摘要：In this study, two clustering algorithms and their success in fault isolation have been investigated in order to use in our fault tolerant control (FTC) system. With so many applications used today, the mathematical model of the system cannot be completely established. Therefore, in this study, fault detection and isolation (FDI) is realized by using knowledg1e-based methods, without the need for any mathematical model. Sensor data, which are taken offline by FDI, are clustered to create knowledge base by means of k-means and farthest first traversal algorithm (FFTA), respectively. The results obtained by the two algorithms are compared and FFTA has found to be more successful in fault tolerance.
其他摘要：Bu çalışmada, geliştirmiş olduğumuz bir arıza dayanımlı denetim (ADD) sisteminde kullanmak üzere iki farklı kümeleme algoritması incelenmiş ve arıza tanılama başarımları araştırılmıştır. Günümüzde kullanılmakta olan birçok uygulamada, sistemin matematikse
关键词：Fault tolerant control;fault detection and identification;k-means;three tank;data mining
其他关键词：Arıza dayanımlı denetim;arıza tespit ve tanılama;k-means;üçlü tank;veri madenciliği
