Model Based Diagnosis
Contact person: Erik Frisk
The diagnosis problem typically consists of detecting and isolating
faulty components given available sensor and actuator
signals. Reliable real-time multiple fault diagnosis of dynamical
systems in the presence of noise is of fundamental importance in many
industrial applications, but is generally associated with a high
The FlexDX diagnosis framework is designed to reduce the computational burden through adaptive reconfiguration, by exploiting the fact that only a subset of all possible tests are required at a specific time instant to detect the faults of interest. The FlexDX framework is illustrated in the figure below.
The basic idea is that in case of an alarm, additional tests can be run, partially on historic sensor
information, in order to further isolate the faulty components. However, special attention has to be given to the issues
introduced by reconfigurability. For example, tests are added and removed dynamically, tests are partially performed on historic data, and synchronous and asynchronous processing are combined. Based on these principles, a diagnosis framework called FlexDX, building on the DyKnow knowledge processing middleware, was developed through a cooperation between the computer science and electrical engineering departments. The work started in MOVIII and has been further improved in this project.
Quantitative Analysis and Test Selection
In a continuation of the ideas explored within the FlexDX project, the
approach to select which new tests to start next is studied. In the
FlexDX project, only qualitative properties were considered and a next
step is to research for quantitative properties that makes it possible
to optimize the approach further. Thus, the challenging question on
how to quantify diagnosability performance is studied in this
project. Such measures has fundamental implications on many aspects of
diagnosis systems design and optimization. In this project, a measure
called distinguishability has been developed for quantifying how
difficult it is to detect and isolate faults. This measure uses a
stochastic characterization of the different fault modes to quantify
diagnosability performance and can be used in test design to achieve
maximum fault to noise ratio. We are currently integrating these
optimal tests into a reconfigurable diagnosis system to achieve high
distinguishability for the complete diagnosis system. Finally, we have
also used the distinguishability measure for evaluating the diagnosis
potential of different sensor configurations.
Mattias Krysander, Fredrik Heintz, Jacob Roll, and Erik Frisk. FlexDx: A Reconfigurable Diagnosis Framework. Engineering Applications of Artificial Intelligence, 23(8):1303--1313, 2010.
Erik Frisk, Anibal Bregon, Jan Åslund, Mattias Krysander, Belarmino Pulido, and Gautam Biswas. Diagnosability Analysis Considering Causal Interpretations for Differential Constraints. IEEE Transactions on Systems, Man, and Cybernetics -- Part A: Systems and Humans, 42(5):1216--1229, 2012.
Carl Svärd, Mattias Nyberg, Erik Frisk, and Mattias Krysander. Automotive Engine FDI by Application of an Automated Model-Based and Data-Driven Design Methodology. Control Engineering Practice, 21(4):455--472, 2013.
Daniel Eriksson, Erik Frisk, and Mattias Krysander. A method for quantitative fault diagnosability analysis of stochastic linear descriptor models. Automatica, (accepted for publication) 2013.