| Popis: |
With the aim of better identify the running conditions of bearings in wind turbine, a two stages classifier approach is suggested. Firstly, a combined feature set is generated through the application of two techniques: statistical time domain parameters and energy features extracted by discrete wavelet transform (DWT). After that Distance evaluation technique (DET) is employed to choice the relevant parameters, Hereafter, a first support vector machine (SVM) is performed to determine the kind of faults. In the last step, three SVMs classify faults severity for each class. To enhance the classification performances, particle swarm optimization (PSO) is applied to jointly calculate the SVM factors. The obtained results have demonstrated the utility and the efficacy of the suggested condition monitoring Scheme. |
