LiU Homepage
Abstract |
Distributed Diagnosis and Simulation Based Residual Generators
Fault diagnosis is becoming increasingly important for many technical
systems. This is for example true in automotive vehicles where fault
diagnosis is needed due to economic reasons such as efficient repair
and fault prevention, and legislations that mainly deal with safety
and pollution. The objective for a diagnostic system is to detect and
isolate faults in the system. A diagnostic system consists of several
specialized parts, for example residual generators, diagnoses
calculation, and communication with other systems.
In embedded systems with dozens of electronic control units that
individually states local diagnoses, it can be computationally
expensive to find which combination of local diagnoses that points at
the correct set of faulty components. A distributed method is
proposed where local diagnoses are extended using networked
information. The extension is done thru the sharing of local conflicts
or local diagnoses between the electronic control units. The number
of global diagnoses grows with the number of local diagnoses.
Therefore, an algorithm is presented that from the local diagnoses
calculates the more likely global diagnoses. This restriction to the
more likely diagnoses is sometimes appropriate since there are
limitations in processing power, memory, and network capacity.
A common approach to design diagnostic systems is to use residual
generators, where each residual generator is sensitive to some faults.
A method is presented that constructs residual generators from sets of
overdetermined model equations, such that simulation can be used to
determine if the residual is zero or not. The method thus avoids the
need to analytically transform the set of equations into some specific
residual generator form. It can also utilize smaller sub sets of
equations like minimally overdetermined sets, and it can further take
advantage of object-oriented simulation tools.
Jonas Biteus
2005
Informationsansvarig: webmaster
Senast uppdaterad: 2021-11-10
