Abstract

This study covers a maintenance decision making model, where condition monitoring is applied to rotating components of horizontal axis wind power generation systems installed in Costa Rica, in order to reduce the incidence of unexpected failures, which are consequences of the run-to-failure policy or following strictly the manufacturers’ suggestions without considering the operational environment, very common practices in the national wind industry.

Aiming the optimization of operation and maintenance costs, a model where two failure probability threshold values are defined is presented. These threshold values allow the component replacement decision making. Moreover, the initial guidelines for executing this strategy in a wind farm are offered in this paper.

The life percentage predictions required by the offered model, are obtained using artificial neural networks for each component (rotor, main bearing, gearbox and electric generator), which use representative condition monitoring variables as inputs.

Keywords: Reliability; Wind Turbines; Artificial Neural Networks; Cost Optimization