Monument Future. Siegfried Siegesmund

velocities because the sample length may vary from samples to sample. Tested samples have a cubic shape and the edge length range from 40–50 mm. Moreover, weathered samples may also suffer superficial weight loss, resulting in a decrease in edge length.

       Results and Discussion

      For the studied sandstones, the arrival time of P-wave signals are easily identifiable with a high SNR, although their waveforms and wavelength values are dependent on the type of rock being 194analysed. The frequency of the samples decreases with particle size (Fig. 1). Inversely, coarse-grained rocks have the highest wavelength values for the studied rocks (Table 1) and also the degree of salt crystallization. The wavelength of output signals depends on grain size, while microstructural components of the rocks such as pores, fractures, grains and the presence of salts operate as a wavelength filter. The frequency associated with the first pulse of the output signal differs from the central frequency of the input elastic waves which is fixed at 1 MHz. This is the result of the interaction between elastic wave and microstructural rock components. Manual picking of the onset time for elastic waves becomes more difficult with increase in wavelength, although this observation is less important for P-waves than S-waves (Figs. 1 and 2).

Table 1: Automatic mean values of grain size, P- (VP) and S-wave (VS) and wavelength (λ) for fresh and weathered sandstones.

      The recorded signals highlight that the microstructural components of rocks and their modification by salt crystallization affect the output signal. Values for manual and calculated P-wave are almost equal (Fig. 3). Without concluding which results are closer to reality, one may compare the results of a proposed method with the results obtained by a human analyst (Sarout et al., 2009). If we assume that manual measuring offers a true or reference value (e. g.: Siegesmund and Dürrast, 2011), then it can be concluded that the proposed method accurately calculates P-waves velocities for a range of studied rocks.

      Their discrepancies are, on average, around 0.8 %, which is within the experimental error of the onset time picking measurements. However, these discrepancies increase for altered samples (2.9 %), which could be related to the surficial roughness and microstructure modification by crystallisation pressure and the presence of remaining salts.

      As grain size and weathering increase, the determination of arrival time of S-waves becomes more problematic due to the contamination of S-waveforms by P-waves, a lower signal-to-noise ratio and an increase of wavelength. These difficulties are more prevalent in the weathered samples, where the manual picking of the onset-time becomes more difficult and time-consuming.

      Despite experimental problems, S-wave velocity values are in agreement with previous published data for similar rock types. Therefore manual measurements can be considered as a “true” or reference value (e. g.: Siegesmund and Dürrast, 2011). When comparing manual and calculated S-wave values, it can be observed that the proposed methodology calculates accurate values S-wave velocity of the studied rock types (Fig. 3). Discrepancies between the manual and automatic methods are within experimental error of the onset time picking measurements of S waves velocities, with a discrepancy of around 5 %. Generally, V_S values are slightly higher when calculated using the automatic method in comparison to the manual method (Fig. 3).

      This methodology successfully distinguishes between P-waves and S-waves based on criteria relating to symmetry, amplitude and duration. One advantage of this method is the limitation of subjectivity of the human analyst. Moreover, this study has identified that the main peak frequency of P- and S-waves are comparatively different (Table 1); a discrepancy that can be used as a further differentiating characteristic. This methodology is recommended for fresh and weathered stones with a medium-coarse grain size (0.5–1 mm). This methodology may be particularly helpful where the quality of the S-waveform signals are poor, resulting in difficulties with manual picking of the onset time.

      195Conclusions

      This paper addresses determination of the picking of the onset of P- and S-waves in transmitted output waveforms on weathered sandstones. Stones weathered by salt crystallisation show an increase in surficial roughness and their microstructural properties are strongly modified by crystallisation pressure and the presence of remained salts. The wavelength of output signals depends on grain size and weathering, while microstructural components of stones and the presence of salts operate as a wavelength filter. Manual picking of the onset time for elastic waves becomes more difficult with increase in wavelength, although this observation is less important for P-waves than S-waves. Particularly, as grain size and stone alteration increase, the determination of arrival time of S-waves becomes more problematic due to the contamination of S-waveforms by P-waves, a lower signal-to-noise ratio and an increase of wavelength.

      Figure 1: P signals for fresh and weathered samples for (a) Doddington sandstone, D; (b) Forest of Dean, F, measured in the parallel direction to bedding; and (c) St. Bees sandstone, BC.

      196

      Figure 2: S signals for fresh and weathered samples for (a) Doddington sandstone, D; (b) Forest of Dean, F, measured in the parallel direction to bedding; and (c) St. Bees sandstone, BC.

      Figure 3: Comparison of P (a) and S (b) wave velocities obtained manually (manual) and automatically (authomatic).

      The automatic onset time from recording P- and S- waveforms is compared to manual picking, which is considered as a true or reference value. The discrepancies between automatic and manual measurements are within the experimental error of the onset time picking measurements.

      This methodology is recommended for fresh and weathered stones with a medium-coarse grain size (0.5–1 mm). This methodology may be particularly helpful in samples where the quality of the S-waveform signals is poor, resulting in difficulties with manual picking of the onset time. The great advantage of this methodology is the accuracy and reproducibility of the obtained results, which do not depend on human subjectivity.

       Acknowledgements

      This project was supported by a mobility scholarship awarded by the University of Glasgow Graduate School, and Historic Environment Scotland for funding the attendance at Stone2020 and the Regional Government of Madrid (Spain) [Top Heritage, grant number S2018/NMT-4372].

       References

      Benavente, D., Galiana-Merino, J. J., Pla, C., Martinez-Martinez, J., Crespo-Jimenez, D., 2020. Automatic detection and characterisation of the first P- and S-wave pulse in rocks using ultrasonic transmission method. Engineering Geology, 66, 105474.

      Benavente, D., Martinez-Martinez, J., Cueto, N., Ordonez, S., Garcia-del-Cura, M. A., 2018. Impact of salt and frost weathering on the physical and durability properties of travertines and carbonate tufas used as building material. Environ. Earth Sci. 77, 147.

      Galiana-Merino, J. J., Rosa-Herranz, J. L., Rosa-Cintas, S., Martinez-Espla, J. J., 2013. SeismicWaveTool: Continuous and discrete wavelet yyysis and filtering for multichannel seismic data. Comput. Phys. Commun. 184, 162–171.

      Sarout, J., Ferjani, M., Gueguen, Y., 2009. A semi-automatic processing technique for elastic-wave laboratory data. Ultrasonics 49, 452–458.

      Siegesmund S., Dürrast H. (2011) Physical and Mechanical Properties of Rocks. In: Siegesmund S., Snethlage R. (eds) Stone in Architecture.