Academic Journals Database
Disseminating quality controlled scientific knowledge

An Adaptive Clustering Procedure with Applications to Fault Detection

Author(s): Milan R. Rapaić | Milena Petković | Zoran D. Jeličić | Alessandro Pisano

Journal: Electronics
ISSN 1450-5843

Volume: 15;
Issue: 2;
Start page: 91;
Date: 2011;
VIEW PDF   PDF DOWNLOAD PDF   Download PDF Original page

Keywords: Classification | Clustering | Novelty identification | Fault detection and isolation (FDI) | Condition monitoring

A novel adaptive clustering procedure is presented in this paper. Among the basic properties of the proposed algorithm is that the number of clusters is not known a priori, it is updated automatically based on the available data. Previous knowledge regarding data source (data generating process), if available, can be used for initialization purposes. However, the algorithm can be used even if such information is not available. The entire data set need not be known in advance, and further, the algorithm does not store previously seen data points. The computation complexity is relatively low and the entire procedure may be implemented recursively (in “real-time”). The proposed procedure is designed primarily for condition monitoring andfault detection in industrial plants. Performances of the proposed algorithm have been demonstrated by an illustrative example.
RPA Switzerland

RPA Switzerland

Robotic process automation


Tango Rapperswil
Tango Rapperswil