Signal processing techniques related to laboratory measurements of ultrasonic S-wave velocities in rocks

This thesis aims to evaluate the effect of signal processing techniques related to ultrasonic laboratory measurements of shear waves. Compressional and shear wave velocities play an important role in static elastic rock deformation behaviour estimation. Onsets of compressional and shear wave signal have to be determined in order to calculate the corresponding wave propagation velocity. Onset estimation by automation is especially problematic in shear wave signals due to noise caused by reflections and refractions, which results in inaccurate onset estimations and, therefore, requires manual onset picking which is time-inefficient and, hence, costly. Akaike Information Criterion (AIC) is the automated picking method applied to the ultrasonic signals in this thesis. By efficiently processing shear wave signals it was tried to optimize the results of the AIC. Ten processing techniques from biomedical engineering, statistical signal processing, audio and speech processing and RADAR applications were thoroughly researched. Their applicability to ultrasonic signals was reasoned based on literature. Six applicable signal processing techniques were eventually applied to 30 synthetic and 30 real ultrasonic signals. The mean and standard deviation of the error related to onset estimation before and after processing was used for evaluation. Visual comparison before and after processing was also executed to evaluate the visual impact of the processing techniques. Results showed that only a Butterworth high-pass filter visually enhances synthetic and real ultrasonic signals and improves the mean and standard deviation with respect to onset estimation. A Chebyshev high-pass filter also improved onset estimation results, but deteriorated the visual interpretation of the time signals. A simple amplitude filter unexpectedly provided the best results with respect to onset estimation. It is concluded from this studies that onset estimation by AIC can be improved by application of related signal processing techniques. This could be beneficial in estimation static deformation behaviour. Potential room for improvement is found within parameter optimisation and synthetic signal production