pdf download - Software and Computer Technology - TU Delft
pdf download - Software and Computer Technology - TU Delft
pdf download - Software and Computer Technology - TU Delft
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Automated Fault Diagnosis<br />
3.1 Overview Techniques<br />
by a sequence of states [4]. The structure is what enables the system to generate the behavior [4].<br />
One of the ideas in system theory is that the structure is as important for implementing the function<br />
of a system as the individual behavior of components. The reason for this is that the interaction<br />
between individual components results in the intended behavior of the whole. These relations between<br />
structure <strong>and</strong> behavior are important to underst<strong>and</strong>, because when creating an approach to<br />
fault diagnosis, various viewpoints on these concepts can be used. Two very important viewpoint<br />
types are black box <strong>and</strong> white box. The definitions of these concepts that are used in this thesis are:<br />
Black Box: A model that describes the externally observable behavior, but does not state anything<br />
about the structure of the system, or behavior of internal components of the system.<br />
White Box: A model that describes the internally non-observable behavior as well as externally<br />
observable behavior. The behavior of the whole system is defined by the structure <strong>and</strong> the<br />
behavior of internal components.<br />
Many views on a system include more information than just system input <strong>and</strong> output, but do not<br />
provide a complete behavioral, or structural description of the system. These views are sometimes<br />
referred to as grey boxes, but terms like these are not well established. The problem with these<br />
terms is that it is not clear what is considered to be a ’complete’ white box, because a model is<br />
never complete. For this reason, in this thesis a model is called white box whenever it uses some<br />
structural or internal behavioral information. Otherwise, it is a black box model.<br />
3.1.4 Abductive Model versus Consistency-based Models<br />
The distinction between black box <strong>and</strong> white box models determines what information is used in<br />
the model. Another distinction determines how the information is stated. A distinction between a<br />
consistency-based <strong>and</strong> an abductive approach to fault diagnosis [18], is as follows:<br />
Abductive model: A model that defines a diagnosis as a set of abnormality assumptions that covers<br />
(or, in terms of logic, implies) the observations. [18, 24]<br />
Consistency-based model: A model that defines a diagnosis as a set of assumptions about a system<br />
component’s abnormal behavior such that observations of one component’s misbehavior are<br />
consistent with the assumption that all the other components are acting correctly [11, 25].<br />
The important difference between the two is that the former, an abductive model, defines effectto-cause<br />
relations between observables, while that latter defines cause-to-effect relations between<br />
observables (also called first principles). This requires different kinds of reasoning schemes.<br />
3.1.5 Effect-to-Cause versus Cause-to-Effect Reasoning<br />
In effect-to-cause reasoning, the effect ”there is no light” can be explained by the cause ”the light<br />
bulb is broken”, knowing that that when a light bulb is broken there is no light. In cause-to-effect<br />
reasoning, the fact ”the light bulb is broken” is the only cause that is consistent with the effect ”there<br />
is no light”. Let s be a symptom <strong>and</strong> f be a fault (e.g., the fault f denotes the cause ”the light bulb is<br />
broken”, <strong>and</strong> the symptom s denotes the effect ”there is no light”). In logic, the basic inference rule<br />
of abductive reasoning can be characterized by the following reasoning pattern [8]:<br />
s<br />
f → s<br />
⊢ f (3.1)<br />
21