Multi-core Accelerated Discriminant Feature Selection for Real-Time Bearing Fault Diagnosis
This paper presents a real-time and reliable bearing fault diagnosis scheme for induction motors with optimal fault feature distribution analysis based discriminant feature selection. The sequential forward selection (SFS) with the proposed feature evalua
- PDF / 1,461,224 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 61 Downloads / 195 Views
Department of Electrical, Electronics, and Computer Engineering, University of Ulsan, Ulsan, South Korea [email protected], [email protected], [email protected], [email protected] 2 School of Electronics and Computer Engineering, Chonnam National University, Gwangju, South Korea [email protected]
Abstract. This paper presents a real-time and reliable bearing fault diagnosis scheme for induction motors with optimal fault feature distribution analysis based discriminant feature selection. The sequential forward selection (SFS) with the proposed feature evaluation function is used to select the discriminative feature vector. Then, the k-nearest neighbor (k-NN) is employed to diagnose unknown fault signals and validate the effectiveness of the proposed feature selection and fault diagnosis model. However, the process of feature vector evaluation for feature selection is computationally expensive. This paper presents a parallel implementation of feature selection with a feature evaluation algorithm on a multi-core architecture to accelerate the algorithm. The optimal organization of processing elements (PE) and the proper distribution of feature data into memory of each PE improve diagnosis performance and reduce computational time to meet real-time fault diagnosis. Keywords: Fault diagnosis Feature selection separability Multi-core architecture
Class compactness
Class
1 Introduction Induction motors are widely used as a low-speed rotating machine to support steady rotational speed with heavy loads in industries [1]. In rotating machines, bearings are the most significant element which supports stationary rotational speed with heavy load. In a bearing, due to variable-speed, improper loading, and rapid rising of voltage pulse, faults are generated, which results in an unexpected shutdown of manufacturing process [2, 3]. Therefore, real-time and reliable bearing fault diagnosis is needed to avoid these unexpected failures. An effective data-driven bearing fault diagnosis model involves three basic steps: (1) data acquisition from bearings during operation, (2) feature extraction for identifying specific fault signatures, and (3) classification of different faults generated from © Springer International Publishing Switzerland 2016 H. Fujita et al. (Eds.): IEA/AIE 2016, LNAI 9799, pp. 645–656, 2016. DOI: 10.1007/978-3-319-42007-3_56
646
Md. Rashedul Islam et al.
the bearing. Many signal processing techniques have been developed in time and frequency domains to extract significant fault information from incoming signals of defective bearings [4–6]. However, these methods are only suitable in different situations depending on signal characteristics [7]. Thus, hybrid feature extraction has been widely used for a fault diagnosis system. Hybrid feature selection can explore maximum possible fault signatures by extracting features using different feature extraction paradigms. The resultant high-dimensional feature vector, however, may have redundant information, which can lead to performance degrada
Data Loading...