Failure prediction method of gearbox based on BP neural network with genetic optimization algorithm

A gearbox is one of the components of double-fed wind turbine which owns the longest downtime and highest maintenance cost. The difficulty of failure prediction of a gearbox based on data is the practical verification of prediction methods. In this paper, a failure prediction method based on the data of vibration feature trend was proposed for a gearbox. The feature values reflecting the condition of a gearbox were extracted through the measured vibration signals. The BP neural network prediction model was established with feature values of a period of time. The structure of BP neural network was optimized by genetic optimization algorithm. The training speed was improved and the prediction error was reduced. The proposed method was verified with practical measured data.