- since 1996 -
|Beta-Version of this Page View|
|NDT.net Issue - 2016-08 - Articles ||NDT.net Issue: 2016-08|
Publication: 8th European Workshop on Structural Health Monitoring (EWSHM 2016), July 5-8, 2016 in Bilbao, Spain (EWSHM 2016)
Session: SHM application to Wind turbines
Efficient neuro-fuzzy damage severity estimation in an experimental wind turbine blade using the Fukunaga-Koontz transform of vibration signal correlationsHoell, Simon; Omenzetter, Piotr
Abstract: Recent international commitments for achieving carbon neutrality in the present century pave the way for further advancements in wind energy. To satisfy energy demands, potential energy outputs can be enhanced by erections of ever larger wind turbines (WTs). The use of novel composite materials is required for these developments in order to reduce the mass of WT blades (WTBs). However, long-term safety and reliability of WTs are affected by higher flexibilities and lower buckling capacities of these WTBs. Furthermore, current international standards and guidelines recommend rigidly defined physical inspections for examining the structural state of WTs, which significantly contribute to operation and maintenance costs. The development of effective structural health monitoring (SHM) systems for WTs can counteract cost increase by reducing the risk of dramatic failures, scheduling maintenance actions according to the actual state and lowering overall inspection efforts. The present paper shows a novel approach for damage severity estimation in WTBs based on vibrational responses. First, acceleration signals of a WTB in the healthy and reference damage states are acquired. Second, initial damage sensitive features (DSFs) are obtained from autocorrelation coefficient estimates of these signals. In order to reduce DSF vector dimensions while retaining discriminatory information, the Fukunaga-Koontz transform (FKT) is applied in the third step. The FKT is an extension of the Karhunen-Loeve expansion and provides a ranking of the resulting transformation vectors with respect to the discriminatory information. This information is then used for sequentially constructing a hierarchical adaptive neuro-fuzzy inference system. Using the ranking, inputs, i.e. the scores from transformed initial DSFs, and hierarchy levels are added sequentially until the desired level of accuracy for estimating the damage severity is achieved. Physical experiments with a small scale WTB are performed in the laboratory. A household fan is used to create a contact-free excitation. Different damage scenarios are simulated non-destructively by attaching small masses, which represent the damage severity in the present study. The results are promising for prospective developments of effective vibration-based SHM systems delivering improved safety and reliability of WTs at lower costs.
Keywords: neural network (60), civil engineering (1025), vibration analysis (155), Feature Extraction (64), damage detection (65), wind turbines (9), Fuzzy computing (1), Damage severity estimation (2), Time series methods (1),