## 1 Introduction

Introducing the stiffness as a feature of construction soils allows to adopt computational techniques commonly used for materials of elastic properties. Fundamental mechanical parameter of the soil which describes its stiffness, is soil modulus. Research for the mechanical properties of soils conducted over the last 30 years have led to the conclusion that the basic parameter determining the soil stiffness is shear modulus *G* [1]. This parameter is the instantaneous value under the conditions specified by strain level and the stress state, which depends on the load history, its velocity and character (static, dynamic, cyclic), drainage conditions as well as the type of structure that transmits the load. The nonlinear nature of the stiffness variability for the shear modulus *G* can be presented as:

where:

*G _{max}* – initial shear modulus;

*F*(*γ*) – stiffness degradation function;

*γ* – shear strain in soil.

Initial shear modulus *G _{max}* is widely recognized as a parameter of the state of the soil [1]. It can be assumed that the value of the dynamic shear module for very small shear strain is the same as the

*G*modulus for static problems in both “drained” and “undrained” conditions [2]. For this reason, many of the techniques used in dynamic and seismic applications have been adopted for estimating the

_{max}*G*module. In dynamic studies based on the measurement of shear wave propagation velocity in a soild medium, the initial shear modulus is derived from a simple relationship:

_{max}where:

*ρ* – volume density of soil;

*V _{s}* – shear wave velocity.

This dependence is used in many uncomplicated research methods, where the main goal is to measure the time of the propagation of generated elastic wave on the separated area of natural subsoil or soil specimen [3, 4]. Determined in this way modulus is a value corresponding to very small shear strain level.

Over the past 50 years, a number of correlation formulas have been developed to estimate the initial *G _{max}* strain modulus based on the results of the velocity measurement of the transverse mechanical wave in soil and the study of other geotechnical characteristics. For example, Hardin and Richart [5] have developed one of the first correlations for uniformly-grained coarse sand (Ottawa sand, No. 2030,

*d*

_{50}= 0.72 mm,

*C*= 1.2):

_{U}where:

*e* – void ratio,

*p*’ – average effective stress [kPa],

*p _{ref}* – reference pressure (

*p*= 100 kPa).

_{ref}On the basis of further investigations of natural and artificially prepared from sanded soils, Hardin and Black [6] reported a universal relationship:

The equations (3) and (4) represent the two specific forms of a so-called Hardin’s formula.

Based on research carried out worldwide in the field, it has been confirmed that the value of the initial shear modulus can be functionally linked to the void ratio and the average effective stress [7]:

where *A* is a constant dependent on soil type, grain properties and distribution, *f*(*e*) is a function of void ratio *e* reflecting the soil density, *m* is the stress exponent reflecting the confining pressure.

A number of studies conducted in the world on noncohesive soils differing in grading from the *d*_{50} diameter and the coefficient of uniformity *C _{U}* have enabled the development of formulas to determine the value of the

*G*modulus. The formulas shown in Table 1 relate to noncohesive soils with specified grain size distribution characteristics (

_{max}*d*

_{50}and

*C*). Based on the recent research, it has been found that for non-uniformly grained soils, these formulas produce overstated values of

_{U}*G*modulus. Research carried out over the last decade, including Wichtmann and Triantafyllidis [8, 9, 10] at the Karlsruhe Institute of Soil Mechanics and Rock Mechanics, have led to develop a universal formula that will take into account the value of the coefficient of uniformity

_{max}*C*by function:

_{U}Empirical formulas for modulus *G _{max}* of sand and gravel, which were derived from the Hardin’s formula.

Author | d_{50} | C_{U} | A | f(e) | m |
---|---|---|---|---|---|

Hardin B.O. & Richart F.E. (1963) [5] | 0.72 | 1.2 | 69 | 0.50 | |

Lo Presti D.C.F. et al. (1993) [11] | 0.22 | 1.35 | 72 | e^{−1.3} | 0.45 |

0.54 | 1.50 | 71 | 0.43 | ||

0.75 | 4.40 | 71 | e^{−1.3} | 0.62 | |

Modoni G. et al. (1999) [12] | 7.90 | 10 | 76 | 0.50 | |

Fioravante V. (2000) [13] | 0.13 | 1.86 | 101-109 | e^{−0.8} | 0.45-0.52 |

0.55 | 1.66 | 79-90 | 0.43-0.48 | ||

Kuwano R. & Jardine R.J. (2002) [14] | 0.27 | 1.67 | 72-81 | 0.50-0.52 | |

Tika T. et al. (2003) [15] | 0.20 | l.10 | 80 | 0.50 | |

1.70 | 62 | ||||

0.32 | 2.80 | 48 | |||

Hoque E. & Tatsuoka F. (2000, 2004) [16, 17] | 0.16 | 1.46 | 71-87 | 0.41-0.51 | |

0.31 | 1.94 | 80 | 0.47 | ||

0.50 | 1.33 | 61-64 | 0.44-0.53 | ||

0.62 | 1.11 | 82-130 | 0.44-0.53 | ||

1.73 | 1.33 | 53-94 | 0.45-0.51 | ||

Chaudhary S.K. et al. (2004) [18] | 0.19 | 1.56 | 84-104 | 0.50-0.57 | |

Wichtmann T. & Triantafyllidis T. (2004) [8] | 0.55 | 1.80 | 275 | 0.42 |

where appropriate constants were proposed as [9]:

The selected most used formulas are graphically represented in Figure 1 in relation to the results obtained from equations 6-9.

In Figure 2 the relationship between *G _{max}* values and coefficient of uniformity

*C*is given and in Figure 3 between

_{u}*G*and average effective stress

_{max}*p*′, for constant value of the void ratio

*e*= 0.37. The value of the

*G*modulus decreases with the increase of the coefficient of graining uniformity

_{max}*C*and increses with the increase of the average effective stress

_{u}*p*’.

This article presents selected issues related to the continuation of non-cohesive soils survey conducted at the Institute of Soil Mechanics and Rock Mechanics in Karlsruhe. The purpose of the present study is to verify the correlation between the characteristics of grain size distribution of non-cohesive soils and the value of the dynamic shear modulus. For this purpose, elastic wave propagation studies were conducted on specimens with artificially generated grain size distribution, using piezoelectric bender elements (BE) and triaxial compression cell. The purpose of the research was also to verify and evaluate the different methods of interpreting BE test results.

## 2 Bender elements

Currently, the use of bender elements is a commonly used measurement method in geotechnical laboratories, although the idea itself is based on the phenomenon of simple and inverted piezoelectricity discovered in 1880 by Jacques and Pierre Curie. However, first practical attempt was carried out more than 100 years later by Donald Shirley and Loyd Hampton in 1977. Initially it was used to measure shear wave velocity in marine sediments using piezoceramic materials. Since then, numerous upgrades of electronic components such as transducers have been made and in spite of the enlargement of the spectrum of uses, the very idea of research still remains unchanged.

Bender elements are two-layer piezoelectric plates covered with a protective resin (Figure 4). Elements work in pairs. By mounting them to the opposite ends of the cylindrical specimen, the system and components of the triaxial compression apparatus (Figure 5) are most commonly used for this purpose.

Applying electrical voltage of variable amplitude (±10V = 20*V _{pp}*) to one of the plates causes alternating and simultaneous shortening of one of the layers of the element and elongation of the second layer. As a result, this leads to a cyclic bending of the entire element and the generation of a elastic mechanical wave inside the tested specimen. The wave propagating along the soil specimen is a shear wave and accompanying compressive waves that propagate perpendicular to the plane of the bender element. The wave reaching the receiver bends the piezoelectric plate generating the voltage that the oscilloscope records (Figure 6).

Measurement of the wave transit time between the transmitter and the receiver is crucial for determining the velocity of propagation of the wave in the tested medium. Then, it leads to the calculation of the practical parameter which is the value of the *G _{max}* module according to the equation (1.2).

The system controlling device allows selection of the waveform generation method in terms of the shape of a single waveform. It is possible to generate rectangular, trapezoidal, triangular waveforms, although sinusoidal waves are most commonly used. It is also possible to generate a continuous or pulse wave according to the selected range of interpretations of the obtained results.

Currently, different methods are used to determine shear wave propagation time. They have been described in detail in [19, 20]. These methods include:

- methods of interpreting the results of a single-pulse test:
- □visual detection method for the first signal level deviation (first arrival FA, visual picking VP) – depends on determining the time that has elapsed since the beginning wave generation until the first recording of the disturbance in the received signal (
*t*,_{FA}*t*); it is most commonly used in practice and also the simplest method of interpretation of results; its major disadvantage is the possibility of interfering with measurement by the disordering influence of compression waves;_{VP} - □peak signal detection (major peak-to-peak, PP) method – depends on the measurement of the time difference (
*t*) between the time the peak occurs in the transmitted signal and the time when the first peak in the received signal is received; similar to the first deviation method, it does not require any analysis beyond the visualization of two characteristic points on the oscillogram; its disadvantage is the ambiguity of the result in the case of receiving a number of closely spaced peaks with small amplitude differences resulting from, for example, the heterogeneity of the specimen of the soil being tested;_{PP} - □cross power spectrum analysis (cross-power spectrum, CS) determines wave propagation time
*t*:_{cs}this method is based on the angle of inclination of a straight line, which is the linear approximation of the phase angle relation$\begin{array}{c}{t}_{CS}=\frac{\alpha}{2\pi};\end{array}$ *φ*from the frequency*f*for the complexity of the cross-power spectrum*G*of the signals_{XY}*X*(*t*) and*Y*(*t*), determined according to the following formula:where$\begin{array}{c}{G}_{XY}\left(f\right)={L}_{X}^{{}^{\ast}}\left(f\right){L}_{Y}\left(f\right),\end{array}$ *L*(_{X}*f*) and*L*(_{Y}*f*) are linear spectra of the signals*X*(*t*) and*Y*(*t*), determined by Fourier transformation (FFT):and$\begin{array}{c}{L}_{X}\phantom{\rule{0ex}{0ex}}\left(f\right)=FFT\phantom{\rule{0ex}{0ex}}\left(X\right(t\left)\right),\phantom{\rule{0ex}{0ex}}{L}_{Y}\phantom{\rule{0ex}{0ex}}\left(f\right)=FFT\phantom{\rule{0ex}{0ex}}\left(Y\phantom{\rule{0ex}{0ex}}\right(t\left)\right)\end{array}$ *L*^{*}_{X}(*f*) is the conjugate coupling of*L*(_{X}*f*); - □cross correlation method (CC) – is a measure of the degree of matching of a generated and received signal in the frequency domain; determining the time
*t*of the best time-domain matching is the result of the analysis of the cross-correlation coefficients_{cc}*CC*of the signals_{XY}*X*(*t*) and*Y*(*t*), recorded at*t*time; this coefficient defines the following formula:_{r}where$\begin{array}{c}C{C}_{XY}\left({t}_{CC}\right)=\underset{{t}_{r}\to \infty}{lim}\frac{1}{{t}_{r}}\underset{{t}_{r}}{\int}X(t+{t}_{CC})\phantom{\rule{0ex}{0ex}}\phantom{\rule{0ex}{0ex}}Y\left(t\right)\phantom{\rule{0ex}{0ex}}dt,\end{array}$ *t*is the time delay between the analyzed signals, indicated unequivocally by the position of the maximum_{CC}*CC*value; having the cross power spectrum_{XY}*G*(_{XY}*f*) of the signals*X*(*t*) and*Y*(*t*),*CC*(_{XY}*t*) is the inverse Fourier transformation*G*(_{XY}*f*):$\begin{array}{c}C{C}_{XY}\left(t\right)=FF{T}^{-1}\left({G}_{XY}\right(f\left)\right)\end{array}$

- □visual detection method for the first signal level deviation (first arrival FA, visual picking VP) – depends on determining the time that has elapsed since the beginning wave generation until the first recording of the disturbance in the received signal (
- methods of interpreting the results based on continuous wave generation:
- □the
*π*-points, phase frequencies method (*π*-points); this method is based on generating sinusoidal waves of smoothly adjustable frequency (0-30kHz) and recording the selected frequency values for which the transmitted and received signals have this the same or the exact opposite phase (colloquially: the signals are in phase or counter-phase); phase compatibility is adjusted visually by observing the screen of the oscilloscope that should display the signal amplitude values in the mutual function relation (ie the x coordinate is amplitude of the transmitted signal, and y – received; graph in straight line means phase compatibility); chosen frequencies should differ by one step:*kπ*– phase matching should appear at constant frequency intervals (angular). Due to*k*value the number of wavelengths*N*that lie on the*L*distance is determined:_{tt}then the graph of dependency$\begin{array}{c}N=\frac{k}{2};\end{array}$ *N*(*f*) is performed; the sought value of the shear wave propagation time is the value of the slope of the straight line approximating the points*N*(*f*). - □continuous sweep input frequency method; it is very similar to cross-power spectrum analysis, except that the wave generation is based on the
*π*-point method, and the received signal is continuously analyzed for power spectrum compatibility and the linearity of the functional relationship between the phase angle and the change in frequency; applying this technique in real time needs to supplement the measurement system with a power spectrum analyzer.

- □the

Regardless of the method used, there must be taken into account the fact that the input signal is different from the signal generated. It is related to elastic reaction of the material (mainly the resin surrounding the piezoelectric) and the length of the bender element (working as a cantilever).

The response of the bender receiver is dependent on:

- ∘stiffness, damping and dispersion properties of the specimen medium,
- ∘distance between transmitter and receiver BE,
- ∘the length and frequency of the generated waveform,
- ∘interference of component waves,
- ∘effects that disorder the measurement of the shear wave component.

It is crucial to highlight that S-wave measurement can be disordered by P-wave interference. Proximity effect of the transmitter is a serious limitation of the method which appears in the case when the transmitter’s distance to the receiver is too small. As the relation *λ/A* increases (*L*-distance between receiver and transmitter, *λ*-generated wavelength), the latency of the arrival of the P-wave to the bender relative to the S-wave is increased – the effect of the proximity phenomenon of the transmitter – the amplitude due to the P-wave decreases, and the final measurement accurately maps the S-wave. According to [21] it is recommended that *L/λ _{S}* be greater than 4 times the shear wave length.

The other limitations were widely described in [20, 22] and might be connected with wave refraction, reflection, amplitude attenuation, frequency shift and electromagnetic coupling (cross-talk).

## 3 Test material and specimen preparation

The geometrical and physical parameters of tested sands are collected in Table 2. The grain size distributions and microscopic photos of sand specimens are shown in Figure 7-11.

Parameters of tested sands.

Sand | G [-]_{s} | d_{50} [mm] | d_{60} [mm] | d_{10} [mm] | C [-]_{U} | e [-] | ρ [g/cm^{3}] |
---|---|---|---|---|---|---|---|

P1 | 2.65 | 0.33 | 0.35 | 0.22 | 1.6 | 0.50 | 1.76 |

P2 | 2.65 | 0.33 | 0.41 | 0.14 | 3.0 | 0.42 | 1.86 |

P3 | 2.65 | 0.33 | 0.48 | 0.12 | 4.0 | 0.37 | 1.93 |

P4 | 2.65 | 0.60 | 0.65 | 0.32 | 2.0 | 0.50 | 1.76 |

P5 | 2.65 | 0.60 | 0.80 | 0.20 | 4.0 | 0.37 | 1.94 |

Each of the sand specimens P1-P5 was prepared from pre-formed non-cohesive soil fractions, according to the accepted grain size curves (approx. weight of each specimen: 1.5kg). A cylindrical specimen was then formed using a template with rubber installed inside it. Each cylindrical specimen (Figure 12a) was compacted with a hand rammer in five layers of dry sand, great care was taken not to damage the bender element installed on the bottom cap.

The weight and geometrical dimensions of formed triaxial specimens were precisely controlled and meticulously recorded. After acceptance of the quality of prepared specimen, it was placed in the compression chamber and isotropic pressure of 100 kPa was applied (effective stress). If changes in displacements are by LVDT the gauge stopped, bender element tests were performed.

After that, a higher level of isotropic pressure was applied (400 kPa) and procedure was repeated. Then the specimen was unloaded to 100 kPa and it was saturated with CO_{2} using back-pressure system. Then the specimen was saturated with vacuum de-aired water and whole procedure of testing was repeated (first with 100 kPa and then with 400 kPa, every time after full stabilization of displacements).

## 4 Results of the research

The results of all performed tests in form of relations between *G _{max}* predicted by Hardin’s formula and

*G*measured using bender elements are collected in Table 3-4 (for all Hardin’s relationships irrespective of their applicable range) and Figure 13. As it can be easily noticed, satisfactory compliance of the results can be observed for specimens in a state of a isotropic pressure of 100 kPa. The points representing the individual comparison of results obtained for specimens tested at isotropic pressure of 400 kPa do not lie on a straight line that is a set of points with identical coordinates (Figure 13). This means that for higher pressures either Hardin’s rule is not correct, or the results of the research are flawed.

_{max}Moduli *G _{max}* [MPa] for tested sand specimens P1-P5 and

*p*= 100 kPa.

Author | P1 | P2 | P3 | P4 | P5 |
---|---|---|---|---|---|

Hardin B.O. & Richart F.E. (1963) [5] | 128.3 | 148.8 | 163.2 | 128.3 | 163.2 |

Lo Presti D.C.F. et al. (1993) [11] | 177.3 | 222.4 | 262.2 | 177.3 | 262.2 |

148.3 | 171.1 | 187.1 | 148.3 | 187.1 | |

174.8 | 219.3 | 258.6 | 174.8 | 258.6 | |

Modoni G. et al. (1999) [12] | 158.7 | 183.2 | 200.3 | 158.7 | 200.3 |

Fioravante V. (2000) [13] | 175.9–189.8 | 202.2–218.2 | 223.7–241.5 | 175.9–189.8 | 223.7–241.5 |

137.5–156.7 | 158.1–180.2 | 175.0–199.4 | 137.5–156.7 | 175.0–199.4 | |

Kuwano R. & Jardine R.J. (2002) [14] | 133.9–150.6 | 155.3–174.7 | 170.3–191.6 | 133.9–150.6 | 170.3–191.6 |

Tika T. et al. (2003) [15] | 148.7 | 172.5 | 189.2 | 148.7 | 189.2 |

115.3 | 133.7 | 146.6 | 115.3 | 146.6 | |

89.2 | 103.5 | 113.5 | 89.2 | 113.5 | |

Hoque E. & Tatsuoka F. (2000, 2004) [16, 17] | 132.0–161.8 | 153.1–187.6 | 167.9–205.8 | 132.0–161.8 | 167.9–205.8 |

148.7 | 172.5 | 189.2 | 148.7 | 189.2 | |

113.4–119.0 | 131.6–138.0 | 144.3–151.4 | 113.4–119.0 | 144.3–151.4 | |

152.5–241.7 | 176.8–280.4 | 193.9–307.4 | 152.5–241.7 | 193.9–307.4 | |

98.5–174.8 | 114.3–202.7 | 125.3–222.3 | 98.5–174.8 | 125.3–222.3 | |

Chaudhary S.K. et al. (2004) [18] | 156.2–193.4 | 181.2–224.3 | 198.7–246.0 | 156.2–193.4 | 198.7–246.0 |

Wichtmann T. & Triantafyllidis T. (2004) [8] | 169. 0 | 209.5 | 238.5 | 169. 0 | 238.5 |

BE test, peak-to-peak (dry/sat.) | 151.4/151.5 | 166.0/147.1 | 152.6/120.6 | 133.8/139.1 | 147.5/145.8 |

BE test, cross correlation (dry/sat.) | 112.6/116.2 | 106.3/108.5 | 99.6/105.3 | 133.1/106.5 | 146.7/111.6 |

BE test, cross power spectrum (dry/sat.) | 94.6/164.7 | 117.9/100.6 | 130.9/145.2 | 135.3/204.2 | 149.2/213.9 |

BE test, p-points (dry/sat.) | 144.0/188.0 | 142.2/130.1 | 173.0/150.6 | 202.5/149.7 | 223.2/156.8 |

Moduli *G _{max}* [MPa] for tested sand specimens P1-P5 and

*p*= 400kPa.

Author | P1 | P2 | P3 | P4 | P5 |
---|---|---|---|---|---|

Hardin B.O. & Richart F.E. (1963) [5] | 256.6 | 297.6 | 326.4 | 256.6 | 326.4 |

Lo Presti D.C.F. et al. (1993) [11] | 330.8 | 415.0 | 489.3 | 330.8 | 489.3 |

269.2 | 310.6 | 339.6 | 269.2 | 339.6 | |

412.9 | 518.0 | 610.8 | 412.9 | 610.8 | |

Modoni G. et al. (1999) [12] | 317.5 | 366.4 | 400.5 | 317.5 | 400.5 |

Fioravante V. (2000) [13] | 328.1–390.2 | 377.3–448.6 | 417.5–496.5 | 328.1-390.2 | 417.5–496.5 |

249.7–304.8 | 287.0–350.5 | 317.7–387.9 | 249.7–304.8 | 317.7–387.9 | |

Kuwano R. & Jardine R.J. (2002) [14] | 267.7–309.7 | 310.6–359.2 | 340.6–393.9 | 267.7–309.7 | 340.6–393.9 |

Tika T. et al. (2003) [15] | 297.5 | 345.1 | 378.4 | 297.5 | 378.4 |

230.5 | 267.4 | 293.3 | 230.5 | 293.3 | |

178.5 | 207.0 | 227.0 | 178.5 | 227.0 | |

Hoque E. & Tatsuoka F. (2000, 2004) [16, 17] | 233.0–328.0 | 270.3–380.5 | 296.4–417.2 | 233.0–328.0 | 296.4–417.2 |

285.4 | 331.0 | 363.0 | 285.4 | 363.0 | |

208.7–248.1 | 242.1–287.8 | 265.5–315.6 | 208.7–248.1 | 265.5–315.6 | |

280.6–503.9 | 325.5–584.6 | 356.9–641.0 | 280.6–503.9 | 356.9–641.0 | |

183.9–354.4 | 213.3–411.1 | 233.9–450.8 | 183.9–354.4 | 233.9–450.8 | |

Chaudhary S.K. et al. (2004) [18] | 312.4–426.1 | 362.3–494.3 | 397.3–542.0 | 312.4–426.1 | 397.3–542.0 |

Wichtmann T. & Triantafyllidis T. (2004) [8] | 302.4 | 375.0 | 426.9 | 302.4 | 426.9 |

BE test, peak-to-peak (dry/sat.) | 318.6/239.5 | 309.2/329.4 | 327.0/256.8 | 380.4/239.5 | 292.4/250.9 |

BE test, cross correlation (dry/sat.) | 122.2/208.8 | 143.9/192.4 | 329.4/170.8 | 92.5/212.5 | 233.1/222.6 |

BE test, cross power spectrum (dry/sat.) | 291.4/216.6 | 384.1/333.5 | 281.0/192.8 | 303.7/155.4 | 295.4/162.8 |

BE test, p-points (dry/sat.) | 208.9/212.8 | 223.8/275.0 | 238.0/358.2 | 195.2/204.8 | 183.1/214.6 |

Analyzing the results of the research on the basis of the literature, it can be seen that non-cohesive soils with identical grain-size distributions and mineral compositions can significantly differ in mechanical properties due to even slight differences in surface geometry and grain shape. The results of the research indicate the need to include in the formulas on *G _{max}* beyond

*C*and

_{U}*e*yet additional geometrical parameters of grains, eg surface smoothness.

On the other hand, it is very important to mention that interpretation methods and techniques are far from ideal ones. For example, the *π*-points method used in signal analysis has proved to be extremely problematic due to the interference occurring in certain frequency ranges. Significant fluctuations in the amplitude of the received signal often prevented the observation due to the high level of electrical noise generated by triaxial apparatus and control system. The usefulness of the remaining methods can be assessed in a simple way, using artificial idealized received signal with constant amplitude and constant timedelay. The results of relative error analysis of the timedelay for frequency variation in the range of 10-30 kHz and noise with the amplitude of 0-100% of pure received signal amplitude for two selected interpretation methods are presented in Figure 14.

As it can be noticed, a cross power spectrum analysis is very sensitive on noise level (see Figure 14b). Satisfactory results can be obtained for noise level below 12% of pure signal amplitude. It is very difficult to achieve this, but it is possible thanks to the use of multiple repeated measurements with averaging of the obtained results (the application which controls oscilloscope ADC-212 has built-in such a function). However, in the case of crosscorrelation method (Figure14a) dark fields indicating a low error value are quite chaotic. In spite of good performance with low-level noise (below 3%), quite accurate results can also be obtained for full-level noise (see left upper corner of the graph). On the other hand, a low level noise can completely degrade the results for certain frequencies (see bright spots on the graph). As far as the peak method is concerned, it is not susceptible to interference and frequency changes because its effectiveness depends only on the proper indication of the peak in the received signal. Unfortunately, due to disturbances caused by S- and P-waves reflections and interferences, it is often very difficult to correctly identify the right peak – such identification is very often not as obvious as it might seem. That is why it is so important to use multiple interpretation methods with detailed analysis of the reliability of the obtained results.

## 5 Concluding remarks

The results of the laboratory tests presented in this paper confirm the correlation between the grain-size characteristics of non-cohesive soils and the value of the dynamic shear modulus for isotropic pressure of 100 kPa. The equation derived by Hardin and extended by Wichtmann & Triantafyllidis has a high applicability in practical estimations of *G _{max}* treated as a nonlinear function of

*C*and

_{U}*e*soil structure parameters. However, it is worth mentioning that the results of the research in the resonant column and the bender elements can be very different. This inconvenience hampers comparative analysis and requires the simultaneous application of many advanced signal processing techniques. The authors used four interpretation techniques and methods to process collected measurement data: peak-to-peak, cross-correlation, cros power spectrum and

*π*-points. Despite the multifaceted processing of measured data, the results obtained for sand specimens at a pressure of 400 kPa have a significant discrepancy. This shows that obtaining unambiguous results from bender element tests is a difficult task in practical applications and direct comparisons with RC results are often not possible. These comparisons are difficult because of the fundamental difference in the generation of small strain: in RC the deformations associated with the torsion develop from the smallest in the center of the specimen to the largest at its circumference; in the BE tests the situation is exactly the opposite. In addition, the RC induces deformation throughout the whole specimen cross section and in BE – it is a matter of the wave propagation phenomenon, which can not be controlled by a researcher. The difficulties with using BE, which were mentioned above, require further research, both in signal processing and in the technical aspects of BE generation of S-waves. The works last.

## Nomenclature

*CC _{XY}* cross-correlation coefficient, [-]

*C _{U}* coefficient of uniformity, [-]

*d* grain size, [mm]

*d*_{50}, *d*_{60} specific grain size, [mm]

*e* void ratio, [-]

*G* shear modulus, [Pa]

*G _{dyn}* dynamic shear modulus, [Pa]

*G _{max}* initial shear modulus, [Pa]

*G _{s}* specific gravity, [-]

*G _{XY}* cross-power spectrum of the signals

*X*(

*t*) and

*Y*(

*t*), [-]

*L _{X}* linear spectrum of the signal

*X*, [-]

*L _{X}*, [-]

*V _{s}* shear wave velocity, [m/s]

*p*′ average effective stress, [Pa]

*p _{ref}* reference pressure (

*p*= 10

_{ref}^{5}Pa)

*t _{CC}* time delay, [s]

*t _{CS}* wave propagation time, [s]

## Greek letters

*γ* shear strain, [-]

*ρ* volume density of soil, [kg/m^{3}]

## References

- [1]↑
Mayne P.W., Coop M.R., Springman S.M., Huang A-B., Zornberg J.G., Geomaterial Behavior and Testing, Proc. of 17-th International Conference on Soil Mechanics and Geotechnical Engineering, Alexandria, Egypt, 5–9 October 2009, 2777-2872

- [2]↑
Tatsuoka, F., Shibuya, S., Deformation characteristics of soils and rocks from field and laboratory tests, Report of the Institute of Industrial Science, Univ. of Tokyo, Japan, 1991, 37 (1)

- [3]↑
Dyka I., Srokosz P., Badania gruntu w aparacie skrętnego ścinania RC/TS. Część 1, Inżynieria Morska i Geotechnika, 2012, 6, 700707 (in Polish)

- [4]↑
Dyka I., Srokosz P., Badania gruntu w aparacie skrętnego ścinania RC/TS. Część 2, Inżynieria Morska i Geotechnika, 2014, 2, 118129 (in Polish)

- [5]↑
Hardin B.O., Richart F.E., Elastic wave velocities in granular soils, Journal of Soil Mechanics and Foundations Divisions, ASCE, 1963, 89 (SM1), 3365

- [6]↑
Hardin B.O., Black W.L., Closure to vibration modulus of normally consolidated clays, Journal of the Soil Mechanics and Foundations Division, ASCE, 1969, 95 (SM6), 1531-1537

- [7]↑
Benz T., Small strain stiffness of soils and its numerical consequences, PhD thesis, Universitat Sttutgart, 2007

- [8]↑
Wichtmann T., Triantafyllidis T., Influence of a cyclic and dynamic loading history on dynamic properties of dry sand, Part I: cyclic nad dynamic torsional prestraining, Soil Dynamics and Earthquake Engineering, 2004, 24 (2), 127-147

- [9]↑
Wichtmann T., Triantafyllidis T., On the Influence of the grain size distribution curve of quartz sand on the small strain shear modulus Gmax, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 2009, 135, (10), 14041418

- [10]↑
Wichtmann T., Soil behaviour under cyclic loading – experimental observations, constitutive description and applications, Habilitation thesis, Karlsruhe Institute of Technology, Karlsruhe, Baden-Wuerttemberg, Germany, 2016

- [11]↑
Lo Presti D.C.F., Pallara O., Lancellotta R., Armandi M., Maniscalco R., Monotonic and cyclic loading behaviour of two sands at small strains, Geotechnical Testing Journal, 1993, 16 (4), 409-424

- [12]↑
Modoni G., Flora A., Anh Dan L.Q., Mancuso C., Koseki J., Balakrishnaier K., Tatsuoka F., A simple experimental procedure for the complete characterisation of small strain stiffness of gravels, Proc. of the 2-nd International Conference on Pre-failure Characteristics of Geomaterials, edit. Jamiolkowski, Lancelotta, Lo Presti, Torino, A.A. Balkema 1999, 123-130

- [13]↑
Fioravante V., Anisotropy of small-strain stiffness of Ticino and Kernya sands from seismic wave propagation measured in triaxial testing, Soils and Foundations, 2000, 40 (4), 129-142

- [14]↑
Kuwano R., Jardine R.J., On the application of cross-anisotropic elasticity to granular materials at very small strains, Geotechnique, 2002, 52, (10), 727-749

- [15]↑
Tika T., Kallioglou P., Papadopoulou A., Pitilakis K., Shear modulus and damping of natural sands, Proc. of the 3-th International Symposium on Deformation Characteristics of Geomaterials IS Lyon 2003, edit. Di Benedetto at all., A.A. Balkema, 2003, 401-407

- [16]↑
Hoque E., Tatsuoka F., Effects of stress ratio on small-strain stiffness during triaxial shearing, Geotechnique, 2004, 54 (7), 429-439

- [17]↑
Hoque E., Tatsuoka F., Kinematic elasticity of a granular material, Proc. of the Int. Conference on Geotechnical & Geological Engineering GeoEng2000, Melbourne, Australia, 19-24 November 2000

- [18]↑
Chaudhary S.K., Kuvano J., Hayano Y., Measurement of quasielastic stiffness parameters of dense Toyoura sand in hollow cylinder apparatus and triaxial apparatus with bender elements, Geotechnical Testing Journal, 2004, 27(1), 1-13

- [19]↑
Sawangsuriya A., Wave propagation methods for determining stiffness of geomaterials, Wave Processes in Classical and New Solids, edit. P. Giovine, InTech, 2012, 157-200

- [20]↑
Srokosz P., Interpretacja wyników badań gruntu elementami bender, Inżynieria Morska i Geotechnika, 2012, 19 (1) 29-38 (in Polish)

- [21]↑
Viggiani G., Atkinson J.H., Interpretation of bender element tests, Geotechnique, 1995, 45 (1), 149-154

- [22]↑
Sanchez-Salinero, I., Roesset, J. M., and Stokoe, K. H., Analytical studies of body wave propagation and attenuation, Report GR 86-15, University of Texas, Austin, 1986