529
3.5 Discussion
It might appear that the final selection of signals will bring some redundant data. There are
two possible ways: we can reduce the number of signals via selection of one of them or “fuse”
data by simple aggregation to increase the power of recognition of hidden pattern. At this
stage, there is no simple answer what option is better. From an implementation point of view,
fewer variables are better, however, data fusion is commonly used to increase the robustness
Figure 3. A procedure for variable selection.
Figure 4. Normalized Autocorrelation functions of all signals.
0.9
0.8
0.7
0.6
8 0.5
0.4
0.3
0.2
0.1
0
-1
Remove samples for
speed>O
Remove samples for
gear <>0
Remove samples for fuel
consumption <10%
Remove zero energy
signals
During unloading
should not drive
Du ring unloading
should park
Du ring unloading
should use fuel
Autocorrelation for the given signal
-0.5
0
lag
0 .5
-
BREAKP
-
ENGCOOLT
-
ENGEXB
-
ENGHOURS
-
ENGOILP
-
ENGRPM
-
ENGTPS
-
FUELUS
-
GAOILP
-
GAOILT
-
HYDDRV
-
HYDOILP
-
HYDOILT
-
INTAKEP
-
INTAKET
-
SELGEAR
-
SPEED
-
SWITCH MOVE
-
TRNAUT
-
TRNBPS
-
TRNLUP