Magdalena Hajdasz
Institute of Building Constructions,
University of Technology, Poznan, Poland
Adam Marlewski
Institute of Mathematics,
Poznan University of Technology, Poland
Extended abstract
Obviously, no expert system can correctly work if its knowledge base
and the inference machine are too poor. On the other hand, a very
wide base and too many rules make the system slower produces result
even if clever heuristics are implemented.
This statement has to be taken into account when we intend to
construct an intelligent program which can be helpful for the manager
who works on a site where a silo is under construction.
Since the very beginnig of the erecting process the manager has to
determine the technology and organisation of the actions concerning
the realization of this very complex process. It means
that he has to take into account all the components involved in this
process: concrete mixers, buckets, slip, cranes, working staff,
atmospherical conditions, perturbations (caused by various failures
of the equipment). There can be arranged various technological
systems and each one of these systems can be evaluated with respect to
different criteria (e.g. maximum efficency, minimal cost of
realisation, minimum losses, maximum balancing).
Naturally, the final result depends on the art of design, the
machinery park and human staff being on disposal. In the considered
case a crane is, in general, the most expensive machine and it makes
that it places on a crucial position in the technological process.
That's why the parts of knowledge and data bases concerning cranes
should be prepared with a special care. Unfortunately, the complexity
of the cranes themselves causes that the data on them are not
supplied in the form which can automatically adapted for the
requirements of a computer program. It concerns, for example, the
behaviour of the crane arm. A knowledge engineer has to know which
is the way followed by the crane arm and how the loading capacity
changes as the operational altitude and horizontal distance vary. The
data on these features are supplied in the form of tables and drafts,
so in the form which is not preferable for computers. Here is the
moment where we see that computer algebra systems can be applied. In
particular, we think here of DERIVE (from Soft Warehouse Inc.,
Hawaii) which is relatively easy to be dominated also by persons
having small experence with computers. In the presentation we want to
show the characteristics of some types of cranes (producing in
Poland, Germany and United Kingdom) and to how to obtain the
equations of curves describing the movement of crane arm and its
loading capacity. In general, the problem reduces to the
determination of the best fitting, so one has to decide in which
approximation class it should be done. Considered cases reveal that
there is no simple advice how to choose this class. Moreover, in some
cases the standard techniques (e.g. the use of the DERIVE built-in
function FIT) comes to unsatisfactory approximations and then a
special approach has to be applied (in our example it is the rational
collocation; we supply it as the function 
).
The schema of the passing from the traditional data to the form
accepted by a computer is shown in the attached figure. Authors'
intention was to elaborate the expert system which produces the
outcome in the user-friendly form which does not stray away from the
traditional one (see Box 8 in Figure).