Timeset is a model representing displacement related to time developed for interpreting pile test results. Linear fractional equations are employed to model both end bearing and shaft friction and combined can then be used to model complete pile behaviour. Once the component functions have been identified and the observed result has been matched, the final settlement for a given load is readily determined (Figure 4).
England observed that Timeset analyses indicate that the mechanism and behaviour of sands and clays are essentially similar, even if the parameters are a little bit different.
There is no discontinuity in the transition from short term to long term behaviour and no division between instantaneous settlement and long term settlement. The presence or absence of water has little apparent effect and it would appear that pore-water dissipation is a function of creep and is really a secondary effect.
Experiments show that peak reaction loads develop during pile tests according to varied rates of penetration. As penetration rates rise, peak reaction forces also increase. This means there must be long term pile behaviour with an additional, rate dependent, reaction load. It is apparent in tests with continuous rates of penetration that an enhanced peak reaction force is reached, followed by a decline as deformation takes place. This phenomenon can be seen in a test carried out at the Bothkennar site (Lyndon et al (1994), Piling and deep foundations, Bruges) shown in Figure 5, which clearly demonstrates that the higher the rate of penetration, the higher the peak.
The effect of sand density on peak resistance variation due to rate is further demonstrated by Figure 6. Lower density sands produce a smaller gain in peak strength than higher density ones.
It is apparent that the peak effect is strongly related to dilation. Dense sands increase in volume during shearing in a shear box, while there is no real volume increase in loose sands. The same mechanism is thought to govern the development of skin friction of piles (Figure 7).
Dilation occurs in soil immediately surrounding the pile shaft in reasonably dense soils. This would naturally produce radial volume change but because of confinement, the lateral force on the surface of the pile is increased. This gives rise to the peak reaction force. As strain becomes large the force diminishes and pile capacity shows post-peak decline. As a consequence, constant rate penetration (CRP) tests are inappropriate for determining pile capacity. A typical example of the analysed long term behaviour versus over-prediction of capacity that may be interpreted from a CRP test is shown in Figure 8.
Interpretation of strength and stiffness
One of the main difficulties in interpreting the fully timecorrected load/settlement relationship is that non-linear behaviour depends on both strength and stiffness. The practical way to deal with this is to define a single value of stiffness related to a fixed fraction of ultimate load. This is often taken as 25% of the ultimate (asymptotic) load. It is possible to deal with stiffness in this way because the non-linear soil functions match closely to linear-fractional functions and these can be defined simply by their start or origin, their final or asymptotic value and one single point (or indeed a slope) in-between. The 25% point is convenient from a practical point of view for pile bases.
A significant problem in geotechnics is that the means of determining stiffness is not part of conventional site investigation. Stiffness and strength have to be inferred from conventional soil tests and this depends on establishing correlations where possible. Even where tests are specifically chosen to look for reliable parameter values, they are usually carried out at test rates which seriously distort the final answer.
Site investigation tools are generally poor for the purpose of deriving stiffness and strength parameters. Inevitably however, they have to be used for foundation design until better methods can be developed. Soil tests that can be are used include:
Undrained shear strength: does not measure strength or stiffness in any absolute way. Approximately equivalent to an index test.
Standard penetration test: a dynamic test with dependence on bore size and which can easily be upset by water inflow, large rock fragments, and even non-standard equipment. Results can be very misleading.
Pressuremeter: should be effective but measures in the horizontal direction. Horizontal and vertical soil properties may differ significantly.
Dutch cone: a dynamic and rate affected test. Its depth may be limited in hard ground. It provides much information about local ground variations and can be a reasonably good strength indicator tool. Stiffness cannot normally be determined.
Seismic methods: uses bulk property response at low strain to infer low strain soil modulus. Directional uncertainty is probable and careful correlation it required. It is a 'young' application of the method in need of further research.
Figure 9 shows a comparison of the installation record of a driven cast-in-place pile with the record of SPT with depth on the same site. The pile driving records showed a very soft layer at 10m below ground (Figure 10 indicates the set change according to depth), whereas the SPT data showed only very scattered results, showing very poor correlation with the driving record and little indication of the soft layer. Piles were driven into the dense sand below the soft layer as indicated by the pile driving record and pile test results were excellent.
Stiffness is obviously influenced by the stress history of soils. The low strain modulus depends primarily on this and is not influenced much by strength. As stress is increased, the influence of strength grows and that of stiffness diminishes until at the ultimate state stiffness is zero. It is very important to have a framework in which this transition can be mathematically modelled.
Pile installation affects both stiffness and strength. A problem is that site investigations are carried out before piling, whereas the real soil properties are not settled until after piling. This difficulty is compounded by other site investigation problems (above) and means that in most cases stiffness and strength are best back-calculated from pile tests.
Fleming said that base stiffness deduced from pile tests was one quarter of the ultimate load. To obtain reasonably accurate values, it is necessary first to be able to define the asymptotic ultimate load with reasonable certainty. This means that the further a test pile is pushed into the ground, the better it is for the analysis. However, the luxury of substantial movement is not typically available and the best has to be done with tests taken to settlements of 20mm or more.
Using the asymptotic definition of ultimate load, it becomes evident that the ultimate capacity of bored and driven piles is not significantly different. However, the stiffness can be quite different and its maximum potential value can be seriously influenced by construction. The base of a bored pile may have residual debris with detriment to stiffness, for example, which means installation technique is very important.
Some construction techniques cause greater cleaning difficulties than others. It is therefore very important to recognise the mechanism by which load is carried. If end bearing is the major issue, then bases must be clean. If shaft friction provides the majority of capacity and is well in excess of the applied working load, then movements will be small and base cleaning is perhaps not so important. Fleming warned of short fat piles with lots of end bearing because they represent the highest risks.
Continuous flight auger piling performs well in regard to base stiffness provided the technique of construction is good, which can only be verified by high quality monitoring. Similarly, driven cast-in-place piles perform well when heave conditions are avoided or are considered.
Precast piles are similar in many respects and perform well if not damaged by over-driving. It is often forgotten that driving through softer soils and heave effects can cause quite severe tensile forces in piles and they need to be fully reinforced accordingly. Nevertheless, pile concrete stiffness is often found to be dramatically reduced due to multiple small cracks.
Fleming then considered various soil conditions and presented tables and graphs relating stiffnesses to soil descriptions and to deduced soil ultimate bearing pressures from pile test analyses using the Cemsolve/Cemset method (Fleming, Geotechnique, September 1992). Many of the results were obtained from tests on CFA and driven cast-in-place piles. There were quite wide scatters in some of the results but for guidance purposes the values chosen seemed appropriate.
Clays (bored piles) E bin terms of undrained shear strength; note that this is the value of the secant modulus at 25% of the ultimate load.
There are difficulties in base cleaning tripod bored piles in many types of soil.
Typical range related to weathering zones (CIRIA report 47) Sands and gravels E bin terms of relative densities Chalk Much depends on the structure of the chalk, its discontinuity apertures and the matrix material within the block structure. The values in the table are reasonably typical. Grades are described as in the CIRIA Funders Report.
For piles constructed without enlarged bases Fleming suggested using values as for well-cleaned bases (eg as for CFA piles) above or if the pile has a driven enlarged base, to multiply by approximately 1.5.It is necessary to consider that when piles are being driven in stiff clays, heave occurs, which can lift piles off the founding stratum unless they are sufficiently anchored into it.
Piles founded in 'blocky' rocks can exhibit unusual behaviour. A column of blocks can be displaced downwards beneath the pile toe, squeezing out the soft interstitial material.Apparent shaft friction may increase and ultimate end bearing decrease accordingly.
Attempts to relate stiffness to shear strength are common. In the case of over-consolidated clays, for example, soil stiffness modulus is commonly quoted as 150 to 400 times the undrained shear strength.
When stiffness modulus values are plotted against ultimate base pressure some approximate ratios are apparent, (where quis the ultimate bearing pressure).Figure 11 shows the plot for sands.
Fleming pointed out that after analysing some 2,000 pile tests, it is apparent that in the Cemset formulation the shaft flexibility parameter M sis a constant. It is of course not possible to found many piles in very soft nor in very hard soils so that there is no proof of its constancy for these cases. However it is a dimensionless parameter and its behaviour may not therefore be surprising. Generally M sis approximately 0.0012.
Site investigation is a poor means of discovering soil stiffness. Equipment development is urgently required.
Refined load control in pile testing is vital if test analysis is to be carried out.
Movements need to be sufficient to allow the ultimate load to be identified.
A method of removing creep or time related behaviour from test results is essential to proper interpretation.
All rapid tests and even those which are carried out as CRP tests distort ultimate loads Dilation is a key issue in causing peak and residual behaviour in pile tests.
Properly carried out maintained load tests rarely show peak and residual behaviour unless the pile is loaded so close to its ultimate value that it begins to accelerate.
Load cycling contributes little to the interpretation of pile test - the results related to repeated loads generally have to be discarded.
Interpretation of strength and stiffness from pile load tests appear to provide a reliable and consistent means of deriving values for design.
Malcom Puller asked about the effect of down-drag and wanted to know if the means of estimating it acceptable.The speakers said because it is a long term effect, which does not present itself in pile testing, it the question was difficult to answer. However, down-drag and heave can affect the results of pile tests. In effect, the mobilised shaft friction may not appear to have the values one might expect. Heave is a common problem with driven piles in many soils, both granular and clayey.
Puller also asked if stabilising fluids were necessary in cleaning pile bases and if bentonite was better than water. The speakers said that because bentonite helped keep solid particles in suspension better it might be said to be more useful.
Michael O'Brien asked if the recommended test procedure in the ICE Specifications for piling and retaining walls was satisfactory for analytical purposes. Fleming and England said the specification was an improvement on the previous version (1988).However, because of the trend towards computer controlled testing it needed reconsideration. The settlement rates with time are onerous and also need some re-thinking.
John Salmon asked for the speakers' views on impact load tests. Piles loaded quickly invariably show enhanced apparent capacities, they said. Rapid tests have no means of dealing with creep and soil stiffness and therefore are at a significant disadvantage compared with maintained load tests.
Richard Young said that the presentation suggested that dynamic testing was questionable. Fleming and England said that dynamic tests do not measure elastic shortening or creep. However it may be possible to correlate dynamic tests against static load tests, for the same pile dimensions and for the same ground conditions.
Simon Quarrell said the presentation was very critical of site investigation. He wondered if there was any remedy to the problem of the declining cost of investigation. The speakers said the industry was poor and employers did not understand the need for improvement. Eventually the needs will have to be recognised and most likely new electronic probes working on the earth-worm principle will be able to measure stress, strain and time functions at the same time.