|
 |

|



Research interests in the modelling and identification field started at the
Engineering School of Bologna University in 1970. The researches performed
in subsequent years have led to the development of procedures and concepts
that have been constantly applied to industrial, natural and economic
processes to evaluate their value as practical tools for engineers.
Some of these applications have concerned:


CSTR reactors for rubber production
(Centro Ricerche Montedison, Ferrara)

Benzene-toluene distillation columns
(CLEDCA-ITALGAS, Fidenza)

Power plant of Pont-sur-Sambre,
Francia (IFAC test case)

Isobuthane-normalbuthane distillation
columns (ANIC, Milano)

Natural gas fields of the Po Valley
(Agip Mineraria, Milano)

Fiberglass production furnaces
(DATAMONT, Milano)

Phosphorus kinetics in fresh waters
(Canada Centre for Inland Waters, Burlington, Canada)

Gasoline consumption on a free market,
(Agip, Roma)

Natural gas reservoirs (Agip
Mineraria, Milano)

Ethane-ethylene distillation columns
(DATAMONT, Milano)

Oil fields (Istituto di Scienze
Minerarie, Università di Bologna)

Primary heat exchangers in a fast
reactor (ENEA, Bologna)

Eutrophication phenomena in the
lakes Erie, Ontario and in the Adriatic Sea (Canada Centre for Inland
Waters, Burlington, Canada and CRISM, Bologna)

Dynamic models for "mini" fuel
injectors (WEBER, Bologna)

Many of these applications have been performed in the context of the
Electronic Engineering curricula or in the context of the
Doctorate curricula on System Engineering at Bologna University.

Present research activities of the team are oriented toward the analysis
of several identification and estimation techniques whose applications
are often described in the literature. Many of these schemes are based on
assumptions which are not verifiable on the basis of the
available input-output sequences.

The Frisch scheme, which does not introduce any a priori hypothesis
on the amount of the additive noises affecting the input-output sequences
of the system, belongs to this context and relies on mathematical descriptions
in the family of the "errors-in-variables" models.
It treats all variables in a symmetric way and leads to the estimation of a
whole set of models compatible with the data.

The extension of the Frisch scheme, originally proposed for algebraic systems,
to the identification of dynamic processes has shown that it is possible,
in this case, to determine a single solution corresponding to the exact model
of the system and of the noise.
In the identification of real systems, however, the assumptions behind
the scheme are never exactly fulfilled and the Frisch scheme leads to
a whole family of possible solutions; the selection of a single model
is possible only introducing criteria based on suitable cost functions.

The most relevant aspects of the researches currently carried out are the
following:

Comparison between the results obtained
by means of errors-in-variables procedures and those obtained by other
methods;

Extension of the Frisch scheme to
multivariable processes;

Development of general errors-in-variables
identification procedures assuming non-white and/or correlated noises;

Computation of the maximal corank of a
covariance matrix in the context of the Frisch scheme;

Frequency-domain approaches for
errors-in-variables models identification;

Stochastic realization for interconnected
systems;

Fault diagnosis in dynamic systems;

Application of identification
procedures to real processes;

Application of advanced tools in
higher education;

Blind identification and equalization;

Fault detection and isolation;

Errors-in-variables filtering and extensions of Kalman filtering;

Identification of noisy ARX and AR models.
|




|
|
|