VI1 THE EXPERIMENTAL RESULTS & DISCUSSIONS
The hybrid vector drive was tested using a speed
command that follows a trapezoidal trajectory. The speed
command ramps up from zero speed to reach a desired
speed in a fixed time interval. It then stays at that speed for a
predetermined period. After that, it ramps down at a fixed
slope. The hybrid drive switches to the neural network
model during the ramp up and the ramp down period. It uses
the fuzzy control model during the constant speed operation
period. The experimental results are shown in fig. 9.
The experimental results shows that the hybrid drive is
able to track the command speeds very closely in all regions
of operation. The lag time during the ramp up and ramp
down regions are much less as compared to a pure fuzzy
logic model shown in fig. 10. Under constant speed
operations, if any disturbance occurs, a neural network
based controller would have to train the network to cancel
the effects of the disturbance. This would be slower than the
fuzzy logic based model, which immediately computes a
reasonably efficient output to combat the disturbance.
Another reason why the neural network model is not used in
this region is that the disturbance will be random and it is
not useful to train the neural network to overcome the
effects of any particular disturbance which may only occurs
once in a blue moon.
Fig. 11 shows the robustness of the hybrid drive towards .
dips in dc-linked voltage. When the motor has reached a
constant speed of 400 rpm, the dc-linked voltage is reduced
gently from 150 V to 60 V at t = 20 sec. The experimental
results only show a very small ripple in the motor speed and
hardly any noticeable variation.
In fig. 12 shows, when the motor has reached a constant
speed of 400 rpm, the dc-linked voltage is reduced gently
from 150 V to60 V at t =20 sec andretumedto 150 V at t =
40 s. The experimental results only show a very small ripple
in the motor speed and hardly any noticeable variatio