- Email: [email protected]
Seismic tests of a pile-supported structure in liquefiable sand using large-scale blast excitation
Nuclear Engineering and Design 228 (2004) 367–376
Seismic tests of a pile-supported structure in liquefiable sand using large-scale blast excitation Naotaka Kamijo a , Hideaki Saito a , Kazuhiro Kusama b , Osamu Kontani b,∗ , Robert Nigbor c a
c
Power Engineering R&D Center, Tokyo Electric Power Co., 4-1 Egasaki, Tsurumi-ku, Yokohama 230-8510, Japan b Kobori Research Complex Inc., 6-5-30, Akasaka, Minato-ku, Tokyo 107-8502, Japan Civil Engineering Department, University of Southern California, University Park, KAP210, Los Angeles, CA 90089, USA Received 8 February 2002; received in revised form 17 June 2003; accepted 23 June 2003
Abstract Extensive, large-amplitude vibration tests of a pile-supported structure in a liquefiable sand deposit have been performed at a large-scale mining site. Ground motions from large-scale blasting operations were used as excitation forces for vibration tests. A simple pile-supported structure was constructed in an excavated 3 m-deep pit. The test pit was backfilled with 100% water-saturated clean uniform sand. Accelerations were measured on the pile-supported structure, in the sand in the test pit, and in the adjacent free field. Excess pore water pressures in the test pit and strains of one pile were also measured. Vibration tests were performed with six different levels of input motions. The maximum horizontal acceleration recorded at the adjacent ground surface varied from 20 Gals to 1353 Gals. These alternations of acceleration provided different degrees of liquefaction in the test pit. Sand boiling phenomena were observed in the test pit with larger input motions. This paper outlines vibration tests and investigates the test results. © 2003 Elsevier B.V. All rights reserved.
1. Introduction In current design practice for reactor buildings in Japan, a reactor building is required to be constructed by a direct foundation on a firm shallow rock layer. Because there are few such locations, it is very difficult to find suitable locations. Pile foundations have become very popular for ordinary buildings, including high-rise buildings in Japan. Pile foundations are popular because piles can be founded on deep or shallow supporting layers by simply adjusting their lengths. If pile foundations ∗ Corresponding author. Tel.: +81-3-5561-2434; fax: +81-3-5561-2345. E-mail address: [email protected] (O. Kontani).
can be introduced in construction of nuclear facilities, the number of suitable sites will be increased. However, before pile foundations can be used in nuclear facilities, it is important to experimentally demonstrate their applicability. Since very strong earthquake motions are employed for designing nuclear facilities, understanding of dynamic nonlinear soil–pile-structure interaction including liquefaction of surrounding soil layers need to be incorporated to structural design of pile foundations. The objectives of this experimental research are (1) to achieve a better understanding of the dynamic nonlinear behavior, including liquefaction, of soil– pile-structure interaction systems under severe ground motions and (2) to provide data for verification and validation of nonlinear response analysis method and
0029-5493/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.nucengdes.2003.06.029
368
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
a structural design method for pile foundation structures. To meet those objectives, seismic tests were conducted on a pile-supported structure in liquefiable sand fill using ground motions from large mining blasts at Black Thunder Mine, Wyoming, USA. This paper outlines seismic tests and investigates test results. The simulation analyses of seismic tests and investigation of a structural design method for pile foundations will be presented in a separate paper (Kontani et al., 2001).
2. Vibration test method using ground motions caused by mining blasts The vibration test method using ground motions caused by mining blasts is shown schematically in Fig. 1. Common ways for performing vibration tests of structures are Shaking Table Tests, Forced Vibration Tests and Earthquake Observations. These methods are useful in many ways. However, none are capable of shaking a large-scale soil–pile-structure system at large amplitude. The vibration test method using ground motions caused by mining blasts has the following advantages over the conventional test methods. • large-scale test structures can be used for the test; • ground motions of various amplitudes can be generated in accordance with distances from the blast area; • three-dimensional effects and perfect soil-structure interaction in the actual ground can be considered. Seismic tests of a pile-supported structure in a liquefiable sand deposit were conducted at Black Thunder Mine of Arch Coal, Inc. Black Thunder Mine is one of the largest coal mines in North America and is located in northeast Wyoming, USA. Since its opera-
tion is very active, there were many opportunities to have large ground motions at the test site. At the mine, there is overburden (mudstone layers) over the coal layers. The overburden is dislodged by large blasts called “Cast Blasts.” These are array explosions with a width of 60 m and lengths up to a kilometer; typical Cast Blasts use 500–2000 tons of chemical explosives. After the cast blasting, the rubble is removed by heavy earthmoving equipment and draglines to expose the coal. After the coal surface is exposed, smaller blasts called “Coal Shots” are applied to loosen coal layers. The coal is then mined out by truck and shovel operations. Ground motions caused by Cast Blasts were used for seismic tests conducted in this research. The smaller Cast Blasts or Coal Shots were used for checking and calibrating instrumentation. The large Cast Blasts near the test site produced ground accelerations up to 4000 Gal (1 Gal = 1 cm/s/s = 0.001 g).
3. Outlines of seismic tests A sectional view and a top view of the test pit and the pile-supported structure are shown in Figs. 2 and 3, respectively. A 12 m × 12 m test pit was excavated 3-m deep with a 45◦ slope, as shown in Fig. 2. A waterproofing layer was made of high-density plastic sheets and was installed in the test pit in order to maintain 100% water-saturated sand. Details of the pile-supported structure are shown in Fig. 4. Four piles were made of steel tube. Pile tips were closed by welding. Piles were embedded 70 cm into the mudstone layer. The top slab and the base mat were made of reinforced concrete and were con-
Test Structure Blast Area
Test Pit
Mudstone Layer
Explosive
Coal Layer
Fig. 1. Vibration test method at mining site.
Fig. 2. Section view of test pit.
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
Triaxial Accelerometer Uniaxial Accelerometer
369
Pore Water Pressure Meter Strain Gage
3m 3m
Blast Area GL -0.0m
3m GL -3.0m 6m
GL -10.0m GL -20.0m
Fig. 3. Top view of test pit.
Top Slab
Base Mat
GL -30.0m
Fig. 5. Instrumentation plan.
nected by H-shaped steel columns. The construction schedule was planned so that the structure received the least influence from mining blasts while under construction. Instrumentation is shown in Fig. 5. Accelerations were measured on the structure and on one of the four piles. Accelerations in the sand deposit and in the native ground adjacent to the sand pit (“free field”) were also measured by arrays of accelerometers. Axial and bending strains of one pile were measured at seven levels; these dynamic strains were used to calculate stress and then to evaluate bending moment distribution within the pile. Excess pore water pressures were measured at four levels in the test pit to investigate liquefaction phenom-
Fig. 4. Details of pile-supported structure.
ena. Sensors embedded in the test pit were sealed and protected against the water–sand mixture. Great care was taken in preparing pore pressure transducers. Each transducer was installed in a plastic cylinder, which was then wrapped in a glass fiber sheet filled with silica sand for protection. The air in the porous stone filter in front of the transducer diaphragms was removed to ensure full saturation of the sensor. This treatment was done by heating and vacuuming in a water-filled glass container (Hushmand and Scott, 1992). Seismic velocity measurements (P- and S-wave velocities) were conducted in the test site to characterize the physical properties of the native soil layers. These measurement results are shown in Fig. 6. The overburden consisted of several layers of siltstone or mudstone over the deeper coal layer. The shear wave velocity at the test pit bottom was about 200 m/s and this increased to 500–700 m/s with increasing depth. Core samples of the native soil layers were collected for laboratory tests. These included index property, dynamic resonant column, and dynamic torsional shear tests. Samples of the backfill sand were also tested in the laboratory for both index properties and for dynamic nonlinear properties. The grain size distribution of the backfill sand is shown in Fig. 7. The sand was found at a borrow pit near Black Thunder Mine, and was aeolian in nature. Great care was taken in backfilling the test pit with the sand, because the sand needed to be 100% water-saturated and air had to
370
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
Shear Wave
Longitudinal Wave
0
Ground Level (m)
10
20
30
Fig. 8. Installation of waterproof layer.
40 0
500
1000 1500 2000 Wave Velocity (m/s)
2500
Fig. 6. Results of seismic velocity (Vp and Vs ) measurement in native ground.
be removed in order to ensure a liquefiable sand deposit. Shear-wave velocity of the sand was measured in situ by an active source and imbedded sensors. This varied from 70 m/s immediately after construction to 93 m/s after the final test. These values are at the low end of the normal range of seismic velocities for sand, and are therefore susceptible to liquefaction. Fig. 8 shows the installation of the waterproof layer. Fig. 9 shows the completed pile-supported structure and the test pit. The water level was kept at 10 cm above the sand surface throughout seismic tests to prevent dry out of the sand deposit.
100 Weight Passed (%)
Mean Grain Size D =0.3mm 50
80
Uniformity Coef. U =3.68 C Fine Contents Fc=10%
60 40 20
Test results of three samples are presented.
rain Size (mm)
0 10
-3
10
-2
-1
10 10 Grain Size (mm)
0
Fig. 7. Grain size distribution of backfill sand.
10
1
Fig. 9. Completion test structure and test pit.
4. Seismic test results Six seismic tests were conducted on the pilesupported structure. The locations of the blast areas for each test are shown in Fig. 10. The blast areas were about 60 m wide and their lengths depended on the mining schedule. The results of the seismic tests are summarized in Table 1. The maximum horizontal acceleration recorded on the adjacent ground surface varied from 20 Gals to 1353 Gals depending on distance from the blast area to the test site. The closest blast was only 90 m from the test site. These differences in the maximum acceleration yielded responses at different levels and liquefaction of different degrees. Sand boiling phenomena were observed in the test
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
371
Table 1 Summary of seismic tests Level of input motions
Test number
Distance (m)a
Maximum acceleration (gal)b EW
a b
NS
Vertical
Distance from blast area to test site. At the ground level of adjacent free field.
4.1. Measurement records at free field and sand deposit The maximum accelerations recorded in the adjacent free field in vertical array configurations are compared for three tests in Fig. 11. The amplification tendencies from GL-32m (GL-32 m = 32 m below Ground Level) to the surface were similar in the mudstone layers for three tests. The maximum accelerations recorded through the mudstone layers to the sand deposit are compared for three tests in Fig. 12. There was a clear difference among the amplification trends in the test pit. Test-1 showed a similar amplification trend to that of the mudstone layers as shown in Fig. 11. Test-5 showed less amplification in the sand deposit. Test-3 showed a large decrease in acceleration in the test pit because of severe liquefaction of the sand deposit. Fig. 10. Locations of test site and blasts in seismic tests (north is up).
0 Ground Level (m)
-5
pit with larger input motions. This is one of the most advantageous features of the test method employed in this project, although the input motions were not controlled mechanically or electrically. In this paper, three tests indicated by highlights in Table 1 were chosen for detailed investigations, because those tests provided three different phenomena in terms of liquefaction of the sand deposit as well as in terms of dynamic responses of the structure. Results from these three tests are detailed in the next section.
-10 -15 -20 -25 -30 Test-1
Test-5
Test-3
-35 0
10
100 200 0 20 30 0 Acceleration (Gal)
400
800
Fig. 11. Maximum acceleration in free-field vertical array.
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
0
100
-2
0 Acceleration (Gal)
Ground Level (m)
372
-4 -6 -8 Test-1
Test-5
Test-3
-10 0
40 0 50 100 150 0 200 400 600 Acceleration (Gal)
20
Sand Surface
Max=84.2Gal
-100 200
Free Field Surface
0 Max=142Gal
-200 100
GL-32m
0 Max=52.0Gal
-100 Fig. 12. Maximum acceleration in test pit (0–3 m) and mudstone layers.
0
2
4 Time (s)
6
8
Acceleration (Gal)
600
Acceleration records at the sand surface, the free field surface and GL-32m are compared for Test-1 (small input level) in Fig. 13. The response spectra from these records are also shown in the figure. The same set of acceleration records and these response spectra are shown in Fig. 14 for Test-5 (medium input level) and in Fig. 15 for Test-3 (large input level).
Sand Surface Free Field GL-32m h=0.05
400 200 0 0.02
0.10
1.00
5.0
Period (s) 50
Fig. 14. Acceleration records of Test-5 (medium input level).
Sand Surface
Acceleration (Gal)
0 Max=27.8Gal
-50 30
Free Field Surface
0 Max=20.0Gal
-30 10
GL-32m
0 Max=6.9Gal
-10 0
2
4 Time (s)
6
8
Acceleration (Gal)
150 Sand Surface Free Field GL-32m h=0.05
100 50 0 0.02
As can be seen from Fig. 13 for Test-1, over all the frequency regions, the responses at the sand surface were greater than those at the free field surface, and the responses at the free field surface were greater than those at GL-32m. From Fig. 14 for Test-5, the responses at the sand surface and the free field surface were greater than those at GL-32m over all frequency regions. The responses at the sand surface became smaller than those at the free field surface for the period less than 0.4 s due to in a certain degree of liquefaction of the sand. From Fig. 15 for Test-3, the responses at the sand surface became much smaller than those at the free field surface and even smaller than those at GL-32m. These responses reductions in the test pit were caused by extensive liquefaction over the test pit, because shear waves could not travel in the liquefied sand. 4.2. Measurement results for the structure
0.10
Period (s)
1.00
5.0
Fig. 13. Acceleration records of Test-1 (small input level).
Fig. 16 compares the maximum recorded accelerations of the pile-supported structure and in the test
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
100
0
Sand Surface
Ground Level (m)
Acceleration (Gal)
0 Max=105Gal
-100 700
Free Field Surface
0 Max=579Gal
-700 400
Acceleration (Gal)
Test-5
Test-3
-3 2
40 5 10 15 0 10 Bending Moment (kN.m)
20
30
Max=272Gal 0
2
4 Time (s)
6
8
Sand Surface Free Field GL-32m h=0.05
2000
1000
0 0.02
0.10
1.00
5.0
Period (s) Fig. 15. Acceleration records of Test-3 (large input level).
pit for three tests. As can be seen for Test-3, there were differences in maximum response between the pile-supported structure and the sand deposit, which means that the pile and surrounding sand did not behave in the same manners. In Test-3, unlike in the other cases, the maximum acceleration decreased as motions traveled upward. The maximum bending mo-
3 Structure
Ground Level (m)
-2
0
-400
Structure
Structure
Sand
0 Pile
Pile Sand
Pile
Sand
Test-1 0
-1
Test-1 GL-32m
0
-3
373
50
Test-5
Test-3
100 0 50 100 150 0 200 400 600 Acceleration (Gal)
Fig. 16. Maximum acceleration of sand fill and test structure.
Fig. 17. Maximum pile bending moment distribution.
ment distributions of the pile are shown in Fig. 17. These bending moments were calculated from the measured bending strains of the instrumented pile at the seven instrumented levels by first calculating stress from strain and then calculating moment from stress. The bending moment took its maximum value at the pile head for all cases. However, the moment distribution shapes differed and the inflection points of curves moved downward in accordance with the input motion levels, in other words, the degrees of liquefaction in the test pit. Fig. 18 compares the acceleration records at the top slab, the base mat and GL-3m of the pile for Test-1 (small input level) . The response spectra from these records are also shown. The same set of acceleration records and these response spectra are shown in Fig. 19 for Test-5 (medium input level) and in Fig. 20 for Test-3 (large input level). As can be seen from Fig. 18 for Test-1, the maximum accelerations increased as motions went upward. For all frequency regions, the responses at the top slab were greater than those at the base mat, and the responses at the base mat were greater than those at GL-3m of the pile. The first natural period of the soil–pile-structure system was about 0.2 s under the input motion level of Test-1. From Fig. 19 for Test-5, the maximum accelerations increased as motions went upward. For almost all frequency regions, the responses at the top slab were greater than those at GL-3m of the pile. The responses at the base mat were greater than those at GL-3m of the pile for periods greater than 0.2 s and were similar to
374
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
100
200
Top Slab
Max=73.9Gal
50
Base Mat
0 -50
Max=42.3Gal
20
GL-3m (Pile)
0
Max=118Gal
-200 100
Base Mat
0 Max=83.2Gal
-100 100
GL-3m (Pile)
0 Max=13.0Gal
-20 0
2
4 Time (s)
6
0
2
4 Time (s)
6
8
600
200
0.10
1.00
5.0
Period (s) Fig. 18. Acceleration records at test structure (Test-1: small input level).
those at GL-3m for periods less than 0.2 s. The peaks in response spectra were at about 0.3 s, and were shifted by 0.1 s from those of Test-1. For Test-3, the maximum accelerations decreased as motions went upward, which were different from those of the other two tests. The responses at the top slab and the base mat became smaller than or similar to the responses at GL-3m of the pile. Compared with the other two tests, it became difficult to identify peaks corresponding to natural periods of the soil–pile-structure system from response spectra diagrams. The structure responses from the three tests described above indicate that soil nonlinearity and liquefaction greatly influence the dynamic properties of pile-supported structures. 4.3. Excess pore water pressure ratio The dynamic changes of excess pore water pressure ratios are shown in Fig. 21. The excess pore
Acceleration (Gal)
Top Slab Base Mat GL-3m(Pile) h=0.05
400
0 0.02
Max=64.9Gal
-100 8
600 Acceleration (Gal)
Top Slab
0
-100
Acceleration (Gal)
Acceleration (Gal)
0
Top Slab Base Mat GL-3m(Pile) h=0.05
400 200 0 0.02
0.10
1.00
5.0
Period (s) Fig. 19. Acceleration records at test structure (Test-5: medium input level).
water pressure ratio is the ratio of excess pore water pressure to initial effective stress, calculated from the laboratory-measured properties of the sand and the depth of each pore pressure sensor. A ratio of one indicates complete liquefaction, while a ratio less than one indicates the initiation of liquefaction and/or partial liquefaction. High-frequency components of the pore pressure ratio that exceed unity arise from large vertical ground accelerations that dynamically weight/unweight the sand mass above each pore pressure sensor. In Test-1, the maximum ratio stayed around zero, which means that no liquefaction took place. In Test-5, the ratios rose rapidly, reaching around one at GL-0.6m and GL-1.4m after the main vibration was finished. Ratios at GL-2.2m and GL-3.0m were about 0.7 and 0.5. The measurement showed that the liquefaction region was in the upper half of the test pit. In Test-3, ratios at all levels rose rapidly, reaching around one, which indicates extensive liquefaction over the entire region.
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
400
Test-1
Top Slab
375
Test-5
Test-3
2 GL -0.6m
0 -400
Max=277Gal
400
Base Mat
0 Max=302Gal
-400 600
GL-3m (Pile)
0 Max=522Gal
-600
Acceleration (Gal)
0
2
4 Time (s)
2000
6
8
Top Slab Base Mat GL-3m(Pile) h=0.05
1000
Excess Pore Water Pressure Ratio
Acceleration (Gal)
1 0 2
GL -1.4m
1 0 -1 2
GL -2.2m
1 0 -1 2 GL -3.0m 1 0 -1 0
2
4
6
Time (sec)
0 0.02
0.10
Period (s)
1.00
5.0
Fig. 20. Acceleration records at test structure (Test-3: large input level).
4.4. Evaluation of natural frequencies and damping factors of soil–pile-structure system Using measured ground motion inputs and structure responses, vibration properties of the soil–pilestructure system were identified by fitting test records to the single-input-multi-output ARX model (the auto-regressive model) (Ikeda, 2000). The soil–pilestructure system was modeled by a simple lumped mass model, as shown in Fig. 22. Acceleration records at the test pit bottom were used as input motions, and acceleration records of the base mat and the top slab were regarded as output motions in this investigation. The dynamic variation of natural frequencies and damping factors evaluated with time are shown for the first natural mode of the system in Fig. 23. The frequencies of the first mode decreased by ground motions and these values were varied in accordance with input motion levels. The first natural frequencies were around 7 Hz for all tests before subject to the ground motions. For Test-3, the frequency was decreased by half after the ground motions.
Fig. 21. Excess pore water pressure ratio at different depths in sand, measured by pore pressure sensors at four levels.
The effective damping factors of the first mode (as estimated by the auto-regressive model) increased during blast ground motions. The damping also varied with the input motion levels of the three tests. For Test-1, the effective damping factor was less than 7%. For Test-3, the effective damping factor was considerably higher at 20–40%. During extensive liquefaction in the test pit, the natural frequency of the soil–pile-structure system Motions at Top Slab
Output (Top Slab)
Output Motions at Base Mat (Base Mat)
Motions at GL-3m
Input (Motions at GL-3m)
Fig. 22. Model for identifying vibration properties of soil–pilestructure system.
376
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
Frequency (Hz) Damping (%)
Test-1 8
40
Test-5
Test-3
4 0
20 0 0
10
20 Time (s)
30
40
Fig. 23. Evaluated natural frequencies and damping.
decreased because the sand lost its stiffness, however, the effective damping factor was increased considerably because of the nonlinear behavior of liquefied sand.
5. Conclusions (1) Seismic tests were conducted of a pile-supported structure in a liquefiable sand deposit six times using large ground motions from large mining blasts. Liquefaction occurred in the sand to varying levels up to full liquefaction. Nonlinear responses of the soil–pile-structure system at different levels were obtained for various levels of liquefaction in the test pit. (2) The maximum horizontal acceleration recorded at the adjacent ground surface varied from 20 Gals to 1353 Gals depending on distance from the test site to the blast areas. This is one of the advantages of the vibration test method employed in this project. (3) In the adjacent free field, motions were amplified from GL-32m to the ground surface, regardless of input motion levels. In the test pit, amplification or reduction of motions depended upon input motion levels. Motions were amplified at small input levels and decreased at large input levels from the bottom to the surface of the test pit. (4) Generation of pore water pressures depended upon input motion levels. Liquefaction started from the shallow part and extended to the deeper part of
the test pit, and finally the test pit was completely liquefied as in Test-3. (5) Bending moments were a maximum at the pile heads, regardless of input motion levels. However, the moment distribution shapes varied and the inflection points of the distribution curves moved downward in accordance with input motion levels, in other words, the degrees of the liquefaction in the test pit. (6) The vibration test method employed in this experimental research was found very useful and effective for investigating the dynamic behavior of large model structure under severe ground motions. Acknowledgements We would like to express our deep appreciation to Dr. Aoyama, Professor Emeritus of University of Tokyo, for his guidance throughout this experimental research. We would also like to thank Dr. Miura, Professor of University of Hiroshima, for his review of the experimental results and the analysis results. The project staff from OYO International, Agbabian Associates, GeoVision, and CE&MT are also thanked for their dedicated assistance with the field studies. Most importantly, we would like to thank the management and staff of Arch Coal’s Black Thunder Mine for their assistance throughout the project. Special thanks go to Paul Lang, to Gordon Shinkle, to Steve Beil and to the late Bob Martin. They were always very generous and provided us with support in many ways. Without their help, this work would not have been possible. References Kontani, O., Tanaka, H., Ishida, T., Miyamoto, Y., Koyamada, K., 2001. Simulation analyses of seismic tests of a pile-supported structure in liquefiable sand using large-scale blast excitation. In: Proceeding of the 16th International Conference on Structural Mechanics in Reactor Technology (SMiRT-16), Washington. Hushmand, B., Scott, R.F., 1992. In-place calibration of USGS pore pressure transducers at Wildlife Liquefaction Site, California, USA. In: Proceeding of 10th World Conference on Earthquake Engineering, pp. 1749–1754. Ikeda, Y., 2000. System Identification of 10-story building and control effect of installed AMDs based on earthquake observation records. J. Struct. Eng., AIJ 46B (2000) 335–344 (in Japanese).
Seismic tests of a pile-supported structure in liquefiable sand using large-scale blast excitation Naotaka Kamijo a , Hideaki Saito a , Kazuhiro Kusama b , Osamu Kontani b,∗ , Robert Nigbor c a
c
Power Engineering R&D Center, Tokyo Electric Power Co., 4-1 Egasaki, Tsurumi-ku, Yokohama 230-8510, Japan b Kobori Research Complex Inc., 6-5-30, Akasaka, Minato-ku, Tokyo 107-8502, Japan Civil Engineering Department, University of Southern California, University Park, KAP210, Los Angeles, CA 90089, USA Received 8 February 2002; received in revised form 17 June 2003; accepted 23 June 2003
Abstract Extensive, large-amplitude vibration tests of a pile-supported structure in a liquefiable sand deposit have been performed at a large-scale mining site. Ground motions from large-scale blasting operations were used as excitation forces for vibration tests. A simple pile-supported structure was constructed in an excavated 3 m-deep pit. The test pit was backfilled with 100% water-saturated clean uniform sand. Accelerations were measured on the pile-supported structure, in the sand in the test pit, and in the adjacent free field. Excess pore water pressures in the test pit and strains of one pile were also measured. Vibration tests were performed with six different levels of input motions. The maximum horizontal acceleration recorded at the adjacent ground surface varied from 20 Gals to 1353 Gals. These alternations of acceleration provided different degrees of liquefaction in the test pit. Sand boiling phenomena were observed in the test pit with larger input motions. This paper outlines vibration tests and investigates the test results. © 2003 Elsevier B.V. All rights reserved.
1. Introduction In current design practice for reactor buildings in Japan, a reactor building is required to be constructed by a direct foundation on a firm shallow rock layer. Because there are few such locations, it is very difficult to find suitable locations. Pile foundations have become very popular for ordinary buildings, including high-rise buildings in Japan. Pile foundations are popular because piles can be founded on deep or shallow supporting layers by simply adjusting their lengths. If pile foundations ∗ Corresponding author. Tel.: +81-3-5561-2434; fax: +81-3-5561-2345. E-mail address: [email protected] (O. Kontani).
can be introduced in construction of nuclear facilities, the number of suitable sites will be increased. However, before pile foundations can be used in nuclear facilities, it is important to experimentally demonstrate their applicability. Since very strong earthquake motions are employed for designing nuclear facilities, understanding of dynamic nonlinear soil–pile-structure interaction including liquefaction of surrounding soil layers need to be incorporated to structural design of pile foundations. The objectives of this experimental research are (1) to achieve a better understanding of the dynamic nonlinear behavior, including liquefaction, of soil– pile-structure interaction systems under severe ground motions and (2) to provide data for verification and validation of nonlinear response analysis method and
0029-5493/$ – see front matter © 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.nucengdes.2003.06.029
368
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
a structural design method for pile foundation structures. To meet those objectives, seismic tests were conducted on a pile-supported structure in liquefiable sand fill using ground motions from large mining blasts at Black Thunder Mine, Wyoming, USA. This paper outlines seismic tests and investigates test results. The simulation analyses of seismic tests and investigation of a structural design method for pile foundations will be presented in a separate paper (Kontani et al., 2001).
2. Vibration test method using ground motions caused by mining blasts The vibration test method using ground motions caused by mining blasts is shown schematically in Fig. 1. Common ways for performing vibration tests of structures are Shaking Table Tests, Forced Vibration Tests and Earthquake Observations. These methods are useful in many ways. However, none are capable of shaking a large-scale soil–pile-structure system at large amplitude. The vibration test method using ground motions caused by mining blasts has the following advantages over the conventional test methods. • large-scale test structures can be used for the test; • ground motions of various amplitudes can be generated in accordance with distances from the blast area; • three-dimensional effects and perfect soil-structure interaction in the actual ground can be considered. Seismic tests of a pile-supported structure in a liquefiable sand deposit were conducted at Black Thunder Mine of Arch Coal, Inc. Black Thunder Mine is one of the largest coal mines in North America and is located in northeast Wyoming, USA. Since its opera-
tion is very active, there were many opportunities to have large ground motions at the test site. At the mine, there is overburden (mudstone layers) over the coal layers. The overburden is dislodged by large blasts called “Cast Blasts.” These are array explosions with a width of 60 m and lengths up to a kilometer; typical Cast Blasts use 500–2000 tons of chemical explosives. After the cast blasting, the rubble is removed by heavy earthmoving equipment and draglines to expose the coal. After the coal surface is exposed, smaller blasts called “Coal Shots” are applied to loosen coal layers. The coal is then mined out by truck and shovel operations. Ground motions caused by Cast Blasts were used for seismic tests conducted in this research. The smaller Cast Blasts or Coal Shots were used for checking and calibrating instrumentation. The large Cast Blasts near the test site produced ground accelerations up to 4000 Gal (1 Gal = 1 cm/s/s = 0.001 g).
3. Outlines of seismic tests A sectional view and a top view of the test pit and the pile-supported structure are shown in Figs. 2 and 3, respectively. A 12 m × 12 m test pit was excavated 3-m deep with a 45◦ slope, as shown in Fig. 2. A waterproofing layer was made of high-density plastic sheets and was installed in the test pit in order to maintain 100% water-saturated sand. Details of the pile-supported structure are shown in Fig. 4. Four piles were made of steel tube. Pile tips were closed by welding. Piles were embedded 70 cm into the mudstone layer. The top slab and the base mat were made of reinforced concrete and were con-
Test Structure Blast Area
Test Pit
Mudstone Layer
Explosive
Coal Layer
Fig. 1. Vibration test method at mining site.
Fig. 2. Section view of test pit.
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
Triaxial Accelerometer Uniaxial Accelerometer
369
Pore Water Pressure Meter Strain Gage
3m 3m
Blast Area GL -0.0m
3m GL -3.0m 6m
GL -10.0m GL -20.0m
Fig. 3. Top view of test pit.
Top Slab
Base Mat
GL -30.0m
Fig. 5. Instrumentation plan.
nected by H-shaped steel columns. The construction schedule was planned so that the structure received the least influence from mining blasts while under construction. Instrumentation is shown in Fig. 5. Accelerations were measured on the structure and on one of the four piles. Accelerations in the sand deposit and in the native ground adjacent to the sand pit (“free field”) were also measured by arrays of accelerometers. Axial and bending strains of one pile were measured at seven levels; these dynamic strains were used to calculate stress and then to evaluate bending moment distribution within the pile. Excess pore water pressures were measured at four levels in the test pit to investigate liquefaction phenom-
Fig. 4. Details of pile-supported structure.
ena. Sensors embedded in the test pit were sealed and protected against the water–sand mixture. Great care was taken in preparing pore pressure transducers. Each transducer was installed in a plastic cylinder, which was then wrapped in a glass fiber sheet filled with silica sand for protection. The air in the porous stone filter in front of the transducer diaphragms was removed to ensure full saturation of the sensor. This treatment was done by heating and vacuuming in a water-filled glass container (Hushmand and Scott, 1992). Seismic velocity measurements (P- and S-wave velocities) were conducted in the test site to characterize the physical properties of the native soil layers. These measurement results are shown in Fig. 6. The overburden consisted of several layers of siltstone or mudstone over the deeper coal layer. The shear wave velocity at the test pit bottom was about 200 m/s and this increased to 500–700 m/s with increasing depth. Core samples of the native soil layers were collected for laboratory tests. These included index property, dynamic resonant column, and dynamic torsional shear tests. Samples of the backfill sand were also tested in the laboratory for both index properties and for dynamic nonlinear properties. The grain size distribution of the backfill sand is shown in Fig. 7. The sand was found at a borrow pit near Black Thunder Mine, and was aeolian in nature. Great care was taken in backfilling the test pit with the sand, because the sand needed to be 100% water-saturated and air had to
370
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
Shear Wave
Longitudinal Wave
0
Ground Level (m)
10
20
30
Fig. 8. Installation of waterproof layer.
40 0
500
1000 1500 2000 Wave Velocity (m/s)
2500
Fig. 6. Results of seismic velocity (Vp and Vs ) measurement in native ground.
be removed in order to ensure a liquefiable sand deposit. Shear-wave velocity of the sand was measured in situ by an active source and imbedded sensors. This varied from 70 m/s immediately after construction to 93 m/s after the final test. These values are at the low end of the normal range of seismic velocities for sand, and are therefore susceptible to liquefaction. Fig. 8 shows the installation of the waterproof layer. Fig. 9 shows the completed pile-supported structure and the test pit. The water level was kept at 10 cm above the sand surface throughout seismic tests to prevent dry out of the sand deposit.
100 Weight Passed (%)
Mean Grain Size D =0.3mm 50
80
Uniformity Coef. U =3.68 C Fine Contents Fc=10%
60 40 20
Test results of three samples are presented.
rain Size (mm)
0 10
-3
10
-2
-1
10 10 Grain Size (mm)
0
Fig. 7. Grain size distribution of backfill sand.
10
1
Fig. 9. Completion test structure and test pit.
4. Seismic test results Six seismic tests were conducted on the pilesupported structure. The locations of the blast areas for each test are shown in Fig. 10. The blast areas were about 60 m wide and their lengths depended on the mining schedule. The results of the seismic tests are summarized in Table 1. The maximum horizontal acceleration recorded on the adjacent ground surface varied from 20 Gals to 1353 Gals depending on distance from the blast area to the test site. The closest blast was only 90 m from the test site. These differences in the maximum acceleration yielded responses at different levels and liquefaction of different degrees. Sand boiling phenomena were observed in the test
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
371
Table 1 Summary of seismic tests Level of input motions
Test number
Distance (m)a
Maximum acceleration (gal)b EW
a b
NS
Vertical
Distance from blast area to test site. At the ground level of adjacent free field.
4.1. Measurement records at free field and sand deposit The maximum accelerations recorded in the adjacent free field in vertical array configurations are compared for three tests in Fig. 11. The amplification tendencies from GL-32m (GL-32 m = 32 m below Ground Level) to the surface were similar in the mudstone layers for three tests. The maximum accelerations recorded through the mudstone layers to the sand deposit are compared for three tests in Fig. 12. There was a clear difference among the amplification trends in the test pit. Test-1 showed a similar amplification trend to that of the mudstone layers as shown in Fig. 11. Test-5 showed less amplification in the sand deposit. Test-3 showed a large decrease in acceleration in the test pit because of severe liquefaction of the sand deposit. Fig. 10. Locations of test site and blasts in seismic tests (north is up).
0 Ground Level (m)
-5
pit with larger input motions. This is one of the most advantageous features of the test method employed in this project, although the input motions were not controlled mechanically or electrically. In this paper, three tests indicated by highlights in Table 1 were chosen for detailed investigations, because those tests provided three different phenomena in terms of liquefaction of the sand deposit as well as in terms of dynamic responses of the structure. Results from these three tests are detailed in the next section.
-10 -15 -20 -25 -30 Test-1
Test-5
Test-3
-35 0
10
100 200 0 20 30 0 Acceleration (Gal)
400
800
Fig. 11. Maximum acceleration in free-field vertical array.
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
0
100
-2
0 Acceleration (Gal)
Ground Level (m)
372
-4 -6 -8 Test-1
Test-5
Test-3
-10 0
40 0 50 100 150 0 200 400 600 Acceleration (Gal)
20
Sand Surface
Max=84.2Gal
-100 200
Free Field Surface
0 Max=142Gal
-200 100
GL-32m
0 Max=52.0Gal
-100 Fig. 12. Maximum acceleration in test pit (0–3 m) and mudstone layers.
0
2
4 Time (s)
6
8
Acceleration (Gal)
600
Acceleration records at the sand surface, the free field surface and GL-32m are compared for Test-1 (small input level) in Fig. 13. The response spectra from these records are also shown in the figure. The same set of acceleration records and these response spectra are shown in Fig. 14 for Test-5 (medium input level) and in Fig. 15 for Test-3 (large input level).
Sand Surface Free Field GL-32m h=0.05
400 200 0 0.02
0.10
1.00
5.0
Period (s) 50
Fig. 14. Acceleration records of Test-5 (medium input level).
Sand Surface
Acceleration (Gal)
0 Max=27.8Gal
-50 30
Free Field Surface
0 Max=20.0Gal
-30 10
GL-32m
0 Max=6.9Gal
-10 0
2
4 Time (s)
6
8
Acceleration (Gal)
150 Sand Surface Free Field GL-32m h=0.05
100 50 0 0.02
As can be seen from Fig. 13 for Test-1, over all the frequency regions, the responses at the sand surface were greater than those at the free field surface, and the responses at the free field surface were greater than those at GL-32m. From Fig. 14 for Test-5, the responses at the sand surface and the free field surface were greater than those at GL-32m over all frequency regions. The responses at the sand surface became smaller than those at the free field surface for the period less than 0.4 s due to in a certain degree of liquefaction of the sand. From Fig. 15 for Test-3, the responses at the sand surface became much smaller than those at the free field surface and even smaller than those at GL-32m. These responses reductions in the test pit were caused by extensive liquefaction over the test pit, because shear waves could not travel in the liquefied sand. 4.2. Measurement results for the structure
0.10
Period (s)
1.00
5.0
Fig. 13. Acceleration records of Test-1 (small input level).
Fig. 16 compares the maximum recorded accelerations of the pile-supported structure and in the test
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
100
0
Sand Surface
Ground Level (m)
Acceleration (Gal)
0 Max=105Gal
-100 700
Free Field Surface
0 Max=579Gal
-700 400
Acceleration (Gal)
Test-5
Test-3
-3 2
40 5 10 15 0 10 Bending Moment (kN.m)
20
30
Max=272Gal 0
2
4 Time (s)
6
8
Sand Surface Free Field GL-32m h=0.05
2000
1000
0 0.02
0.10
1.00
5.0
Period (s) Fig. 15. Acceleration records of Test-3 (large input level).
pit for three tests. As can be seen for Test-3, there were differences in maximum response between the pile-supported structure and the sand deposit, which means that the pile and surrounding sand did not behave in the same manners. In Test-3, unlike in the other cases, the maximum acceleration decreased as motions traveled upward. The maximum bending mo-
3 Structure
Ground Level (m)
-2
0
-400
Structure
Structure
Sand
0 Pile
Pile Sand
Pile
Sand
Test-1 0
-1
Test-1 GL-32m
0
-3
373
50
Test-5
Test-3
100 0 50 100 150 0 200 400 600 Acceleration (Gal)
Fig. 16. Maximum acceleration of sand fill and test structure.
Fig. 17. Maximum pile bending moment distribution.
ment distributions of the pile are shown in Fig. 17. These bending moments were calculated from the measured bending strains of the instrumented pile at the seven instrumented levels by first calculating stress from strain and then calculating moment from stress. The bending moment took its maximum value at the pile head for all cases. However, the moment distribution shapes differed and the inflection points of curves moved downward in accordance with the input motion levels, in other words, the degrees of liquefaction in the test pit. Fig. 18 compares the acceleration records at the top slab, the base mat and GL-3m of the pile for Test-1 (small input level) . The response spectra from these records are also shown. The same set of acceleration records and these response spectra are shown in Fig. 19 for Test-5 (medium input level) and in Fig. 20 for Test-3 (large input level). As can be seen from Fig. 18 for Test-1, the maximum accelerations increased as motions went upward. For all frequency regions, the responses at the top slab were greater than those at the base mat, and the responses at the base mat were greater than those at GL-3m of the pile. The first natural period of the soil–pile-structure system was about 0.2 s under the input motion level of Test-1. From Fig. 19 for Test-5, the maximum accelerations increased as motions went upward. For almost all frequency regions, the responses at the top slab were greater than those at GL-3m of the pile. The responses at the base mat were greater than those at GL-3m of the pile for periods greater than 0.2 s and were similar to
374
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
100
200
Top Slab
Max=73.9Gal
50
Base Mat
0 -50
Max=42.3Gal
20
GL-3m (Pile)
0
Max=118Gal
-200 100
Base Mat
0 Max=83.2Gal
-100 100
GL-3m (Pile)
0 Max=13.0Gal
-20 0
2
4 Time (s)
6
0
2
4 Time (s)
6
8
600
200
0.10
1.00
5.0
Period (s) Fig. 18. Acceleration records at test structure (Test-1: small input level).
those at GL-3m for periods less than 0.2 s. The peaks in response spectra were at about 0.3 s, and were shifted by 0.1 s from those of Test-1. For Test-3, the maximum accelerations decreased as motions went upward, which were different from those of the other two tests. The responses at the top slab and the base mat became smaller than or similar to the responses at GL-3m of the pile. Compared with the other two tests, it became difficult to identify peaks corresponding to natural periods of the soil–pile-structure system from response spectra diagrams. The structure responses from the three tests described above indicate that soil nonlinearity and liquefaction greatly influence the dynamic properties of pile-supported structures. 4.3. Excess pore water pressure ratio The dynamic changes of excess pore water pressure ratios are shown in Fig. 21. The excess pore
Acceleration (Gal)
Top Slab Base Mat GL-3m(Pile) h=0.05
400
0 0.02
Max=64.9Gal
-100 8
600 Acceleration (Gal)
Top Slab
0
-100
Acceleration (Gal)
Acceleration (Gal)
0
Top Slab Base Mat GL-3m(Pile) h=0.05
400 200 0 0.02
0.10
1.00
5.0
Period (s) Fig. 19. Acceleration records at test structure (Test-5: medium input level).
water pressure ratio is the ratio of excess pore water pressure to initial effective stress, calculated from the laboratory-measured properties of the sand and the depth of each pore pressure sensor. A ratio of one indicates complete liquefaction, while a ratio less than one indicates the initiation of liquefaction and/or partial liquefaction. High-frequency components of the pore pressure ratio that exceed unity arise from large vertical ground accelerations that dynamically weight/unweight the sand mass above each pore pressure sensor. In Test-1, the maximum ratio stayed around zero, which means that no liquefaction took place. In Test-5, the ratios rose rapidly, reaching around one at GL-0.6m and GL-1.4m after the main vibration was finished. Ratios at GL-2.2m and GL-3.0m were about 0.7 and 0.5. The measurement showed that the liquefaction region was in the upper half of the test pit. In Test-3, ratios at all levels rose rapidly, reaching around one, which indicates extensive liquefaction over the entire region.
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
400
Test-1
Top Slab
375
Test-5
Test-3
2 GL -0.6m
0 -400
Max=277Gal
400
Base Mat
0 Max=302Gal
-400 600
GL-3m (Pile)
0 Max=522Gal
-600
Acceleration (Gal)
0
2
4 Time (s)
2000
6
8
Top Slab Base Mat GL-3m(Pile) h=0.05
1000
Excess Pore Water Pressure Ratio
Acceleration (Gal)
1 0 2
GL -1.4m
1 0 -1 2
GL -2.2m
1 0 -1 2 GL -3.0m 1 0 -1 0
2
4
6
Time (sec)
0 0.02
0.10
Period (s)
1.00
5.0
Fig. 20. Acceleration records at test structure (Test-3: large input level).
4.4. Evaluation of natural frequencies and damping factors of soil–pile-structure system Using measured ground motion inputs and structure responses, vibration properties of the soil–pilestructure system were identified by fitting test records to the single-input-multi-output ARX model (the auto-regressive model) (Ikeda, 2000). The soil–pilestructure system was modeled by a simple lumped mass model, as shown in Fig. 22. Acceleration records at the test pit bottom were used as input motions, and acceleration records of the base mat and the top slab were regarded as output motions in this investigation. The dynamic variation of natural frequencies and damping factors evaluated with time are shown for the first natural mode of the system in Fig. 23. The frequencies of the first mode decreased by ground motions and these values were varied in accordance with input motion levels. The first natural frequencies were around 7 Hz for all tests before subject to the ground motions. For Test-3, the frequency was decreased by half after the ground motions.
Fig. 21. Excess pore water pressure ratio at different depths in sand, measured by pore pressure sensors at four levels.
The effective damping factors of the first mode (as estimated by the auto-regressive model) increased during blast ground motions. The damping also varied with the input motion levels of the three tests. For Test-1, the effective damping factor was less than 7%. For Test-3, the effective damping factor was considerably higher at 20–40%. During extensive liquefaction in the test pit, the natural frequency of the soil–pile-structure system Motions at Top Slab
Output (Top Slab)
Output Motions at Base Mat (Base Mat)
Motions at GL-3m
Input (Motions at GL-3m)
Fig. 22. Model for identifying vibration properties of soil–pilestructure system.
376
N. Kamijo et al. / Nuclear Engineering and Design 228 (2004) 367–376
Frequency (Hz) Damping (%)
Test-1 8
40
Test-5
Test-3
4 0
20 0 0
10
20 Time (s)
30
40
Fig. 23. Evaluated natural frequencies and damping.
decreased because the sand lost its stiffness, however, the effective damping factor was increased considerably because of the nonlinear behavior of liquefied sand.
5. Conclusions (1) Seismic tests were conducted of a pile-supported structure in a liquefiable sand deposit six times using large ground motions from large mining blasts. Liquefaction occurred in the sand to varying levels up to full liquefaction. Nonlinear responses of the soil–pile-structure system at different levels were obtained for various levels of liquefaction in the test pit. (2) The maximum horizontal acceleration recorded at the adjacent ground surface varied from 20 Gals to 1353 Gals depending on distance from the test site to the blast areas. This is one of the advantages of the vibration test method employed in this project. (3) In the adjacent free field, motions were amplified from GL-32m to the ground surface, regardless of input motion levels. In the test pit, amplification or reduction of motions depended upon input motion levels. Motions were amplified at small input levels and decreased at large input levels from the bottom to the surface of the test pit. (4) Generation of pore water pressures depended upon input motion levels. Liquefaction started from the shallow part and extended to the deeper part of
the test pit, and finally the test pit was completely liquefied as in Test-3. (5) Bending moments were a maximum at the pile heads, regardless of input motion levels. However, the moment distribution shapes varied and the inflection points of the distribution curves moved downward in accordance with input motion levels, in other words, the degrees of the liquefaction in the test pit. (6) The vibration test method employed in this experimental research was found very useful and effective for investigating the dynamic behavior of large model structure under severe ground motions. Acknowledgements We would like to express our deep appreciation to Dr. Aoyama, Professor Emeritus of University of Tokyo, for his guidance throughout this experimental research. We would also like to thank Dr. Miura, Professor of University of Hiroshima, for his review of the experimental results and the analysis results. The project staff from OYO International, Agbabian Associates, GeoVision, and CE&MT are also thanked for their dedicated assistance with the field studies. Most importantly, we would like to thank the management and staff of Arch Coal’s Black Thunder Mine for their assistance throughout the project. Special thanks go to Paul Lang, to Gordon Shinkle, to Steve Beil and to the late Bob Martin. They were always very generous and provided us with support in many ways. Without their help, this work would not have been possible. References Kontani, O., Tanaka, H., Ishida, T., Miyamoto, Y., Koyamada, K., 2001. Simulation analyses of seismic tests of a pile-supported structure in liquefiable sand using large-scale blast excitation. In: Proceeding of the 16th International Conference on Structural Mechanics in Reactor Technology (SMiRT-16), Washington. Hushmand, B., Scott, R.F., 1992. In-place calibration of USGS pore pressure transducers at Wildlife Liquefaction Site, California, USA. In: Proceeding of 10th World Conference on Earthquake Engineering, pp. 1749–1754. Ikeda, Y., 2000. System Identification of 10-story building and control effect of installed AMDs based on earthquake observation records. J. Struct. Eng., AIJ 46B (2000) 335–344 (in Japanese).