Use this URL to cite or link to this record in EThOS:
Title: Development of an open source drillability framework : using offshore pile top large diameter drills as a case study
Author: Tonkins, M.
ISNI:       0000 0004 7969 5414
Awarding Body: University of Exeter
Current Institution: University of Exeter
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Thesis embargoed until 29 Jul 2024
Access from Institution:
Offshore socket/shaft development is carried out from a specialised form of drilling utilising Pile Top Large Diameter Drills (PTLDD), and like all forms of offshore developments carry additional risks compared to the development of land-based infrastructure. Therefore, to reduce risks to drilling cost centres, estimation of the drills rate of penetration is required for project planning using empirical drillability models. At the time of writing, there is no drillability model to predict the performance of offshore PTLDDs. Machine learning methods of drillability have proved to increase accuracy of drillability estimates with rotary drills and Tunnel Boring Machines. In addition, the use of machine learning allows for automatic learning and model selection within confined boundaries, saving both time and increased accuracy. A drillability KDD process has been modified from the standard KDD process by Fayyad (1996). This process outlines the three key stages of development; Engineering Geological Model Creation, Drill mapping and machine learning. Each of these stages have been developed within the Python ecosystem allowing for open-source model creation and machine learning using SciKit Learn (Pedregosa et al., 2011). This workflow was applied to the historical PTLDD case studies which cover a range of geological terrains and engineering applications. This is the first time such as process has been applied to this niece style of drilling. The resulting performance estimations for the Rate of Penetration have been successfully estimated with a coefficient of determination (r2) of 0.7 from a combination of Hoek and Brown rock mass strength, unit weight, hole diameter, depth from drill and drill rig size. This model can be further improved by using a risk-based approach from the random sampling of geological parameters resulting in an r2 of 0.93. This model has the advantage of increasing observations from different realisations of the ground conditions.
Supervisor: Coggan, J. ; Johanning, L. Sponsor: Engineering and Physical Sciences Research Council (EPSRC)
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Drillability ; Machine Learning ; Offshore