Condition monitoring- Fault diagnosis Literature review

Condition monitoring- Fault diagnosis - Literature review Example onlinear behavior of the manufacturing device has regression performed to filter out noise through the utilization of a kernel based Bayesian structure. The GA tabulates the near optimal control parameters in order to maximize the required objective (Yuan et al., 2007). Rotating machinery fault diagnosis has been attempted using thermal imaging processed through RVM methods in combination with bi-dimensional empirical mode decomposition (BEMD) and generalized discriminant analysis (GDA). The BEMD enhanced thermal image is treated with GDA to reduce features after which RVM is implemented for fault classification (Tran et al., 2013). RVM has been compared to support vector machine (SVM) methods to demonstrate its robustness for gear fault detection. Compared to SVM, the RVM method required lesser kernel functions and learning time while demonstrating comparable performance (He et al., 2009). RVM combined with GA has been utilized in state classification of roll bearings. The GA is applied to determine training parameters for RVM. Experimentation and analysis revealed that the application of GA in combination with RVM produced better results than back propagation neural networks and SVM (Li & Liu, 2010). A comparison of multi class RVM and SVM methods for low speed bearing fault detection revealed that RVM methods held great promise for accurate fault classification. Component analysis was carried out in order to classify features and to reduce the dimensions of the raw data set. Fault diagnosis was carried out with feature extraction and without it (Widodo et al., 2009). Wavelet packet feature extraction was applied in tandem with RVM for detecting gear faults. Using the Fisher criterion, the discrimination power of the features is tabulated and two optimal features are selected in the time domain and wavelet domain. These are used as inputs to the RVM. Comparisons with SVM revealed that the RVM based method produced better results for online classification (Li