ANNALS OF THE UNIVERSITY OF CRAIOVA

Georgeta Budura

Abstract: Nonlinear adaptive filtering techniques are widely used for the nonlinearities identification in manny applications. This paper investigates the performances of the Volterra estimator by considering a nonlinear system identification application. The Volterra estimator parameters are compared with those of a linear estimator. For the nonlinear estimator, based on a second order RLS Volterra filter, a new implementation is proposed. The experimental results show that the proposed Volterra identifier has better performances than the linear one. Different degrees of nonlinearity for the nonlinear system are considered.
Keywords: Recursive least squares algorithm, Nonlinear estimator, Linear estimator, Volterra filter.

Petr Chalupa, Vladimír Bobál, Petr Dostál

Abstract: The paper deals with decentralized control with logical supervisor. The decentralized control is an approach of controlling multi input multi output systems with the same number of inputs as the number of outputs. The logical supervisor was proposed as a possible way of improving the stability and the quality of control courses. The theoretical results are applied to a real-time laboratory control system – DTS200 Three Tank System.
Keywords: Decentralized control, Logical supervisor, Self-tuning control, On-line identification

Daniela Danciu

Abstract: The paper is concerned with the improvement of sufficient conditions for the exponential stability of Hopfield-type neural networks displaying interaction delays. The results are based on a method obtained in our previous work that combines an idea suggested by Malkin for studying the absolute stability of a nonlinear system via their linearisations and a procedure proposed by Kharitonov for construction of an “exact” Liapunov-Krasovskii functional used in the analysis of uncertain linear time delay systems. Since the Liapunov function method give only sufficient conditions for stability, the improvement of these criteria is obviously necessary. These less conservative conditions are suitable for the implementation of recurrent neural networks.
Keywords: Hopfield neural network, Time delays, Asymptotic stability, Rrobust stability.

Petr Dostál, Vladimir Bobál

Abstract: The paper presents one approach to design of controllers for time delay systems. The proposed method is based on the time delay Padé approximation. The controllers are derived using the polynomial approach and LQ control technique. Resulting proper and stable controllers obtained via polynomial Diophantine equations and spectral factorization techniques ensure setpoint tracking as well as load disturbance attenuation. The procedure is developed for stable and unstable first order time delay systems and the results are verifyed by simulations in MATLAB – Simulink.
Keywords: Time delay system, Time delay approximation, Polynomial method, LQ control.

Eva Dulf, Tiberiu Colosi, Mihail Abrudean, Mihaela Unguresan

Abstract: Continuing the papers (Colosi et al, 2002; Colosi et al, 2003; Bogdan, 2005), are presented significant aspects of modeling and numerical simulation of some categories of processes, defined by partial differential equations (pde) of I and II order, with frequent applications in technique. The originality of the paper is the definition and the use of the (Mdpx) operating matrix, which, beside the disadvantage of a relatively high volume of calculus, cumulate as the main advantage the quasi-general applicability of the method for a large category of (pde), linear or nonlinear. This main advantage is attested by numerous examples, run on computer.
Keywords: partial differential equations, state variables, Taylor series, modeling and numerical simulation.

Teodor DUMITRIU, Nicolaos Antonio CUTULULIS, Emil CEANGA

Abstract: Friction is to be considered an essential process for micrometer scale tracking servo systems. Therefore, a high performing and robust control of friction becomes an important issue in mecatronics and robotics field of research. This paper is a contribution to the development of a real time friction simulator (RTFS) using Matlab/SIMULINK® and DS1103 PPC Controlled Board. The simulator is designed to generate friction torque on its mechanical shaft and to provide the static and/or dynamic characteristics of a given friction model for further friction behavior test and simulation in order to implement different friction compensation strategies.
Keywords: Hardware in the loop, Model-based compensation, Limit cycles, Rreal time system, Friction simulator, Digital signal processor, Electromechanical tracking system.

Anna Filasová, Juraj Klacik, Martin Sás, Michal Racek

Abstract: Purpose of this paper is to present an application of algorithm which has been used to solve the control problem of nonlinear system. Algorithm is based on Dual Heuristic Programming (DHP) which belongs to the Adaptive Critic Design family (ACD). ACD is a group of Reinforcement Learning methods which use a criterial function. DPH is a neural network approach which has been used to solve Bellman equation. Usual application of DPH is in field of control. The model of a power system has been used to verify the effectivity of DHP algorithm.
Keywords: Optimal control, Reinforcement learning, Dynamic programming, Neural network control, Aadaptive control.

Jozef FUTÓ, Karol KOSTÚR, Ján KACUR

Abstract: Rock disintegration is on of the most important operation s of mineral resources excavation and treatment technologies. This paper deals with of rock drilling control and optimization by acoustic signal. During the disintegration process, the thrust of the drilling tool, revolutions, drilling speed and disintegration power have been monitored. This paper analyses experiment results for disintegration process optimization with use of acoustic signal.
Keywords: Optimisation, Drilling, Acoustic signal.

Hani Hamdan, Gerard Govaert

Abstract: In this paper, we present a new and original mixture model approach for Acoustic Emission (AE) data clustering. AE techniques have been used in a variety of applications in industrial plants. These techniques can provide the most sophisticated monitoring test and can generally be done with the plant/pressure equipment operating at several conditions. Since the AE clusters may present several constraints (different proportions, volumes, orientations and shapes), we propose to base the AE cluster analysis on Gaussian mixture models which will be, in such situations, a powerful approach. Further more, the diagonal Gaussian mixture model seems to be well adapted to the detection and monitoring of defect classes since the weldings (potential critical zones) of cylindrical pressure equipment are lengthened horizontally and vertically. The EM algorithm applied to the diagonal Gaussian mixture model provides a satisfactory solution but the real time constraints imposed in our problem make the application of this algorithm impossible if the size of data becomes very large. The solution that we propose is to use the CEM algorithm which converges faster and generates comparable solutions in terms of resulting partition. To choose the number of mixture model components, we propose a new criterion that we call WST, which allows to obtain rather comparable practical results than the ICL criterion and is faster. The practical results on real data are very satisfactory from the experts point of view.
Keywords: CEM algorithm, Mixture model, Clustering, Flaw diagnosis, ICL criterion, WST criterion, Acoustic emission.

Ionela Iancu, Eugen Iancu

Abstract: The use of bolus subcutaneous injections of insulin for the control of blood glucose concentration in a diabetic patient is a normal clinical practice. Better physiological response should be obtained if the glucose level is monitored regularly and insulin delivered in a regime more closely resembling to the normal release mechanism of the pancreas. The pump of insulin represents the solution of the physiological request. In this work, different structure for control of the pump with maintained of blood-glucose level are analysed.
Keywords: Blood glucose control, Insulin pump, Modeling and simulation, Mellitus diabetes.

Constantin Marin, Emil Petre, Dan Selisteanu, Dorin Sendrescu, Anca Petrisor

Abstract: This paper presents a procedure for time variant nonlinear system identification based on distribution theory. Some of the system parameters change in time according to unknown laws. These laws are expressed as finite degree time polynomials whose parameters are included in the set of parameters to be identified. Mainly it is an extension of the procedure developed in (Marin et al., 2005). Considering functionals weighted by polynomials, it is possible to transform a time variant differential system of equations to an algebraic system in unknown parameters. The hierarchical structure identification method for rational expressions, proposed in (Marin et al., 2005), is now extended to time variant systems. An application for parameter identification of a wastewater biodegradation process, considering time variant yield coefficients is presented.
Keywords: Identification, Distribution theory, Functionals, Bioprocesses.

Letitia Mirea, Lavinia Ferariu

Abstract: The paper investigates the development of the a type of a Co-Active Neuro-Fuzzy System (CANFS) and its application to Fault Detection and Isolation (FDI). Hybrid learning, based on a fuzzy clustering algorithm and a gradient-like method, is used to train the CANFS. The experimental case study refers to the component fault diagnosis of a Three-Tank System. A neuro-fuzzy simplified observer scheme is used to generate the residuals (symptoms) in the form of the one-step-ahead prediction errors. These are further analysed by a neural classifier in order to take the appropriate decision regarding the behaviour of the considered process.
Keywords: Fault diagnosis, Neuro-fuzzy systems, Fuzzy clustering, Neural networks, Neural classifier

Octavian Pastravanu and Mihaela Matcovschi

Abstract: A sufficient condition that is frequently used for exploring the exponential stability of an interval system requires the stability of a unique test matrix, adequately built from the interval-type coefficients. We prove that the fulfilment of this sufficient condition guarantees a stronger property for the interval system, called diagonal stability, which, concomitantly with the standard exponential stability, ensures the flow (positive) invariance of certain time-dependent sets with respect to the state-space trajectories. Our approach covers both cases of discrete- and continuous-time interval systems.
Keywords: Robustness, Interval systems, Diagonal stability, Exponential stability, Timedependent invariant sets

Emil Petre

Abstract: In this paper the problem of identification and nonlinear adaptive control for a class of biotechnological processes for which the dynamics is incompletely known or completely unknown is investigated. The nonlinear controller design is based on the input-output linearizing technique. The only information required about the process is the measurements of the state variables and its relative degree. The unknown controller functions are approximated using radial basis neural networks. For this, firstly, a state transformation that allows a separate identification of the kinetic models of the reaction rates even if they occur simultaneously in the reactor is presented. Then, the model structure of each reaction rate is separately identified by using a radial basis neural network. The presented technique is applied to a nonlinear and time varying biotechnological process model corresponding to alcoholic fermentation. Simulation results are included to show the effectivenes of the proposed controller by comparison to an exactly linearizing controller.
Keywords: Nonlinear systems, Identification, Nonlinear control, Adaptive control, Neural networks.

Catalin Petrescu, Ciprian Lupu , Dumitru Popescu

Abstract: The paper presents the results of the research performed by the authors on systems control and optimization for the operating process of the cowpers from the ISPAT-SIDEX steel plant. This system was developed on two relevant levels interconnected in a hierarchical control structure. The acquisition and control level was designed using specialized microcontrollers. The supervisor level for the optimization of the combustion process was implemented on an operator console. The solution of the optimization problem represents the optimal decisions, translated in real-time procedure to the acquisition and level control.
Keywords: Supervisory control, Identification, Optimization, Plant, Steel industry

Vladimir Rasvan

Abstract: Existence of nonlinear oscillations (periodic and almost periodic solutions) for systems with deviated argument and sector restricted (Lurie type) nonlinearities) has been discussed using the Liapunov functional approach as well as the approach of the frequency domain inequality due to V.M. Popov. In this paper a comparison of the computational issues of the two approaches is performed using a standard example supplied by an electrical circuit containing a single lossless transmission line. It is shown that the computational di±culties of the frequency domain method may be overcome when a suitable Liapunov function(al) - here the stored electromagnetic energy - is readily available.
Keywords: Nonlinear oscillations, Ppropagation, Liapunov functional, Eelectrical circuits

Monica ROMAN and Eugen BOBASU

Abstract: In this paper we will study the structural controllability and observability of bond graph models. The concept of causality is an important concept embedded in bond graph theory and it presents the dependencies between elements. Through causal manipulations on the bond-graph model, the structural analysis of dynamic systems can be done. Starting from causal bond graph model and using some theorems we will study the controllability and the observability of the system. Also, a matriceal method is used in order to test the controllability and the observability of such systems. The structural properties are studied using illustrative physical examples.
Keywords: Bond Graphs, Causality, Controllability, Observability

Sas Martin, Filasova Anna, Klacik Juraj

Abstract: The paper presents multiple model control strategy based on supervisor. The process outputs are fed back to a set of controllers. At a certain point in time only one control signal is used. Which control signal generated by each controller that is finaly used is decided by the supervisor. The pole placement method is used to derive controllers. There is the k-nearest neighbours algorithm in the role of supervisor. The algorithm is ilustrated in simulation enviroment MATLAB using the four tank system model.
Keywords: Switching based control, Multiple models, k-nearest neighbour, Pole placement

Sendrescu Dorin and Woihida Aggoune-Bouras

Abstract: A simple car-following model replace human drivers and their low-predictable reaction time with respect to traffic problems (reaction time 0.25 – 1.25 sec needs an inter-vehicle spacing around 30 m or more at 60 km/hour). A way to solve the problem is to organise the traffic into platoons, that is groups of vehicles consisting of a leader and a number of followers “tightly” spaced, all moving in longitudinal direction. The stability of the system in the parameter space is analysed using a specialized software DDE-BIFTOOL v. 2.00.
Keywords: Stability, Cruise control, Time delay system

Abstract: For a no periodically, rapid damped, evolution of reaction between Cu²⁺ and S₂O₃^2-, the paper proposes a variant of numerical modelling and simulation, based on two exponential functions. The method assures a good approximation of the experimental solution, with a remarkable flexibility for analyses and synthesis elaborated in the paper.
Keywords: Kinetic of rapid reactions in solutions, Nonlinear-exponential functions, Systems of nonlinear equations, Numerical simulation.

Ioan Ursu, Felicia Ursu

Abstract: This paper continues recent research of authors, considering the adaptive control synthesis in the presence of parametric uncertainties, with application to electrohydraulic servos actuating primary flight controls. To account for these parametric uncertainties in the model, a parameter adaptation scheme is essential. So, certain parameters are adjusted on line; using in synthesis the methods of Control Lyapunov Functions and backstepping, the obtained dynamic update law for parameters, together with the system’s model, would render the closed loop system stable and would achieve the regulation of the output. The work also illustrates how the main theory can be brought or adapted to control design practice as defined by a given mathematical model. Barbalat's Lemma is used in the proof of control law structure. Numerical simulations are reported from viewpoint of servo time constant performance.
Keywords: Nonlinear control synthesis, Adaptive backstepping, Control Lyapunov Function, Barbalat's Lemma, Electrohydraulic servo.

Matei Vinatoru

Abstract: This paper is analysing aspects regarding the control of the power and frequency of a synchronous generator coupled to an automatic power grid by a steam or hydraulic turbine. The mathematical modelling of the system allows choosing the adequate control structures function of the inter-influence that appears between the power grid and the steam power-blocks reported to the variables that allow the energy transfer: frequency and active power.
Keywords: Power system, Modelling, Optimisation, Power control.