Turbine fault diagnosis of the oscillating water column wave energy converter based on multi-lead residual neural networks

Abstract

Turbine fault diagnosis of the oscillating water column (OWC) wave energy converter has been a critical engineering challenge. This paper proposes a new approach for diagnosing turbine faults which is based on a correlation analysis of ensemble empirical modal decomposition (EEMDCA) and the fusion of multi-lead residual neural (MLRN) networks. The correlation coefficient between the intrinsic modal function (IMF) components and the original sample data is used to determine the number of lead branches in the MLRN structure. The numerical simulations were conducted for 14 cases (one intact and 13 faulty cases). Based on the data processed by EEMDCA, the diagnostic accuracies of MLRN, single-lead residual neural (SLRN) networks, and CNN are compared. Additionally, the noise immunity of the different methods (MLRN, SLRN, CNN, EEMDCA-MLRN, EEMDCA-SLRN and EEMDCA-CNN) is studied considering four different noise levels. Furthermore, experimental research of the OWC model in a wave flume is conducted under three wave conditions to validate the effectiveness of the proposed EEMDCA-MLRN approach for turbine fault diagnosis. Results indicate that the proposed method can provide references for real-time operation and maintenance monitoring of OWC in practical engineering.