Submitted by tkujala on November 9, 2007 - 10:15
HIIT seminars in fall 2007 will be held in hall **B222** of Exactum, on Fridays starting at 10:15 a.m. Coffee available from 10.
Fri Nov 9
Anna Pernestål
KTH School of Electrical Engineering, Automatic Control
Bayesian fault diagnosis -- utilizing data and prior information
Abstract:
Fault diagnosis is the task of detecting if there are faults present in a system, and if so determine which. Due to noise, limited information, and model errors it is often difficult, or even impossible, to exactly point out the faults that are present. Hence, we are forced to reason under uncertainty.
In this presentation we show how Bayesian methods can be applied to perform fault diagnosis. In particular we discuss how training data can be used for learning. One challenge is that the amount of data available is often limited. Therefore we also discuss how to use prior information: which information that may be available, and how it could be integrated in the Bayesian framework for learning.
The methods are applied to the diagnosis of an automotive diesel engine, and we will present some preliminary results, based on authentic data from real driving situations.
Fri Nov 9
Anna Pernestål
KTH School of Electrical Engineering, Automatic Control
Bayesian fault diagnosis -- utilizing data and prior information
Abstract:
Fault diagnosis is the task of detecting if there are faults present in a system, and if so determine which. Due to noise, limited information, and model errors it is often difficult, or even impossible, to exactly point out the faults that are present. Hence, we are forced to reason under uncertainty.
In this presentation we show how Bayesian methods can be applied to perform fault diagnosis. In particular we discuss how training data can be used for learning. One challenge is that the amount of data available is often limited. Therefore we also discuss how to use prior information: which information that may be available, and how it could be integrated in the Bayesian framework for learning.
The methods are applied to the diagnosis of an automotive diesel engine, and we will present some preliminary results, based on authentic data from real driving situations.
Events:
Last updated on 8 May 2008 by Martti Mäntylä - Page created on 9 Nov 2007 by Teija Kujala