Abstract
To date there has been no published reliability data for Tidal Stream Turbines (TST’s). Thus, reliability assessments of failure critical components are highly uncertain. To reduce the Levelised cost of energy (LCoE) of TST technology it is critical to be able to perform accurate reliability assessments of devices at the design stage.
Based on experience from the wind industry, the power take off (PTO) is a failure critical assembly representing a high proportion of total turbine failures and downtimes. Many studies have found that the pitch system (PS) contributes to most turbine failures [1], [2]. Bayesian reliability methods are of interest in industries where data is scarce or commercially sensitive as they allow for the use of surrogate data sources along with domain knowledge; they also allow for this knowledge to be updated as new data becomes available [3].
This research develops a Bayesian reliability model of a Horizontal axis TST PS using state of the art surrogate failure data and domain knowledge. The paper discusses the rationale behind Bayesian modelling and provides a framework for TST developers, researchers, consultants and the like to use their own device failure data when it becomes available to make robust PS reliability assessments with quantified uncertainty levels. The components of the PS focussed on in this research are the Dynamic Seal, Roller Bearing and Electric Motor. These are seen to be failure critical areas. Empirical Physics of Failure (PoF) equations are used to determine individual failure rates for these parts and then Monte Carlo methods are used to calculate the combined failure rates. This combined part failure rate distribution is then updated using representative ‘real’ PS failure data (from the wind industry) to highlight the Bayesian updating process. This paper represents a step change in current reliability assessment methods in the Tidal industry by presenting a method to attribute specified levels of uncertainty to the underlying parts (e.g. design parameters and correction factors) of the failure rate calculations.