A data-driven approach for fault diagnosis of drivetrain system in a spar-type floating wind turbine based on the multi-point acceleration measurements

Abstract

This paper deals with the condition monitoring of a floating wind turbine drivetrain using multi-point acceleration measurements. Single sensor data obtained from drivetrain system may provide insufficient information about the health condition due to the complicated structure and applied loading on this system. As a result, multi-point measurements are required to be employed for reliable fault diagnosis. The shared information between the multi-point measurements can be used for identifying the system's condition. In this study, the fault diagnosis of the floating wind turbine drivetrain system is performed using a data-driven approach. Fault cases are considered in bearings most likely to damage. A combined principal component analysis (PCA) and deep convolutional neural network (CNN) is proposed to extract common and fault-related information between the measurements on the one hand and to classify different health conditions of the drivetrain on the other. It will be demonstrated that PCA-based information provides more satisfactory fault diagnosis results than individual sensor data. The method is numerically validated using the acceleration responses obtained from a 5-MW reference drivetrain model installed on a spar-type floating wind turbine.