Condition monitoring - fault detection and diagnosis - Literature review Example

Fault Identification and Diagnosis for Reciprocating Compressors Based on Support Vector Machine and Genetic Algorithm Support vector machine (SVM) methods have been used along with grid search and other learning techniques. Similarly, SVM methods have been applied in tandem with Genetic Algorithms (GAs) to obtain fault classification for fault prone software components. Comparably, SVM in tandem with GAs produced the best classification results proving its superiority over other methods for fault classification (Martino et al., 2011). Mathematical modelling and testing of SVM methods with GA indicate their superiority over regular SVM methods in dealing with unbalanced classes to produce higher classification and faster learning (Amer et al., 2011). Hybrids of SVM methods such as combined SVM (CSVM) have been used extensively for process control such as in the Eastman process.

Results indicate the superiority of SVM based methods over other methods of control (Tafazzoli & Saif, 2009).SVM methods have been employed extensively in order to classify reciprocating compressor faults. SVM methods were employed in order to classify faults of reciprocating refrigeration compressors through the application of wavelet transform and statistical methods. Significant features were extracted from both raw noise signals and vibration signals. The selection of relevant RBF kernel parameters was carried out through iteration (Yang et al., 2005). In a similar application, SVM methods were applied to reciprocating compressors butterfly valves to classify cavitation faults (Yang et al., 2005). A comparable research was performed on reciprocating compressor valves to classify faults through vibration signals alone.

Data for this purpose was gathered from the surface of the valve and the resulting vibration signals were decomposed by applying local wave methods (Ren et al., 2005).One of the larger problems posed by reciprocating compressor valves is the non stationary and non linear characteristics of the extracted vibration signals. In order to deal with the non stationary and non linear nature of such data, information entropy with good fault tolerance potential was utilised as the feature parameter fed to a SVM.

This was utilised as being a comprehensive characteristic of the raw vibration signal. The resulting decision function was used to solve the limits of traditional fault classifications. The added strength of the SVM was its ability to be trained with only a few input samples to deal with multiple new faults (Chen & Lian, 2010).The small linear pattern recognition performance and relatively small data sets extracted from reciprocating compressor valves present unique problems for fault classification.

SVM has been utilised to deal with such limitations by employing information entropy since it is flexible as well as being liberal in terms of non linear behaviour. The SVM was trained by using a small vibration data set from reciprocating compressor valves characterised by information entropy. The fault classification results of SVM methods proved accurate enough for fault classification in valve failure of reciprocating compressors (Cui et al., 2009).Another limitation of vibration data from reciprocating air compressors is the small amount of fault data that can be extracted from regular runs compared to a large number of excitation sources.

SVM was applied in tandem with Statistical Learning Theory (SLT) in order to overcome this challenge. Vibration signals were extracted from the rolling bearing in a reciprocating compressor’s crankcase through the utilisation of a test bed. A SLT scheme was developed to extract features that were then fed to the SVM for intelligent fault classification. Results showed that the application of these methods identified faults immediately with significant accuracy (Sheng et al., 2009).BibliographyAmer, F.Z. et al., 2011. A real-valued genetic algorithm to optimize the parameters of support vector machine for classification of multiple faults in NPP.

NUKLEONIKA, 56(4), pp.323-32.Chen, Z. & Lian, X., 2010. Fault Diagnosis for Valves of Compressors Based on Support Vector Machine. In 2010 Chinese Control and Decision Conference. Xuzhou, 2010. CCDC.Cui, H., Zhang, L., Kang, R. & Lan, X., 2009. Research on fault diagnosis for reciprocating compressor valve using information entropy and SVM method. Journal of Loss Prevention in the Process Industries, 22, pp.864-67.Martino, S.D., Ferrucci, F., Gravino, C. & Sarro, F., 2011. A Genetic Algorithm to Configure Support Vector Machines for Predicting Fault-Prone Components.

Product-Focused Software Process Improvement Lecture Notes in Computer Science, 6759, pp.247-61.Ren, Q., Ma, X. & Miao, G., 2005. Application of support vector machine in reciprocating compressor valve fault diagnosis. Lecture Notes in Computer Science, 12, p.81–84.Sheng, F., Jing, L. & Yabin, Z., 2009. Fault Diagnosis System for Reciprocating Air Compressor Based on Support Vector Machine. In Proceedings of the 2009 International Workshop on Information Security and Application. Qingdao, 2009. IWISA.Tafazzoli, E.

& Saif, M., 2009. Application of combined support vector machines in process fault diagnosis. In 2009 American Control Conference. St. Louis, 2009. ACC.Yang, B.S., Hwang, W.W., Kim, D.J. & Tan, A.C.C., 2005. Condition classification of small reciprocating compressor for refrigerators using artificial neural networks and support vector machines. Mechanical System and Signal Processing, 19(2), p.371–390.Yang, B.S., Hwang, W.W., Ko, M.H. & Lee, S.J., 2005. Cavitation detection of butterfly valve using support vector machines.

Journal of Sound Vibration, 287(1-2), p.25–43.