Use this URL to cite or link to this record in EThOS:
Title: Developing anomaly detection, diagnostics and prognostics for condition monitoring with limited historical data in new applications such as tidal power
Author: Galloway, Grant Stewart
ISNI:       0000 0004 6422 5997
Awarding Body: University of Strathclyde
Current Institution: University of Strathclyde
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Tidal power is a promising source of renewable energy worldwide, more stable and predictable than alternatives such as offshore wind power. However, the harsh operating environment makes maintenance of tidal turbines difficult and costly to perform. Intelligent condition monitoring systems, utilised as part of a condition-based maintenance strategy, can provide operators with timely and accurate indications of faults before serious damage occurs. Nevertheless, tidal technology is a new application, where deployments are currently in its infancy. Therefore, there is limited practical experience of how faults will develop within tidal turbines in operation. This thesis aims to investigate the requirements of condition monitoring methods, from both theory and knowledge of similar fields (such as offshore wind), to develop an approach for applying condition monitoring in new applications. This work first investigates the response of a commercial-scale tidal turbine in operation through a data mining analysis of condition data from the Andritz Hydro Hammerfest HS1000 tidal turbine. Data mining was performed following the CRISP-DM methodology and was used to build models of the normal condition response of turbine components, from which anomalous behaviour indicative of the development of faults can be detected. This approach was then expanded to include both diagnostic and prognostic modelling, where faults can be automatically classified and the remaining useful life of equipment undergoing degradation can be estimated. This has resulted in a generalised framework, based on CRISP-DM, that can be applied to perform condition monitoring in new applications, learning from the response of machinery over time during its operation.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral