Automation in ambient vibration analysis for soil characterisation

Abstract

For many geo-technical applications knowledge of the soil characteristics is essential for assessing the suitability of a site for various constructions. Recently, various non-invasive methods have been proposed to infer the shear wave velocity (Vs) profile from active and passive sources. Non-invasive methods have the advantage of being low in monetary cost and represent a potential method for improving the low coverage of invasive methods. However, there are many concerns regarding the reliability of applying non-invasive methods to stratigraphic profiling. This research aims to develop effective algorithms to improve the coverage of site assessment and to reduce the amount of user expertise required for geo-technical applications. The proposed approach is to identify suitable methods of site assessment and the challenges that result from the choice of methods. In this case the passive ambient vibration Horizontal-to-Vertical Spectral Ratio (HVSR) method and Evolutionary Algorithms are chosen. The challenges associated with the HVSR method are identified here as the Algorithm Selection Problem and objective function manipulation. The advantage of the proposed approach is a practical reduction in manual modelling and expert interpretation, while improving coverage. This is achieved by an integrated approach to the inversion process. With the proposed evolutionary strategies and objective function manipulation, good approximations to the shear wave profiles have been retrieved automatically for real data sets. The automated HVSR inversion results are in close agreement with the profiles produced by expert manual modelling. The results are also verified by comparison with invasive geo-technical tests done nearby.

Date of Award	2014
Original language	English