the onset-time becomes more difficult and time-consuming. In general, the automatic method obtains S-wave values slightly lower than the manual method and their discrepancies are, on average, lower than 5 %.
Keywords: Wavelet analysis, Building stones, Salt weathering, Primary wave velocity, Shear wave velocity Ultrasound, Non-destructive testing
192Introduction
These non-destructive methods based on elastic wave measurements are commonly used in stone conservation investigations in both laboratory and field experiments. The study of compressional and shear wave velocities is considered a reliable method of determining the elastic, physical-mechanical, and durability properties of studied samples. Elastic wave velocities are frequently investigated either individually or in combination. The compressional or primary (P) wave velocity, Vp, (also termed as ultrasonic pulse velocity) is a frequently studied parameter that is simple to measure in the laboratory. In addition, Vp is also used as an indicator of rock strength and degree of weathering. Obtained in combination with Vp, shear or secondary (S) elastic wave velocity, VS, is used when calculating the Young and Poisson dynamic elastic moduli.
Modification of microstructural properties of rock caused by salt weathering, combined with the presence of crystallised salts and water, strongly affects the rocks elastic wave velocities. (Benavente et al., 2018). The determination of P-velocities in weathered stone samples could present difficulties due to waveform attenuation, which results in output signals showing a low signal-to-noise ratio (SNR). On the other hand, the generation and acquisition of pure S-waves in rocks is difficult (Wang et al., 2009). The most critical problems come from the contamination of S-waveforms by P-waves (P- wave appears before the S- wave arrivals), the reduction of waveform amplitude (lower SNR) and increase of wavelength.
As a result, picking of onset time of the S-wave, for both fresh and weathered samples, and the P-wave in highly weathered samples, becomes a complicated process. All of these difficulties limit the use of S-waves, particularly when calculating the dynamic elastic moduli, which considers a tabulated value of the Poisson’s ratio, yielding dubious estimations of elastic rock properties. Manual picking is a tedious and time-consuming process, which is subject to human error. Furthermore, manual operation relies on the attention of a trained technician, which can be a disadvantage when analysis of large volumes of data is required.
In this paper, we record P- and S-waveforms in three sandstones used in the Scottish heritage and obtain the P- and S-velocities as well as the calculation of wavelength of the P-waves in both fresh and salt weathered samples. We use an automatic methodology for the calculation of onset time that includes the signal pre-processing and the analysis, in the time-domain, of the first pulse symmetry, amplitude and duration criteria. This triple check provides greater confidence in the obtained results. This method limits human error, while improving the accuracy and reproducibility of P and S wave onset time estimations.
Materials and Methods
Stones
In this paper three sandstones used in the Scottish heritage are investigated:
Doddington sandstone (D) is a common building stone used throughout Scotland and is currently used as a replacement stone at Jedburgh Abbey, Scotland. The sandstone is quarried at Doddington quarry near Wooler in the Scottish Borders region; and forms part of the Fell formation, deposited during the early-mid Carboniferous. The sandstone is mineralogically mature and textural immature; with rounded-sub rounded grain shape and an average grain size of 0.25 mm. Mineralogy consists of: quartz (90 %); feldspar (mainly orthoclase) (10 %); and minor amounts of muscovite, lithic fragments and clay within the matrix. The sandstone is classified as a quartz-arenite.
St. Bees sandstone (BC) is currently used as a replacement stone at Arbroath Abbey and is quarried on the west coast of Cumbria, England. The red-dark red sandstone is part of the Chester sandstone formation, deposited during the early Triassic. The sandstone is texturally and mineralogically immature; with (sub)rounded – (sub)angular grain shape and an average grain size of 0.12 mm. The sandstone is classified as a lithic arkose. Mineralogy consists of: quartz (60 %); feldspar (orthoclase and plagioclase) (20 %); lithic fragments (10 %); mica (biotite and muscovite) (10 %); and clay-rich matrix. BC exhibits planar 193bedding on a scale of 1–2 mm. Therefore, during this study BC samples were measured in two directions: parallel and perpendicular to bedding, to determine the degree of anisotropy and potential effects on output signals.
Forest of Dean (F) is a distinctive grey-green sandstone used throughout the UK, for example in the restoration of Dunkeld Cathedral, Scotland. The sandstone forms part of the Pennant sandstone formation deposited during the mid-late Carboniferous; and is currently quarried at Barnhill quarry in the county of Gloucestershire, England. The sandstone is mineralogically and texturally immature; with (sub) angular grain shape and an average grain size of 0.19 mm. Mineralogy consists of: quartz (70 %); feldspar (orthoclase and plagioclase) (10 %); lithic fragments (15 %); muscovite (5 %); and a clay-rich matrix. The sandstone is classified as a (sub)litharenite.
Ultrasounds
The ultrasonic measurements were carried out by means of the transmission method, which consists of two piezoelectric sensors coupled to the sample at constant pressure. Compressive (P) and shear (S) waves were measured using polarised Panametric transducers (1 MHz). Emitting-receiving equipment (Panametrics-NDT 5058PR) and an oscilloscope (TDS 3012B-Tektronix) were used to acquire and digitalize the waveforms to be displayed, manipulated and stored. Two types of visco-elastic couplants were used to achieve good coupling between the transducer and the sample: one of a fluid consistency for the P-wave transducers (eco-gel), and another of a more viscous consistency for the S-waves (SWC, shear wave couplant, GE Panametrics).
In this work, we have applied a new method for automatic characterisation of the first pulse in both compressional and shear waves described in Benavente et al., (2020). Firstly, the recorded signal is pre-processed in the wavelet domain. This step filters noise and removes low-frequency disruption. Secondly, the recorded signal is analysed in the time-domain. This step identifies and characterises the first pulse, allowing for an estimation of onset time. Using the automatic approach allows for all detected pulses in the output signal to be analysed, where the first pulse of the P or S wave is determined based on criteria related to symmetry, amplitude and duration. The method outlined above is carried out using Matlab based on Galiana-Merino et al. (2013), where parts of the pre-processing stage are carried out using the Wavelet Toolbox. This triple check provides greater confidence in the results. We obtain the automatic onset time from recorded P- and S- waveforms and they are compared to manual picking, which is considered as a true or reference value.
Finally, the wavelength, λ, was calculated using the velocity, v, and frequency, f, of the first pulse as λ=v/f.
Salt crystallisation test
Salt crystallisation tests were in concordance with EN-12370 (1999) recommendations, respectively. 4 cm cubic samples of each lithotype were tested and underwent cycles of saline immersion (14 % w/w Na2SO4 solution, at 20 °C for 4 hours), drying (at 60 °C for 16 hours) and cooling at room conditions (20 °C, for 4 hours). Samples were exclusively cleaned at the end of the test (after the 15 cycles). P and S waves were measured before and after the durability test. Every measurement of the P and S waves was repeated three times in order to test the reproducibility of the experiments and the corresponding results. Figures 1 and 2 display, respectively, P and S signals for fresh and weathered samples. They aim to show the evolution of waveform rather than wave velocities
