Fault Detection with CP-Decomposed Qualitative Models

作     者：Müller, Thorsten Lichtenberg, Gerwald 

作者机构：Fraunhofer Institute for Solar Energy Systems - ISE Freiburg79110 Germany Hamburg University of Applied Sciences Hamburg21033 Germany 

出 版 物：《IFAC-PapersOnLine》 

年 卷 期：2016年第49卷第5期

页      面：309-314页

核心收录：

基　　金：284920 

主　　题：Fault detection Automata theory Digital control systems Discrete time control systems Signal detection Stochastic models Stochastic systems Tensors Discrete time systems Exponential growth Fault detection algorithm Qualitative model Real measurements Stochastic Automata Tensor decomposition Tensor structures 

摘      要：The paper shows how a fault detection algorithm based on stochastic automata as qualitative model can be improved by non-negative CP tensor decomposition to make it applicable to large discrete-time systems. Because exponential growth of the number of transitions of the automaton with a rising number of states, inputs and outputs of the system can usually not be avoided, tensor decomposition methods enable the reduction of the amount of data to be stored by an order of magnitude. In order to exploit the full potential of the decomposition, a fault detection algorithm that is applicable to the decomposed tensor structure is defined. An example based on real measurement data shows the functionality of the algorithm. © 2016