Jeremiah Corrigan
Data-driven techniques to improve industrial process modelling.
Email: [email protected]
Project title
Data-driven modelling of industrial processes with applications in the nuclear waste processing industry
Supervisors
- School of Engineering
- Sellafield Ltd
- Katy Spencer
Project description
Data-driven modelling using machine learning techniques is a useful tool for applications in the process industry. Mechanistic models face limitations, such as a lack of theoretical knowledge or time/cost issues. In an age where an ever-increasing amount of data is being collected, this makes data-driven approaches more appealing. Data-driven models can improve applications in industrial processes. These include process control, monitoring and optimisation. But for the creation of such models, we need robust and reliable techniques.
This research will develop data-driven techniques to improve the robustness and reliability of models when applied to industrial process data. We will apply developed techniques to the nuclear waste vitrification process at Sellafield Ltd. Developed models can provide predicted measurements of key quality variables. The measurements are more frequent and have less delay. Such models are often known as a “soft sensor”. They can also aid process optimisation.
EPSRC and Sellafield Ltd are funding the project.
Qualifications
MEng in Chemical Engineering with Honours in Process Control