Fault diagnosis for machinery based on feature extraction and general regression neural network
- 59 Downloads
Fault diagnosis for the maintenance of machinery is more difficult since it becomes more precise, automatic and efficient. To tackle this problem, a new feature extraction method for signal processing is developed and a general regression neural network (GRNN)—based method is proposed in this paper. Features are extracted from vibration signals that collected from mechanical systems and a feature selection method based on Euclidean distance technique (EDT) is applied. Then, the selected features are processed by the fault characteristic frequencies of mechanical components. And a part of processed data is as train samples and the others as test samples. Finally, the samples are inputted to GRNN to train and verify the model. The proposed method is applied as a fault diagnosis method for both planetary gearbox and bearings datasets, and the performance of it is validated by compared to such methods as radial basis function neural networks (RBFNN), probabilistic neural network (PNN) and a combination model (EMD–EDT). The experimental results show that the GRNN-based method has an advantage over other similar approaches.
KeywordsFeature extraction Signal process Fault diagnosis General regression neural network
- Fault data sets (2013) http://www.mfpt.org/FaultData/FaultData.htm. Accessed on 10 April 2013
- Feng ZP, Chu FL, Song XG (2004) Application of general regression neural network to vibration trend prediction of rotating machinery. In: International symposium on neural networks, Dalian University of Technology, China, 29 August 2004, 767–772Google Scholar
- Lebold M, McClintic K, Campbell R et al (2000) Review of vibration analysis methods for gearbox diagnostics and prognostics. In: Proceeding of the 54th meeting of the society for machinery failure prevention technology, Virginia Beach, VA, 1–4 May 2000, pp 623–634Google Scholar
- Polat O, Yildirim T (2007) Recognition of patterns without feature extraction by GRNN. In: International conference on adaptive and natural computing algorithms, Warsaw University of Technology, Poland, 11 April 2007, pp 161–168Google Scholar