Consideraciones finales

ultrasonic transducer, than for example other broadband pulse - spike pulse. It has also an advantage of being short, comparing to sweeps and bursts, which is essential for fast testing of dynamic systems.

 A square wave of short rise/fall times and of high amount of points will produce high quality of spectrum.

 Signal acquisition needs to be planned, in order to process a noisy signal using averaging techniques. Running average and ensemble average techniques are useful for applying on ultrasonic signals for noise removal.  Matlab simulation has illustrated possible mechanism of ultrasonic testing of

a medium using a square wave. The mechanism is related to detection of extra frequencies produced by medium, on the background of the incident square wave frequencies. However, as practical testing has shown such mechanism does not apply, because the amount of incoming and outcoming frequencies is constant and only peak amplitudes at fixed wave frequencies are essential for testing.

 Post-processing of ultrasonic data requires writing Matlab programmes of an advanced level, which include memory clearing commands, variable matrices, stop conditions and calculation loops.

2.3 Feasibility testing of broadband ultrasonic

spectroscopy

2.3.1 Challenges

 Find sensitivity of the ultrasonic broadband time-domain spectroscopy method for measurement of liquids of different physical properties;

 Find sensitivity of the ultrasonic broadband time-domain spectroscopy method for measurement of particle size;

2.3.2 Feasibility test for different liquids at static conditions

According to ISO - Ultrasonic Attenuation Spectroscopy (2006), the sound attenuation spectroscopy can be used for characterizing properties of fluids of

different physical properties, as well as for dispersed particles of various sizes. In order to verify this statement, a few liquid samples of different physical properties (glycerol, silicon oil, NaCl brine and water) were tested using the sound attenuation spectroscopy. A square pulse was used as an incident ultrasonic signal. Measurements were performed in TT mode, using two 5 MHz transducers. Sampling rate for acquisition of data points was 200 MS/s. Liquid samples were measured at static conditions in plastic square box of side 10 cm. The square wave was designed in Visual Basic, ASCII Arb mode and generated by Agilent Signal Generator. Emitted wave travelled through the box and liquid and was received on the other side of the box. Data was acquired in LabView program. Ultrasonic results are presented in Figs. 22-23 and physical properties of liquids are listed in Table 6. The original reference signal acquired directly from the signal generator is presented for comparison in Figs. 22a and 23a.

Table 6 Physical properties of liquids.

Water Glycerol Silicon oil Density, kg/m3 970 1230 930 Viscosity, Pa·s 0.001 1.5 0.01055

Fig. 22 a) Reference 1MHz PRF square signal acquired from a signal generator with its spectrum. Measured signal and spectrum obtained for b) glycerol, c) NaCl brine.

Fig. 23 a) Reference 1MHz PRF square signal acquired from a signal generator with its spectrum. Measured signal and spectrum obtained for b) water, c) silicon oil.

Based on the tests, the following observations have been made:

 Amplitudes of harmonic frequencies were attenuated by different rates in tested media. This can also be seen as different signal shapes of time-domain signals acquired for individual liquids.

 Liquids of high viscosity (glycerol, silicon oil) have dampened ultrasonic signal in a greater degree, than of low viscosity (water), see Figs. 8-9 and Table 6.

 Amount of input wave frequencies (fundamental frequency of 1 MHz and its harmonics) has remained unchanged in output signals.

 Acquired reference square wave contains in addition to odd frequencies also even ones of smaller amplitudes. This is a practical phenomenon, in case of a measured square wave.

 Results are specific for the plastic box used in experiment.

Literature gives a lot of examples of measured attenuation of sound via frequency spectrum, for media of different properties. For example, Dukhin, Goetz

and Travers (2005) measured dairy products, which resulted in greater sound

attenuation in high-fat milk, than in a low-fat one. In principle, this example confirms the obtained results for low and high viscosity liquids. Assuming that sound loss is only due to viscosity (and other losses are negligible), which is generally true for liquids (Dukhin and Goetz, 2002), sound losses in water, glycerol and silicon oil have been calculated in Table 7using Stokes equation (see Eq. 2.15 in section 2.6.5). The equation takes into consideration liquid viscosity and density, sound frequency and sound speed parameters. Calculated values indicate that viscous losses are the largest in glycerol and the smallest in water (independently on frequency) and this confirms the obtained results.

Literature does not give any examples of studying broadband pulse shapes in time domain. However, this feature of signal seems to be quite valuable, since it is like a “signature” or a pattern, which is specific for individual liquid.

Table 7 Calculated viscous sound attenuation in water, glycerol and silicon oil using Stokes equation.

Liquid Viscous sound attenuation, αS, np/cm

At 10 kHz At 10 MHz

Water 7.8e-9 7.8e-3

Glycerol 4.4e-6 4.4e0

Silicon oil 1.2e-7 1.2e-1

2.3.3 Feasibility test for different particle sizes in liquid-

particle flow

In the next part of feasibility study, glass particles of different sizes were tested: 20-40 µm and 106 µm. They were suspended in water by mixing, using two types of mixers: a magnetic stirrer, rotating at speed 0.33 m/s (for 106 µm particles) and a high-shear mixer rotating at speed 1.56 m/s (for 20-40 µm particles). Data corresponding to 2 %wt particle mass fraction are plotted in Fig. 24. A reference signal

was acquired for case of flow of pure water, in order to calculate a relative change of signal acquired for particles. This means that water amplitude is 100%rel, and

amplitude for particles is <100%rel for frequency range < 10 MHz (within the useful

sound absorption region).

Detailed analysis of sound spectra for particles is preformed in Paper II, where various effects influencing spectra, like for example particle size, are discussed. An acoustic theory that relates particle size with sound loss is discussed in section 2.6.4 (see the subsections: the critical frequency, particle interactions according to the Viscous and Thermal Loss theories).

Generally, particle size influences the critical frequency at which maximum sound loss occurs (~8.9 MHz for 20-40 µm particles and ~6.3 MHz for 106 µm particles). This is due to the fact that small particles attenuate mostly high sound frequencies (low frequencies passes through them without significant energy loss), which relates to their size, whereas big particles attenuate mostly low frequencies (Dukhin and Goetz, 1996). Particle size is related to the sound critical frequency, however, only for the case, where particles interactions do not occur (for a dilute

system). In case of more concentrated systems, high particle volume fraction

influences the critical frequency and it cannot be related directly to particle size, but to the inter-particle distance, which is smaller than the particle size. That is the case in the performed tests, as calculated particle sizes (from Eq. 2.13 in section 2.6.4)are: 0.19 and 0.22 µm for 20-40 and 106 µm particles, respectively. The inter-particle distance is not only influenced by high particle fraction, but also by flow effects (this issue is discussed in Paper II).

Fig. 24 Attenuation spectra for 20-40 µm and 106 µm glass particles in water, at mixer speeds 1.56 and 0.33 m/s.

2.3.4 Conclusions

 The improved pulsed broadband sound attenuation spectroscopy method is suitable for testing liquids of different physical properties, through study of time signals and their frequency spectra.

 Test has shown that liquids of high viscosity attenuate ultrasonic signal in higher degree than of low viscosity. This has been confirmed by many examples available in literature.

 The output ultrasonic signal contains the same amount of frequencies as the input signal. This is in contradiction with the simulation results presented in section 2.2.9. However, the effect of medium response on the background of the reference square pulse spectrum is clearly distinguishable, through attenuation of frequency peaks.

 The improved pulsed broadband sound attenuation spectroscopy method is sensitive to particle size.

 In order to measure particle size of liquid-particle flow using sound attenuation spectroscopy, an advanced signal interpretation is needed.