Information Technology Reference
In-Depth Information
for the roughing process is in average (but not always) smaller than the finishing time. Thus
a slab can be output by the roughing mill while the rolling of the previous coil is being
completed: this fact may cause a collision or may force the second slab to remain stuck while
its temperature decreases, which makes its successive rolling more difficult. On the other
hand, the time between the input of two successive slabs to the roughing time cannot be
excessive, in order to keep productivity and avoid energy losses. Ideally, a slab should be
input to roughing mill exactly at a time instant that will allow it to arrive at the entrance of
the finishing mill when the rolling of the previous coil is just terminated. (Colla et al. 2010)
applied various neural networks-based approaches to predict the time
i
(
1≤i≤6
) required by
the slab to pass each one of the 6 stages that form the roughing mill (see Fig. 3).
Fig. 3. Scheme of the first stages of a steel hot rolling mill.
In particular, the most successful solution performs a sequential prediction, namely bases
the prediction of
i
(for
i>1
) not only on product and process parameters, but also on the
prediction of the times required to pass the previous stages, i.e.
k
with
1≤k≤i
. Moreover,
neural networks have been applied also to predict the time required for passing the
finishing mill.
In this case, the application of neural networks were actually advantageous for the following main
reasons: i) neural networks proved to outperform more traditional approaches; ii) the neural system is
naturally adaptable to the changing operating conditions thanks to its capability of self-learning from
data. However the on-site real-time implementation has not been easy and required considerable
efforts because it has been difficult to interface the system with the control system of the mill.
4.5 Estimate of train position and speed from wheels velocity measurements
Within an Automatic Train Protection (ATP) system, two subsystems are usually included: a
ground subsystem, which provides updated information on the train position and the line
gradient by exploiting fixed balises or another source of absolute information (e.g. GPS), and
an on-board subsystem, which estimates the actual train position and speed, according to
the scheme depicted in Fig.4.
The ground subsystem communicates to the on-board one the distance from the next
reference point on the line, the gradient of the line and the allowed speed at the next
reference point. The on-board subsystem then evaluates the distances from the next
information point and the minimum distance that allow compliance with the speed limit at
the next reference point. If it turns out that the train cannot meet the target speed at the next
reference point (as the residual breaking resources of the train are not sufficient), the on-
