Dr Pete Green, Department of Civil Engineering and Industrial Design, University of Liverpool, UK

On the adoption of Machine Learning for Engineering Applications
When Feb 18, 2019
from 02:00 PM to 03:00 PM
Where LR1, Thom Building
Contact Name
Contact Phone +44 (0) 1865 273030
Add event to calendar vCal

It is clear that data-intensive approaches, facilitated through Machine Learning algorithms, has the potential to revolutionalise a number of engineering disciplines. This is perhaps most obvious within the manufacturing sector, where the transition to digitalised data-based systems has been hailed as the 4th industrial revolution (or ‘industry 4.0’).

There are, however, several significant barriers to adoption that must be overcome before the full potential of data-based discovery can be realised. This talk examines some of these barriers through a series of case studies. A range of issues are explored including:

  • Bad data: potential problems with ‘standard’ Bayesian approaches to Machine Learning when unexpected errors occur in the data-acquisition processes.
  • Decision making: the crucial role that uncertainty quantification has to play if Machine Learning is to be applied in risk-averse engineering sectors.
  • Combining Big Data with Sparse Data: the role of ‘semi-supervised learning’ in combining large cheap-to-obtain datasets with measurements that are expensive and time consuming to acquire.
  • Knowledge Transfer: the non-technical barriers that often prevent Machine Learning expertise from being effectively transferred into industry.

Case studies include applications in additive manufacturing, aerospace and structural dynamics.