Search

Substructuring methods are a very common and efficient way to solve complex vibroacoustic problems. The Condensed Transfer Function (CTF) approach is a substructuring method based on the concept of subsystem condensed transfer functions (corresponding to admittances or impedances) that allows assembling acoustical or mechanical subsystems coupled along lines or surfaces. For certain practical applications, it may be more efficient to subtract or to decouple a subsystem to a global system rather than assembling different subsystems. In this paper, a reverse formulation of the CTF method is proposed. This formulation allows us to predict the behavior of a subsystem that is part of a larger system, from the knowledge of the condensed transfer functions (CTFs) of the global system and of the residual subsystem that must be removed. For purposes of validation, the scattering problem of a rigid sphere in an infinite water domain impacted by an acoustic plane wave is considered. Comparisons with theoretical calculations are used to validate the formulation proposed and permit studying its accuracy for two types of condensation functions defining the CTFs.

Past initiatives with respect to realising control libraries are analyzed. A new collaborative initiative is introduced. The design of a new control library SLICOT is presented and the first release of this library is discussed

The field of Computer Aided Design in Control Systems has become a major topic of research in modern control engineering. Current CADCS packages cover the whole spectrum of classical and modern control theory and are being continuously updated with the latest advances in numerical control algorithms. As a result todays control engineers have access to many powerful and advanced control techniques to meet the still increasing demands from the industry. However the actual implementation of these modern techniques is still very much limited, especially in time-critical on-line situations, by the physical limits of todays sequential processors. The arrival of new computer architectures and technologies, such as vector computers and parallel processing, can offer a spectacular breakthrough to the current computing limitations. In this paper it is demonstrated that a considerable speedup is obtained when vector processing is applied to two classical but powerful long-range predictive control techniques; namely Model Algorithmic Control and Dynamic Matrix Control.

The use of parallel processing techniques offers interesting perspectives for the application of advanced control algorithms in time-critical situations. In this paper the parallelization of some long-range predictive control (LRPC) algorithms is examined. A study is made of the speed-up that can be obtained through proper parallelization of the underlying mathematical matrix routines. Two LRPC-methods are applied to the position control of a DC-servomotor. >

In system identification we are often confronted with non-stationary timeseries. In the area of econometrics the method of cointegration analysis has become the most popular way to deal with models involving nonstationary data. The main disadvantage of the methodology based on cointegration is that it is basically a linear method and hence it is only capable to find linear relationships. Therefore. in this paper we propose a new method for nonlinear multivariate cointegration analysis.

Currently, multilayer feedforward neural networks with backpropagation learning rule are the most popular artificial neural networks for handling optimisation and learning problems. This success is largely based on the simplicity to implement these networks. However, there are some major drawbacks: there is no guarantee for training in finite time and there is no guarantee that the optimal solution will be reached. In the sequel, the application of genetic algorithms as learning rule (thus, as optimiser) is compared to the error backpropagation learning rule (i.e., gradient descent optimiser).

In this paper two CACSD-packages for the analysis and synthesis of multivariable linear control systems are presented. Their theoretical background and specific implementation are discussed. The main characteristics of these programs are their user-friendliness and robustness. The packages can be applied in an educational environment as well as in an industrial context.

During the last years a number of interesting control methods were published all having the same common idea : the control action is based on a long-range prediction of the process output. Four of these methods are compared in this paper. At least for three of them industrial applications are known, indicating that long-range predictive control methods are likely to be a useful tool for real-life applications. These adaptive control methods are compared by means of a simulation study. Special attention is paid to their robustness w.r.t. unmodelled dynamics, parameter variations, process noise and changing dead-time. Also the number of tuning parameters that remain, as well as their efficiency and interpretation to the user are dealt with.

In this paper two CACSD-packages for the analysis and synthesis of multivariable linear control systems are presented. Their theoretical background and specific implementation are discussed. The main characteristics of these programs are their user-friendliness and robustness. The packages can be applied in an educational environment as well as in an industrial context.