Chapter VII. Results
Nearly 53 megabytes of data were collected during the test. MATLAB® software version 4.2b by The Mathworks Inc. was used for the data reduction. MATLAB is a powerful software package that executes FORTRAN-type commands to perform mathematical operations on raw data and to generate desired output.
After recalibrating the piles, load rod, and load cells, new calibration factors for the load cells were used to calculate the magnitudes of all of the loads in cycles 1, 5, 10, 25, and 50, which were the cycles chosen for analysis. It was found that the actual maximum load delivered to the pile group in each cycle was consistently near 2135 N (480 lb.). Based on this finding, the data starting at 667 N (150 lb.) and at every 222 N (50 lb.) increment thereafter, up to 2000 N (450 lb.), in the cycles mentioned above, were selected for analysis. The load at 667 N (150 lb.) was chosen as the minimum because at lower loads, the data acquisition system noise levels caused inconsistent output.
Four categories of output were produced from the test output:
Separate output was produced for the half-cycle in which the pile group was pulled to the left and for the half-cycle in which the group was pulled to the right. All figures, except for those showing load versus deflection, reflect this output format and if unlabelled, the curves on the left half of plots are for the left half-cycle and vice versa. The piles were not organized in the pile group in numerical order. Because of hardware defects, either in the piles themselves or in the data acquisition input boards, a number of strain gages did not function properly. The piles were ordered in the group according to their apparent reliability. That is why the group order was Pile 6, Pile 2, Pile 1, Pile 5, Pile 4, going from left to right. Referring to Figure 25 during this discussion will be helpful in keeping straight the pile order used in testing.
Some readily expected and also unusual output was produced from the strain gages mounted on the load rod, which measured the loads taken up by each pile in the group. Overall, results from the left loading and right loading show few similarities other than the fact that for almost all of the load increments in most of the cycles, the lead pile takes more of the load than any of the other piles. In cycle 1-Left, the lead pile (#6) takes more of the load for all load increments up to 2000 N (450 lb.), but the two trailing piles (#5, #4) take more load than the other two (#2, #1) as shown in Figure 27. Cycle 1-Right does not show a trend of this nature and all of the piles except the lead pile (#4) seem to assume the load fairly evenly.
In cycle 5 in both directions, the load distributions are similar to what was seen in cycle 1. The lead pile again takes more of the load than the other pile, with random distributions of load spread among the four other piles. In cycle 5-Right, as the load passes through the 1780 N (400 lb.) and 2000 N (450 lb.) increments, a very nice stair step distribution is shown. See Figure 28.
Cycle 10-Right shows probably the most consistent load distribution pattern for any of the cycles analyzed. The lead pile (#4) takes more of the load, with the second pile (#5) fairly consistently taking the next highest amount and then followed generally in turn by the fourth (#2), the third (#1), and then the trailing pile (#6). Cycle 10-Left shows only the lead pile (#4) taking the most load.
As the load increases to 1334 N (300 lb.) in both directions for cycle 25, the load is distributed evenly through the whole group, and not until the 1557 N (350 lb.) load increment does the lead pile in each direction again start to take more of the load than the other piles. Perhaps this is due to the displacement of soil from around the piles to the point where resistance to movement in the lower load regions is equal for all piles, until the lead pile again encounters undisturbed soil at the higher loads.
In cycle 50, it can clearly be seen up through the 1557 N (350 lb.) load increment that the trailing piles take more of the load than do the other piles. The leading pile (#6) in cycle 50-Left takes the least load of all of them until the 1780 N (400 lb.) load increment is reached, when it resumes and takes more load up through the last load step. The same phenomenon can be seen for the lead pile (#4) in cycle 50-Right.
Other than the consistent demonstration that the lead piles take more of the load than the other piles, except when much movement occurs where they may take the least load of all, no strong trend can be seen. It appears that piles other than the lead pile take about the same load with no real trend in hierarchy. Complete load distribution output can be found in Figures B-1 through B-5.
Unlike the output for load distribution, strong trends are easily seen in the group moment distribution output plots. The plots all show closely grouped moment curves for all the piles. The shape of the curves is similar for the left and right half-cycles. Usually, the moment curves for the lead pile in each loading direction stand out slightly from those of the other piles because the lead piles take more of the total load than do the other piles. This can be seen in Figure 29.
All through cycle 1, it can be seen that the piles are acting as "long" piles with points of counter-flexure starting at a depth of around 560 mm (22 in.) at a load of 667 N (150 lb.) and moving downward, until at a load of 2000 N (450 lb.), the points of counter-flexure are at a depth of about 760 mm (30 in.). A second point of counter-flexure can be seen in each pile in all load increments up to 1557 N (350 lb.), after which only one point can be seen.
Output for cycle 5 shows that only one point of counter-flexure exists in each pile, even at the lowest load increment. The points of counter-flexure are at approximately the same depth, 760 mm (30 in.), as they were for the 2000 N (450 lb.) load in cycle 1. As in cycle 1, the points of counter-flexure move down the piles, ending at a depth of about 890 mm (35 in.) at the highest load. Even though the points of counter-flexure move down the piles as the load increases, the depths at which the maximum moments occur remain fairly constant through the whole cycle.
In cycle 10, the output plots are much the same as from cycle 5. The points of counter-flexure move down the piles as the load increases, and the depth at which the maximum moment occurs is a little deeper than in cycle 5.
At a load of 1557 N (350 lb.) in cycle 25, the points of counter-flexure disappear altogether, and a moment of 0 is registered at the lowest strain gages in the piles, as seen in Figure 30. This demonstrates that the piles have become "short" piles. The points of counter-flexure occur in the loads lower than this in this cycle, but remain unseen in the loads higher than this. The depth at which the maximum moment occurs is again lower than in the previous cycles.
Cycle 50 shows some interesting results. At the load of 667 N (150 lb.), the bending moment output for the lead piles in each half-cycle is distinctly lower than that of the other piles as shown in Figure 31. This correlates well with the load distribution data for this same cycle and load. There also are no counter-flexure points at this or any other load in this cycle. Unlike the situation discussed for cycle 25 at the 1557 N (350 lb.) load, where the moment is 0 at the bottom of the piles, moment readings occur at all depths through this whole cycle. Because of this, it can be reasoned that even though the piles act as short piles, the bottoms of the piles are still fixed; otherwise, there would be no reaction to cause the bending that is registered at the pile tips.
By examining the data as a whole, several other trends can be seen. One trend that cannot be readily explained is the often occurring mismatched magnitudes of bending moment from the left and right half-cycles. While on some plots the moment data matches up well, on other plots the moment data in one half-cycle is larger than in the other half, with the trend showing larger moments for the left half-cycle. Soil inhomogeneity is a possible reason but not an absolute explanation. It was interesting to note that from cycle to cycle, the depth at which the maximum moment occurs increases, but remains relatively constant as the load increases in each cycle, while the depth at which counter-flexure occurs increases during the cycle. For complete group moment distribution output, refer to Figures B-6 through B-16.
Moment distribution output from the individual piles at each load increment and for all cycles creates an interesting plot, and some good comparisons can be made. At each of the seven load increments, the output for each pile for all five cycles can be examined for trends and similarities.
The plots of the lead piles, Pile 6 when loading was to the left, and Pile 4 when loading was to the right, have symmetric curves for all loads and cycles as seen in Figures 32 and 33. The magnitudes of the moment data do not match as well as the curve shapes, but the effects of acting as a leading pile and as a trailing pile are evident. When a pile is leading, the maximum moment in each cycle occurs at a shallower location than it does in the same cycle when trailing. At the lowest loads, 667 N (150 lb.) and 890 N (200 lb.), in cycle 50, it can be seen that more bending moment occurred when the piles were acting as trailing piles than when they were acting as leading piles. This correlates well with the load distribution output already discussed.
The output from the inner piles (#2,#5) shows some interesting trends. For the early cycles, the moment in each of these piles is almost exactly the same regardless of loading direction. As the number of cycles increases and at lower loads, each of the piles has a greater moment when acting as a trailing pile than when acting as a leading pile. For example, when loading was to the left in cycle 50 at 1112 N (250 lb.), Pile 5 had more moment than Pile 2, and when loading was to the right in this same cycle at this same load, Pile 2 had more moment than Pile 5. This trend continued until the load approached its maximum value, at which each pile took almost the same amount of the load, whether leading or trailing.
The moment curve shapes for the middle pile (#1) are more symmetric than any of the other piles, but the magnitudes at the maximum values do not agree for left and right loading. As stated in the last section, this phenomenon has yet to be explained. Complete test output for individual pile moment distribution is contained in Figure D-1.
The load versus deflection output was straight forward. As load magnitude increased in each of the five cycles, the deflection increased in a linear fashion. If the deflection at each load increment is compared from cycle to cycle, the deflection does not increase in a linear fashion. For example, deflection of the pile top at the maximum load of 2000 N (450 lb.) at cycle 10 was not 10 times greater than it was in cycle 1 at this same load. This can be seen in Figure 34.
The load versus deflection curve for cycle 50 showed a lesser deflection at a load of 667 N (150 lb.) than the deflection at this load for cycle 25. This can be explained by the conditions during testing. In the latter cycles of the test, when the target load was reached, the pile group was pulled back in the direction opposite the one in which it was being loaded, to unload it. It was pulled back until the load on the active load cell dropped below the cutoff threshold, whereupon the hydraulic cylinders stopped moving. Since the pile group had been pulled so far to one side, soil resistance continued to hold the pile group in a displaced position, and after the cutoff threshold was reached, the piles moved back slightly in the direction in which they had just been loaded. This produced a starting point for the next half-cycle that was slightly displaced in the direction opposite to the one in which the next loading would occur. For complete load versus deflection output, refer to Figure B-16.
Florida Pier is a 3-D, nonlinear, finite element analysis program developed at the University of Florida for designing piles, pile groups, and drilled shafts. Predictions were made using this software and the properties of the model pile group and of the test soil, to validate the Florida Pier program through a comparison with the testing results. Florida Pier is capable of modeling the conditions under which the testing occurred, so no scaling of the test output was necessary for making the comparison. Predictions also were made using the COM624P software by following the pile group method outlined by the Federal Highway Administration (U.S. Department of Transportation, 1996). This prediction also was made for a linear, five-pile group, laterally loaded in a cyclic fashion 50 times, using the soil properties of the test soil. Predicted bending moment output from each of the computer programs only applied to the lead pile in the group. Pile head deflection was modeled for pinned pile connections and the output reflects the displacement at the cap level, 280 mm (11 in.) above the top of the soil, where the load was applied.
In cycle 50 at a load of 2000 N (450 lb.), the bending moment on the lead piles reached an average maximum of 237 N-m (2100 in-lb.) in the test. Florida Pier predicted a value of about 180 N-m (1600 in-lb.) and COM624P predicted a value of 264 N-m (2340 in-lb.). Figure 35 shows the results of the predictions compared with the actual test results. Measured response was bracketed on both sides by the predictions with the COM624P prediction being the closer of the two. The measured response was quite linear, whereas both of the predictions had curved lines.
For pile group top displacement, Florida Pier predicted 32 mm (1.25 in.) of deflection, COM624P predicted 29 mm (1.12 in.) of deflection, and testing ceased at 62.2 mm (2.45 in.) of deflection. The predictions were far under the measured displacement after 50 cycles. Figure 36 shows the deflection predictions and test results.
As a matter of interest, the measured results for the lead piles from cycle 1 also were compared with predictions. Maximum measured bending moment reached just under 170 N-m (1500 in-lb.), and was quite close to the prediction made using Florida Pier. Displacement measured 15.7 mm (0.62 in.), and was close to the prediction made by COM624P up to the 1334 N (300 lb.) load increment.