Automated Back Analysis of Geotechnical Infrastructure Performance using Optimization

The recent development of powerful, high resolution full-field measurement techniques (e.g. digital imaging) now gives access to a very large quantity of data during an experimental test or when monitoring field construction or field deformations (e.g. earthworks). Pilot studies now indicate that it is possible to utilise this data to directly characterise the properties of the soil in the system without having to sample the soil and test it in the laboratory. This adds a significant additional dataset to any problem and gives the engineer valuable additional capability with which to assess the problem in hand. Such a technique may be applied at laboratory model test level or at full scale in the field. The proposed PhD project is focused on the former.
In the context of laboratory model tests, modern digital image correlation techniques allow determination of large dataset of high resolution displacement and strain fields from a set of images. At present these are typically generated using plane strain models with transparent walls. Pilot studies at Sheffield (Gueguin et al., 2015) have shown that it is possible to utilise optimization techniques in combination with these strain fields and the force-displacement response of a structure to reconstruct the stress-strain response of an undrained (clay) soil without any need for a specific analytical or numerical model of the system being tested, i.e. this is a problem agnostic process.
This obviates the need to undertake a separate direct measurement of the stress-strain response of the soil by sampling and testing. Such a process introduces disturbance of the soil and uncertainty in results and may be a significant cause of the lack of fit that has been seen between some experimental data and theory. (Sampling and testing can however still be used to provide additional contextual data).
The aim of the PhD is to develop and explore this technique further, calibrating it on selected physical model studies and extending it to granular materials. This will be enabled through design and testing of specific physical models using the state of the art 1g and centrifuge physical modelling facilities, together with transparent and photo-elastic soil capabilities in the geotechnics research group. Analytical/numerical developments will take advantage of the leading capabilities of the CMD group in applying optimization in Civil Engineering. Such work will strengthen Sheffield's status in physical modelling and use of optimization.
Whilst field application of the technique is not the direct aim of this PhD, industry bodies such as Network Rail have shown interest in the technique where aerial and ground based digital imaging, LIDAR and wireless sensor data can be combined to generate a detailed dataset. One aspect of the research project will be to explore these possibilities in outline in order to provide a wider context for the proposed work.