Automatic detection and characterisation of the first P- and S-wave pulse in rocks using ultrasonic transmission method

Abstract In this paper, we propose a new methodology for the automatic picking of P- and S- wave arrivals. The measurement of elastic wave velocities is crucial for characterisation of the elastic properties of rocks, as well as their physico-mechanical properties and durability. P-waves are easy to generate and acquire, whereas the acquisition and subsequent analysis of the S-waves present major difficulties in most investigations of rocks. In our proposed method the recorded signal is first pre-processed in the wavelet domain, removing low-frequency disturbances and filtering noise. The signal is then analysed in the time-domain to detect and characterise the first pulse and, subsequently, estimate the corresponding onset time. The automatic approach analyses all pulses detected in the output signal and selects the first pulse of the P or S wave relative to symmetry, amplitude and duration criteria. This triple check provides greater confidence in the obtained results. We record P- and S- waveforms through the transmission method for a broad range of sedimentary, igneous and metamorphic rocks used in buildings. Results are compared to manual picking, which is considered as a true or reference value. The recorded signals show that microstructural components of rocks have a strong influence on the output signal. Mineralogical composition, porosity and particle size affect the wave velocity, attenuation, wavelength and waveform, which in turn influences the manual picking of the onset time. The proposed methodology provides precise values of the onset time of P- and S- waves. The discrepancies between automatic and manual P- and S- wave picking are 0.7 and 4.4%, respectively. P-wave signals have a high signal-to-noise ratio and their arrival times are clear and easy to determine. However, the arrival time of S-waves presents significant problems, mainly for medium to coarse-grained rocks. In this case, the output signal is contaminated by P-waves and has a lower signal-to-noise ratio. Besides, propagation of the S-wave through the rock affects the frequency and waveform of the first pulse, which complicates manual picking of the onset time. The developed methodology distinguishes between the S-wave and the remaining P-waves contained in the transversal signal, independently of the skill and subjectivity of a human analyst.