Search

The identifiability (structure detectability and consistent parameter estimation) of linear multivariable stochastic system from the "joint process" statistics for general closed-loop characteristics is considered. Necessary and sufficient conditions for the identifiability of classes of systems are proven. Experimental tests to verify the identifiability and the use of dithers which guarantee it when the feedback is unknown are discussed.

In the identification of linear invariant systems the optimum choice of order and structure of a state variable model representation is strictly dependent on the level of uncertainty present in the system and on the identification experiment. This concept is formalized here by applying re cent results found in numerical analysis literature.

This paper presents a package for modeling and identification of linear multivariable systems from only the knowledge of input - output data. The method, that is original, belongs to the least squares family and applies new efficient factorization algorithms drawn from the numerical linear algebra literature. Performances of the package and comparisons with other methods on a simulated real size power plant are illustrated.

The paper is addressed to the approximation in norm of stochastic processes of arbitrary dimension by reduced order models. After a systematic review of the basic geometry of stationary process representation in Hilbert space, the approximation problem is developed along the following items:1)Using Hankel norm, a closed solution of the approximation problem is available in the literature, based on singular value decomposition. This offers the opportunity to compare diffe rent approaches based on Hankel matrix decomposition: i) factorization of the covariance Hankel matrix, proposed by Rissanen and Kailath and Faurre; ii) canonical correlation analy_ sis, recently discussed by Akaike.2)By singular value decomposition, the internal energetic structure of the process distinct from the canonical algebraic structure, is also elucidated.3)Often the process model is needed for prediction; this suggests an interesting extension to optimum predictor approximation.

The planning problem in batch and repetitive manufacturing environments is approached from the Large Scale Systems point of view. The optimal production schedule, according to a given cost function, is sought for a sequence of workshops constituting a multistage production process. The workshops themselves may be Flexible Manufacturing Systems able to process different product types at the same time.

This paper presents the outline of an environment for the specification and prototyping of manufacturing systems software, which is under development in the framework of a research project sponsored by Italian National Research Council.

The hardware-software architecture of the MODIAC system is described. MODIAC is a process control computer which represents the result of a joint research carried out by many Italian industries and university institutions within the framework of the Computer Science Program supported by the National Research Council (CNR). The system is based on a local area network interconnecting mono- or multiprocessor nodes. Token Passing or CSMA/CD strategies can be adopted for the exchange of the messages among the nodes. All the nodes are implemented on a high-performance bus, whose characteristics enable the use of 16-bit microprocessors of different types. A special unit is provided with a powerful programming environment, thus allowing the development of advanced control algorithms. Two application cases, belonging to quite different automation areas, are presented to enhance the system flexibility.

The best fitting problem, i.e. the problem of transferring a rigid body in the Euclidean space as close as possible to a given target is approached and solved in this paper.