Title

Back Analysis using Machine Learning for the Observational Method – Lessons Learnt and Future Directions

 

Special session organizers

Franz Tschuchnigg, Assoc Prof. Graz University (ISSMGE TC309 and 206).

Duncan Nicholson, Arup Fellow, (ISSMGE TC206 Chair)

Antonio Cañavate Grimal, Principal geotechnical engineer, Ove Arup and Partners Ltd.

 

The ISSMGE TC206 on Interactive Design and the Observational Method (OM) is currently focusing on field monitoring data with TC220 and back analysis with TC103 (numerical methods) and TC309 (Machine Learning, ML and Big Data, BD).

 

Overview

The Observational Method (OM) process is well established over the last 70 years and is recognised by Eurocodes. It covers a wide range of engineering applications. These include, retaining walls, tunnels, embankments on soft ground, ground treatment, dewatering, piling, adjacent infrastructure/ asset protection, etc.

 

The application of Numerical modelling tools and the advanced soil constitutive models have improved predictions, however there are often uncertainties in ground and groundwater conditions affecting the design parameters, and uncertainties in actual construction data, structural and ground-structure interaction behaviours.

 

Field monitoring techniques have also been improved with automated data logging enabling Big Data (BD) techniques to be applied. This will enable an overview of field behaviour to be established as construction progresses. These can be compared with the numerical model predictions on a ‘real-time’ basis.

 

Many OM applications require back analysis to be undertaken quickly during the construction. This is to enable model input parameters to be reassessed, calibrated and fed into the reanalysis of remaining and future construction work. Machine Learning (ML) is being introduced to optimise and speed up this back analysis process.

 

A future goal is to apply ML techniques to facilitate the “real time back analysis” as the construction work progresses.

 

This special session aims to pull together experience in both practice and academic fields.

 

Topics

This special session invites authors to submit at least 7 high quality research papers. Oral presentations will be made. The following topics shall be included:

• Back analysis case histories where the analysis was carried out during construction for the OM. Contractual issues should be discussed in these papers.

• The back analysis of case histories using ML tools and manual interpretation.

• Comparisons of soil parameters from site investigations and back analysis.

• The use of BD to interrogate precision and accuracy of data.

• The use of BD to visualise / integrate output for comparison with numerical models.

​

​This special session aims to pull together experience in both practice and academic fields. Extended abstracts (maximum length 4 pages) should be submitted until 01.12.2022.

​