IDENTIFICATION OF DIAGNOSTIC STATISTICAL STATE MODELS FOR NETWORKED SYSTEM

Miika Rajala, Risto Ritala

Abstract

The diagnostics of networked systems has concentrated on diagnostics of network nodes. We present a statistical modelling method which supports analysis and diagnostics of the network as whole, in particular, of coher- ent states and state transitions. We describe nodes with dis- crete states statistically affected by external load. In a net- work the effective load on a node is a sum of the external load and an effect due to the states of neighbouring nodes. Such effective loads introduce statistical dependence be- tween states, and eventually coherent behaviour of the net- work. This presentation describes how such models are identified from data and discusses how the models support network-wide diagnostics.

Keywords

networked systems, statistical models, Ising model, diagnostics

Download

IMEKO-TC10-2005-31.pdf

DOI

-

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings