Dissimilarity-Based Sequential Backward Feature Selection Algorithm for Fault Diagnosis

• Published in: Neural Information Processing Vol. 10637; pp. 393 - 401
• Main Authors: Xue, Yangtao, Zhang, Li, Wang, Bangjun
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2017; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Dissimilarity; Fault diagnosis; Feature selection; Support vector machine
• ISBN: 9783319700922; 3319700928
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-70093-9_41

The aim of feature selection applied to fault diagnosis is to select an optimal feature subset that is relevant to the faults. The optimal feature subset with fewer features contains more discriminative information which can improve the performance of fault diagnosis models. A novel sequential backward feature selection method based on dissimilarity is proposed to detect the difference of features between normal and fault data. The proposed feature selection method can be used to find relevant features with fault. Furthermore, the fault diagnosis model combines the proposed feature selection method with support vector machine. Experimental results on a chemical process indicate that the proposed feature selection method is useful and superior in fault diagnosis.