Advanced Dynamic Soil Testing – Introducing the New Marum Dynamic

advertisement
Advanced Dynamic Soil Testing –
Introducing the New Marum Dynamic
Triaxial Testing Device
S. Kreiter, T. Moerz, M. Strasser, M. Lange, W. Schunn,
B.F. Schlue, D. Otto, and A. Kopf
Abstract Soil mechanical and submarine mass-movement initiation studies often
use static and quasi-static approaches to determine the strength of soils against
external mechanical stresses. However, many natural processes pose time variant stresses on soils, and hence exert key roles for submarine slope stability and
submarine mass-movement initiation. Prominent examples are earthquake-, wind-,
wave- and current-forces and alternating man-made loading on offshore constructions. Most soils show a weaker response to periodic loading – making dynamic
and cyclic loading experiments mandatory for offshore natural hazard and risk
assessment. Dynamic and cyclic triaxial testing are essential in liquefaction studies
of granular soils and creep investigations of cohesive and granular sediments. So
far, competing setups are used with mechanical spindles, pneumatic actuators or
full hydraulic drives.
The new MARUM dynamic triaxial testing device (DTTD) unit is addressing this
increasing demand by enabling a wide range of test configurations. At its core it
contains an ultra fast, hydraulically-driven ±20 kN cylinder and a 5 kHz real-time
controller. This enables up to ±0.5 mm strokes at up to 50 Hz. Advantages to commercial systems are (1) the high flexibility in test setup, (2) the possibility to feed
arbitrary signals derived from in situ measurements, and (3) full system access to
all controls to expand and adjust the system abilities on the hard and software level.
Applications so far include cyclic creep studies for offshore wind farms, liquefaction experiments on artificial sand–clay mixtures and studies on the behavior of
submerged soils under dynamic stress conditions to evaluate slope stability and
submarine landslide initiation.
S. Kreiter, T. Moerz (), M. Strasser, M. Lange, W. Schunn, D. Otto, and A. Kopf
MARUM – Center for Marine Environmental Sciences and Faculty of Geosciences,
University Bremen, Leobener Str. 28334 Bremen, Germany
e-mail: tmoerz@uni-bremen.de
B.F. Schlue
ARCADIS Consult GmbH, Europaplatz, 3, 64293 Darmstadt, Germany
D.C. Mosher et al. (eds.), Submarine Mass Movements and Their Consequences,
Advances in Natural and Technological Hazards Research, Vol 28,
© Springer Science + Business Media B.V. 2010
31
32
1
S. Kreiter et al.
Introduction
The response of continental margins to time variant stresses has become an important
element in offshore risk assessment. Dynamic and cyclic loading exerts a key role
in submarine slope stability and tsunamigenic submarine landslide initiation
(Sultan et al. 2004; Biscontin et al. 2004). Cyclic loading is important for long-term
securing of offshore installation such as wind energy plants (Lesny and Hinz 2006;
Lesny and Wiemann 2005). The time variant stresses from seismic or storm loading
may either affect the natural system directly or be transmitted through manmade
constructions. In the last few years, large research efforts have been undertaken to
understand the influence of rapid and transient stress changes on slope stability in
the marine realm (Sultan et al. 2004, 2008; Biscontin et al. 2004, Biscontin and
Pestana 2006; Azizian and Popescu 2006, Stegmann et al. 2007; and many others).
Under cyclic loading, the sediment behavior is influenced by the intensity and duration of the cyclic stress. The soil properties relevant for the cyclic and dynamic
behavior are the stress history, density or consistency respectively, sensitivity (for
cohesive soils), degree of saturation, grain size distribution, and fine content
(Kramer 1996). Cyclic stresses may lead to plastic deformation and to degradation
in the stress–strain and strength properties of the sediment, which may result in the
initiation of sediment mobilization. Additionally sands may show total liquefaction
(Ishihara 1985; Pestana et al. 2000).
To simulate and study the wide variety of cyclic and episodic natural processes
in the laboratory a testing system with maximum flexibility is required. Various
dynamic shear systems have been developed over the previous decades. The majority
of them are limited in their performance with regard to the frequency of the load
applied. During recent years, results of undrained monotonic and cyclic ring shear
tests on saturated cohesionless soils have been reported in the literature by Trandafir
and Sassa (2004, 2005a, b), which are limited to 5 Hz. Research cyclic triaxial test
systems at e.g. the Kyoto University or at the University of Duisburg-Essen work
with pneumatic actuators or a combination of servo motor and pneumatic actuaors
to apply the time variant loads up to approximately 1 Hz. The latest generation of
commercially available dynamic triaxial systems are driven by electro-hydraulic
servo actuators at 0.5 kHz, limiting their maximum frequency to 10 Hz or less (see
www.geocomp.com and www.oce.uri.edu). In order to overcome these limitations,
to avoid the disadvantage of restricted modification access using commercial systems, and to address sediment deformation with earthquake loads, which frequency
spectra contain relevant energies up to 20 Hz, the MARUM dynamic triaxial testing
device (DTTD) project was initiated.
2
MARUM Dynamic Triaxial Testing Device
The MARUM DTTD project was started in 2005 when the German Science foundation granted the second phase of the DFG Research Center MARUM (http://
www.marum.de). The system was developed in house from components with the
Advanced Dynamic Soil Testing
33
initial guidance of the industry partner Hansa-Flex Hydraulik GmbH – a company
specialized in hydraulic system design and setup (http://www.hansa-flex.com).
The test equipment compasses a servo-driven hydraulic cylinder with a free
configurable real time controller, a hydraulic power unit, displacement-, load- and
pore pressure transducers, a load frame, a pneumatically controlled confining and
backpressure unit, and a control station with a user interface to initiate the testing
scenarios (Fig. 1). At the heart of the system is a servofloat® quality hydraulic testing
cylinder built by Herbert–Hähnchen GmbH & Co. KG. This cylinder contains a
patented annular gap seal and is characterized by stick-slip-free cylinder movements and ultra low-friction at all piston speeds. The test cylinder operates at up to
32 MPa (25 kN) and the maximum piston speed is 4 m/s. The test cylinder is physically controlled by a 10 l/min D765 series servo MOOG valve directly mounted on
to the cylinder via a voltage interface. The system performance is enhanced by two
1 l accumulator tanks (orange bulbs on the photo of Fig. 1) that prevent insufficient
flux at the feed and the return side of the MOOG valve.
The oil cooling unit of the (11 kW) hydraulic power unit proved to be valuable
in providing a constant oil viscosity and cylinder response. The unit delivers
28 MPa with a maximum flux of 15 l/min, allowing maximum oil flux at full servo
valve opening which leads to frequencies up to 50 Hz and distance amplitude in
access of ±0.5 mm.
The confining- and backpressures are provided via three independent servocontrolled valves (AirCom GmbH) of the PQ type with a pressure range from 0 to 2 MPa.
Fig. 1 Photograph and schematic drawing of the MARUM dynamic triaxial testing device
34
S. Kreiter et al.
The valves are equipped with separate pressure transducers that report the pressure
back via a 0–10 V interface. The heavy load frame is built around two parallel guiding rods that allow height adjustment for different experimental cells of up to 40 cm
width. The load frame is mounted to massive concrete walls at the basement of the
MARUM building, internal vibrations are suppressed via rubber shock-absorbers.
The real time data acquisition, processing and control unit (ADwin-light-16)
comprises a SHARC™ digital signal processor from Analog Devices with a 40 MHz
32-bit FPU. Load and distance information is analog to digital converted and digitally processed by low phase digital filtering send to the control PC and to the PID
(proportional, integral and differential) servo algorithm (Fig. 2). The resulting control signal is connected to the MOOG valve via an analog output of the ADwin
system. All calculations are repeated every 0.2 ms resulting in a 5 kHz feedback
loop and two orders of magnitude oversampling at maximum frequency of 50 Hz.
The distance information is currently provided via (1) an internal inductive displacement sensor (WLH 250-K-0.5%, MESSOTRON) within the hydraulic cylinder
and (2) an external laser optical triangulation system from MICRO-EPSILON.
Force Controlled
3.3
4
3
3.1
Load
Load
3.2
3
2.9
Dynamic
Load
Adaptation
2
1
2.8
0
2.7
0
0.1
0.2 0.3
Time
0.4
0.5
0
0.1
Desired Output Response
0.3
0.4
+
PID
Controller
Digital Analog
Converter
Analog
Digital
Converter
Force Sensor
Process
–
15
4
10
3.5
5
Load
Load
0.2
Time
3
0
–5
2.5
–10
2
–15
0
0.1
0.2
Time
0.3
0.4
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Time
Feedback
Displacement
3.3
Displacement Controlled
3.2
3.1
3
2.9
2.8
2.7
0
0.1
0.2
0.3
Time
0.4
0.5
Desired Output Response
PID
Controller
Digital
Analog
Converter
Analog
Digital
Converter
Displacement
Sensor
Process
–
4
Displacement
+
3
2
1
0
0
0.1
0.2
Time
0.3
0.4
Feedback
Fig. 2 Flowcharts of the force and displacement control processes in the real-time 5 kHz circuit
Advanced Dynamic Soil Testing
35
The internal inductive displacement control proved to be too temperature-sensitive
and its resolution did not allow us to quantify sample deformations with the desired
precision. However, the inductive sensor output is still used for safety reasons
where distance control over the full stroke length is required. The external class II
laser sensor ILD 1700 (Fig. 1) with a measuring range of 100 mm has a dynamic
resolution of 0.003 mm (Fig. 3) at 2.5 kHz sampling frequency – accurate and fast
enough to track the strain of stiff materials under dynamic loads of up to 50 Hz.
Load information is acquired via high-precision force transducers from GTM
(Gassmann Testing and Metrology GmbH) with a range of 20 kN and 0.002% full
scale accuracy. The GTM load transducers are directly mounted to the hydraulic
cylinder rod (Fig. 1). Two ultra-low noise carrier frequency HBM amplifiers are
used to collect the data of the force- and differential pore pressure transducers
(Validyne) prior to analog/digital conversion. The input signal is logged and processed at 5 kHz by the ADwin unit.
Process control and data storage is accomplished with the scenario planning &
control station, a PC-based system running a Labview (National Instruments™)
routine developed in-house. The control station coordinates the real-time ADwin
unit with the back- and side pressure unit. This main user interface allows for the
performance of user-defined scenarios in force-controlled and displacementcontrolled modes (Fig. 2).
3
Performance Examples
To evaluate the actual status (capability, compliance, etc.) of the MARUM DTTD,
various performance tests have been conducted on materials of variable stiffness.
In Figs. 3 and 4, experimental results on the stiffest end member, an aluminum
cylinder placed inside a standard triaxial testing cell, are presented. It can be seen
that at 1 Hz and 60 kPa cyclic loading, the input function and system response are
in excellent agreement. At 10 Hz and 60 kPa dynamic loading, a phase offset and
slight waveform distortion can be observed.
Fig. 3 System response using an aluminum test cylinder at 1 Hz left and at 10 Hz right. Red solid
lines: predefined stress; blue dashed lines: measured stress, green dotted lines: sample deformation
36
S. Kreiter et al.
Fig. 4 Summary of the MARUM dynamic triaxial device response on an aluminum test cylinder
at different frequencies
Frequencies of up to 10 Hz are well reproduced on stiff material (Fig. 4).
However, the phase offset increases linearly for higher frequencies. Phase and
amplitude offsets of the system response at higher frequencies or softer materials are
tackled in an ongoing development effort to automatically modify the PID settings
using a dynamic load adaptation (Fig. 2) This procedure will be especially valuable
in experiments where soil properties are subject to change during complex runs.
The arbitrary signal function offers the possibility for testing soils with selfgenerated or recorded time – load signals to analyze the soil response to complex,
but realistic multi-frequency dynamic loading scenarios under laboratory conditions.
The performance example given here simulates a ground-motion record of the 1989
Loma Prieta earthquake (Mw 7.1) of a stratum 2 m subbottom. The time versus
stress input function was derived using a procedure proposed by Seed and Idriss
(1971) based on the empirical stress reduction factor from Iwasaki et al. (1978):
The shear stress function t(t) on a potential horizontal failure plane at depth z of
a soil column moving horizontally as a rigid body according to a horizontal acceleration function a(t) (i.e. the signal recorded by strong-motion seismometers) can
be calculated using Newton’s laws of motion. In reality the soil column is internally
deformable and the actual shear stress is less than that estimated from the above
described rigid body model. This discrepancy is empirically corrected by the stress
reduction coefficient rd. The shear stress evolution t(t) is therefore defined by
τ (t) =
a(t) ⋅ γ
g⋅z
rd .
(1)
where g is the unit weight and g is the gravitational acceleration. On the basis of a
variety of soil-response data sets, Iwasaki et al. (1978), proposed rd = 1 − 0.015
z/m. This allows for the assessment of shear stress history of a soil at a given depth
from seismometer acceleration recordings.
Figure 5 shows the input arbitrary signal function and the measured system
response measured on a soft foam test specimen (Young’s modulus 235 kPa).
Advanced Dynamic Soil Testing
37
Fig. 5 Load time history: red solid line: arbitrary signal input function derived from acceleration
time history of strong motion record from the Gilroy station #1, Channel 1 – at 090° for the 1989
Loma Prieta earthquake (Mw 7.1; 11 km from the rupture area) at Gilroy (CA), data from the
Centre for Engineering Strong Motion Data (CESMD) http://strongmotioncenter.org. Blue dashed
line actual load measured on the load sensor
The satisfactory reproduction of the input signal confirms that the DTTD system is
able to adequately reproduce the ground acceleration of a given earthquake. Hence,
any other arbitrary load signal derived from storm wave and tidal loading, traffic
load (railways and motorways), foundations (bridges, dams, offshore platforms,
etc.), or data from in situ measurements can be simulated.
4
4.1
Data Examples
Liquefaction of Sand
When initially loose, saturated granular sediment is exposed to strong dynamic or
cyclic stresses (e.g. during an earthquake), the grains may reorganize to take up less
space viz. the sediment may contract. Under undrained conditions the tendency
toward pore water release reduces the effective stress by increasing the excess porewater pressure. High cyclic stress amplitudes on loose granular soils may cause a
drop in the effective stress to zero, which implies that the stress is taken up by the
fluid rather than the particles: the soil is then liquefied (e.g. Ishihara 1985; Ishihara
and Tsukamoto 2004).
Two variables are often used in cyclic failure assessment: the cyclic stress ratio
(CSR) and the cyclic resistance ratio (CRR) (Kramer 1996). The cyclic stress
ratio CSR is defined as sd,cyc/2s′3,i, where sd,cyc is the cyclic deviator stress and s′3,i
is the initial effective minor principal stress at the start of the cyclic loading. Hence,
the CSR is a measure of the cyclic load. The CRR is the CSR for a given number
of cycles at failure. Hence, the CRR is a measure for the capacity of a sediment
38
S. Kreiter et al.
layer to resist liquefaction and stress softening (Sultan et al. 2004). The ratio
between both values is the factor of cyclic safety.
The CRR of natural quartz-rich dune sand has been determined with the new
MARUM DTTD for various configurations. The sand was initially cleaned of
organic material and sieved to a grain size ranging from 0.1 to 0.45 mm. The void
ratio was emax = 0.59 in its densest state and emin = 0.96 in its loosest state. The sand
was prepared dry in the cell and subsequently saturated by 500 ml flow of deaired
water through the sample. Multiple samples were then back-pressured up to
500 kPa until full saturation (B-values > 95%) was reached and consolidated
to the desired initial effective minor principal stress s′3,i. Then the samples were
subjected to cyclic loading by applying load cycles with different CSR values and
frequencies.
Figure 6 presents an example result of such a cyclic triaxial test with dense sand
at a density index ID = 0.87, corresponding to an initial void ratio of e0 = 0.64, a
B-value = 0.96 at 500 kPa backpressure, consolidated s’ 30 kPa, equivalent to an
in-situ depth of ∼5 m. The CSR was set to 0.3 at 1 Hz, which is typical for strong
earthquakes (see Seed and Idriss 1971, and Eq. 1). The normalized pore pressure
(pore pressure/total stress) increases continuously with increasing numbers of
cycles, reaching 1 after ∼17½ cycles (dashed line Fig. 6). Deformation starts after
15¾ cycles, when normalized pore pressure values are around 0.8 (solid line in
Fig. 6). The negative strain reveals a dilatational behavior of the sand during liquefaction. As proposed by Kramer (1996) liquefaction of a soil element is considered
to occur at excess pore pressures equal to 90% of the mean total stress (dotted
vertical line in Fig. 6). Hence, the cyclic resistance of this sand under the given
circumstances is equivalent to 16½ cycles, resulting in a CRR of 0.3 at 16½ cycles.
Fig. 6 Cyclic triaxial test of an artificial sand sample. Top graph: axial strain; bottom graph: pore
pressure evolution. Vertical lines mark different liquefaction definitions: Solid line is onset of
axial deformation; dotted line is pore pressure reaches 90% of total stress; dashed line is pore
pressure reaches 100% of total stress
Advanced Dynamic Soil Testing
39
Fig. 7 Cyclic triaxial test results of samples from a North Sea wind farm area. Cyclic frequency
is 5 Hz, cyclic s1 is stepwise increased every ~10,000 cycles. Blue solid line: axial strain; green
dotted line: CSR. (a) Slightly overconsolidated silty clay, s3,i is 200 kPa. (b) Highly overconsolidated
clay, s3,i is 100 kPa
4.2
Cyclic Creep in Clays
The behavior of cohesive sediments under cyclic load is fundamentally different to
the usual cyclic response of granular media. Commonly stiffness and shear strength
of clayey sediments degrade under cyclic loads. Figure 7a illustrates cyclic stressinduced deformation behavior of slightly overconsolidated silty clay, tested with
every 10,000 cycles increased cycles cyclic stress ratios. At low CSRs the clay
deforms incrementally during the first 1,000 cycles, followed by a shake down in
deformation. At higher cyclic stress levels the deformation does not cease completely, and the clay deforms at lower and lower rates not likely leading to failure
(stabilization). Under even higher CSRs the plastic deformations of clays may
accumulate with ongoing cyclic loading and lead to progressive failure; however
such a behavior has not been observed here for the CSR tested (Fig. 7a). Shake
down, stabilization and progressive failure are the three most common behavior
types of cohesive sediments (Mallikarjuna 1992; Lesny and Richwien 2004).
But cohesive soils show a variety of creep deformation responses. Figure 7b
illustrates a CSR independent creep of strongly overconsolidated plastic clay. Here
the number of stress cycles seems to be more important for the deformation behavior
than the incrementally increased CSR. So far it is unclear whether (1) a critical
consolidation ratio exists, (2) if a certain type of mineralogy is required for CSR
independent creep behavior, and (3) if such a behavior could be described by a
cyclic deformation rate, describing the material deformation by a rate to be
multiplied by the number of stress cycles, not regarding the CSR. These are open
questions currently addressed in an ongoing study.
5
Summary and Conclusion
The increasing awareness of the importance of cyclic and dynamic loading in hazard
initiation stimulated worldwide efforts to build adequate testing devices. In this
paper, the setup of the MARUM DTTD is presented and the crucial components
40
S. Kreiter et al.
and control functions are discussed in the context of commercial systems. The full
dynamic capabilities of servo-valve controlled hydraulic systems, however, can
only be exploited if the force and displacement transducers have excellent accuracy
and dynamics. The programmable digital signal processor of the DTTD allows
flexible system operation modes. A series of performance tests have been performed to benchmark the system’s capabilities. Frequencies up to 10 Hz can be
used with confidence for soil testing. Higher frequencies up to 50 Hz are easily
reached by the DTTD system, but lead to large phase offsets. The linear trend in
acceleration suggests that the hydraulic system is capable of even higher dynamics.
An earthquake equivalent stress signal is reproduced confirming that the MARUM
DTTD is capable of accurately simulating multi-frequent arbitrary signals. Test
carried out using standardized sand samples resulted in high quality strain and pore
pressure data that allow a reliable CRR determination. Tests on highly overconsolidated clays revealed critical cyclic behavior of CSR-independent creep, stimulating
further in depth studies. Cyclic creep, cyclic softening and liquefaction tests are
now routinely conducted at the MARUM in the context of various ongoing projects,
such as (1) the design of offshore wind power plants, and (2) slope stability analysis
and submarine landslides studies (e.g. 1979 Nice Airport Slide, Cretan Sea, Nankai
Trough). However, future work is required to implement the dynamic load adaptation routine that modifies the PID settings according to the changing stiffness of the
samples.
Acknowledgments The authors want to thank the reviewers Kerstin Lesny and Kate Moran for
improving this paper by their helpful comments. This study was funded by the Deutsche
Forschungsgemeinschaft as part of the DFG-Research Center MARUM at the University of
Bremen.
References
Azizian A, Popescu R (2006) Three-dimensional seismic analysis of submarine slopes. Soil Dyn
Earthq Eng 26:870–887
Biscontin G, Pestana JM (2006) Factors affecting seismic response of submarine slopes. Nat Haz
Earth Syst Sci 6:97–107
Biscontin G, Pestana JM, Nadim F (2004) Seismic triggering of submarine slides in soft cohesive
soil deposits. Mar Geol 203:341–354
Ishihara K (1985) Stability of natural deposits during earthquakes. In: Balkema AA (ed.) 11th
International Conference on Soil Mechanics 1:321–376
Ishihara K, Tsukamoto Y (2004) Cyclic strength of imperfectly saturated sands and analysis of
liquefaction. Proc Japan Acad Ser B 80:372–391
Iwasaki T, Tatsuoka F, Tokida K, Ysauda S (1978) A practical method for assessing soil liquefaction potential bas on case studies at various site in Japan. In: Proc. 2nd International Conference
on Microzonation for Safer Construction: Research and Application
Kramer SL (1996) Geotechnical Earthquake Engineering. Prentice-Hall, Upper Saddle River, NJ
Lesny K, Hinz P (2006) A Concept for a Safe and Economic Design of Foundations for Offshore
Windenergy Converters. In: Forkiewicz M (ed.) New Approach to Harbour, Coastal Risk
Management and Education – Littoral, Gdansk
Advanced Dynamic Soil Testing
41
Lesny K, Richwien W (2004) Mindestanforderungen an die Baugrunderkundung. 3. Tagung
Offshore-Windenergie, Germanischer Lloyd, Hamburg
Lesny K, Wiemann J (2005) Design Aspects of Monopiles in German Offshore Wind Farms. In:
Gouvernec S, Cassidy M (eds.) Proceedings of the International Symposium on Frontiers in
Offshore Geotechnics, Perth
Mallikarjuna RK (1992): Behavior of Vertical Piles Subjected to Static and Cyclic Lateral
Loading. Ocean Engineering Centre, Indian Institute of Technology, Madras
Pestana JM, Biscontin G, Nadim F, Andersen K (2000) Modeling cyclic behavior of lightly overconsolidated clays in simple shear. Soil Dyn Earthq Eng 19:501–519
Seed HB, Idriss IM (1971) Simplified procedure for evaluation soil liquefaction potential. Soil
Mech Found Eng SM 9:1249–1273
Stegmann S, Strasser M, Kopf A, Anselmetti FS (2007) Geotechnical in situ characterisation of
subaquatic slopes: The role of pore pressure transients versus frictional strength in landslide
initiation. Geophys Res Lett 34:L07607 1–5
Sultan N, Cochonat P, Canals M, Cattaneo A, Dennielou B, Haflidason H, Laberg JS, Long D,
Mienert J, Trincardi F (2004) Triggering mechanisms of slope instability processes and sediment failures on continental margins: A geotechnical approach. Mar Geol 213:291–321
Sultan N, Cattaneo A, Urgeles R, Lee H, Locat J, Trincardi F, Berne S, Canals M, Lafuerza S
(2008) A geomechanical approach for the genesis of sediment undulations on the Adriatic
shelf. Geochem Geophys Geosyst 9:Q04R03 1–25
Trandafir AC, Sassa K (2004) Undrained cyclic shear response evaluation of sand based on undrained monotonic ring shear tests. Soil Dyn Earthqu Eng 24:781–787
Trandafir AC, Sassa K (2005a) Seismic triggering of catastrophic failures on shear surfaces in
saturated cohesionless soils. Can Geotech J 42:229–251
Trandafir AC, Sassa K (2005b) Evaluation of catastrophic landslide hazard on gentle slopes in
liquefiable soils during earthquakes. In: Proceedings of the International Conference on
Landslide Risk Management, Vancouver (Canada) A.A. Balkema
Download