pdf download - Software and Computer Technology - TU Delft
pdf download - Software and Computer Technology - TU Delft
pdf download - Software and Computer Technology - TU Delft
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
5.6 MBD Implementation 3<br />
Diagnosing the Beam Propeller Movement<br />
of the Frontal St<strong>and</strong><br />
Y<br />
Observations<br />
Diagnosis<br />
e pos CS P.V. POS. SP. CUR. Possible single faults<br />
Y0 0 0 0 0 0 None: entire system healthy<br />
Y1 0 0 0 0 1 !h_EXT, !h_MVR, !h_MBU<br />
Y2 0 0 0 1 0 !h_EXT, !h_MVR, !h_MBU, !h_St<strong>and</strong><br />
Y3 0 0 0 1 1 !h_EXT, !h_MVR, !h_MBU<br />
Y4 1 0 0 1 0 0 !h_EXT, !h_MVR, !h_MBU, !h_St<strong>and</strong>, !h_PEU<br />
Y5 0 0 0 1 0 0 !h_EXT, !h_PEU<br />
Y6 0 0 1 0 1 !h_EXT, !h_MVR, !h_MBU<br />
Y7 0 0 1 1 0 !h_EXT, !h_MVR, !h_MBU, !h_St<strong>and</strong><br />
Y8 0 0 1 1 1 !h_EXT, !h_MVR, !h_MBU<br />
Y9 0 1 0 0 0 !h_PEU<br />
Y10 0 1 0 0 1 None: only multiple faults<br />
Y11 0 1 0 1 0 None: only multiple faults<br />
Y12 0 1 0 1 1 None: only multiple faults<br />
Y13 0 1 1 0 0 !h_PEU<br />
Y14 0 1 1 0 1 None: only multiple faults<br />
Y15 0 1 1 1 0 None: only multiple faults<br />
Y16 0 1 1 1 1 None: only multiple faults<br />
Y17 1 !h_EXT<br />
Table 5.7: Partial diagnosis Look-Up Table of the Beam Propeller Movement for single faults<br />
LUC_Extension has highest probability of being at false. Thus, the fact that LUC_Extension is<br />
at false is on the top of the list of both implementations. This implies that the accuracy is equal<br />
for both implementations. As a result, the diagnostic performance does not increase if MBD-2 is<br />
implemented. The benefits of the entropy calculation are clear: we humans would have chosen to<br />
implement MBD-2 instead of MBD-1. As a consequence, needless costs would be spend for the<br />
development of the tool that derives the value of ctr_speed. The entropy calculation predicts that<br />
the accuracy will not increase for future fault scenarios, <strong>and</strong> saves the higher development costs of<br />
MBD-2.<br />
5.6 MBD Implementation 3<br />
This section presents how to use entropy to determine which additional measurements on the system<br />
can increase the diagnostic accuracy the most. Adding measurements means increasing the<br />
observation space. The observation space can be divided in a spatial dimension <strong>and</strong> a temporal<br />
dimension [21]. The spatial dimension refers to the number of observation variables (sensors). The<br />
temporal dimension refers to the number of samples per observation variable in time. Here, the<br />
spatial dimension of the beam propeller movement system is examined.<br />
Figure 5.3 shows all components, observables <strong>and</strong> internal variables with respect to the beam<br />
propeller movement. The observation space of the model used for implementation MBD-2 is determined<br />
by all permutations of the following observables: CURRENT_ERROR, SPEED_ERROR, e_pos,<br />
POSITION_ERROR, POSVAL_ERROR <strong>and</strong> ctr_speed. It is interesting to known what variables shown<br />
in the block diagram of Figure 5.3 are best to add to this observation space. In other words, an<br />
engineer is interested what measurements of the target system improve the accuracy of the MBD<br />
outcome the most. This is interesting because there are costs for each measurement that is added to<br />
the observation space. For example, adding the variable real_speed requires that a sensor is added<br />
to the target system, the sensor signal is being logged, possibly discretized, <strong>and</strong> finally inserted<br />
64