25.07.2014 Views

pdf download - Software and Computer Technology - TU Delft

pdf download - Software and Computer Technology - TU Delft

pdf download - Software and Computer Technology - TU Delft

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

4.4 Modeling Model-Based Fault Diagnosis<br />

//input is assumed to be correct<br />

correct_w = true;<br />

}<br />

// Connect the 3 inverters<br />

invA ( correct_w, hA, correct_x);<br />

invB ( correct_x, hB, correct_y);<br />

invC ( correct_x, hC, correct_z);<br />

The user of the diagnostic system should be able to determine whether a specific output is<br />

correct or not. These are the observables.<br />

Weak Fault Model This model only defines the nominal behavior of the system. The description<br />

of the 3-inverters in Section 4.2 is a weak model; it does not define how an inverter behaves if<br />

something has been broken (if h=0).<br />

Strong Fault Model Description of the system that defines all modes of operation. A mode of<br />

operation is a state of a component in which it obeys an unique behavioral rule. The nominal<br />

behavior of a weak model could specify more than one nominal modes of operation. A strong<br />

fault model also defines all known false modes of operation. The corresponding LYDIA code<br />

follows the form !h => s. Examples are stuck-at-zero, stuck-at-one, etc. This way, it is<br />

possible in MBD to define an abductive model [18], as discussed in Chapter 3. A strong<br />

model of the 3-inverters is:<br />

system inverter(bool i, h, o) {<br />

h => (i = !o); // If healthy, output equals inverse of the input<br />

!h => (o = 0); // If unhealthy, the output is stuck-at-zero<br />

}<br />

Model that is not from First Principles It might be that it is practically impossible, for certain<br />

parts of the system, to define the correct behavioral rules. In these cases it is possible to<br />

include mappings of symptoms on broken components (of the form ’s → f ’). MBD diagnosis<br />

does not forbid to use these explicitly specified statements, that are abductively or deductively<br />

derived by humans. A (partial) model that is not based on first principles, <strong>and</strong> that is equal to<br />

the weak model of the 3-inverter example is:<br />

// mapping of symptom (y=0, z=1) on diagnosis<br />

((y=0) <strong>and</strong> (z=1)) => (<br />

!hB<br />

or (!hA <strong>and</strong> !hC)<br />

)<br />

In theory, the expressive power of rules that are not based on first principles is the same. After<br />

all, the rules are equivalent to a consistency-based model (’s → f ’ corresponds to ’¬ f → ¬s’).<br />

40

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!