作者: Zengbing Xu , Jianping Xuan , Tielin Shi , Bo Wu , Youmin Hu
DOI: 10.1016/J.ESWA.2009.01.063
关键词: Fuzzy logic 、 Bootstrapping 、 Artificial intelligence 、 Feature (computer vision) 、 Bearing (mechanical) 、 Computer science 、 Euclidean distance 、 Network model 、 Stability (learning theory) 、 Pattern recognition 、 Fault (power engineering)
摘要: Considering different importance of the feature parameters to fault conditions bearing, a modified fuzzy ARTMAP (FAM) network model based on feature-weight learning is presented in this paper. The features time-domain, frequency-domain and wavelet-domain are extracted from vibration signals characterize information relevant bearing. By improved distance evaluation technique optimal selected corresponding feature-weights which assigned indicate their bearing obtained. Then they combined with FAM described by weighted Manhattan applied seven-class diagnosis To assess effectiveness stability network, bootstrapping method employed quantify performance statistically. Diagnosis results show that can more reliably accurately recognize classes.