Eliana-Dina Andreica*, Mugurel Ionut Andreica**
* Computer Science Department, Politehnica University of Bucharest, Splaiul Independentei 313, 060042, sector 6, Bucharest, Romania (email: eliana.andreica@cs.pub.ro) ** Computer Science Department, Politehnica University of Bucharest, Splaiul Independentei 313, 060042, sector 6, Bucharest, Romania (email: mugurel.andreica@cs.pub.ro)

Abstract: In this paper we present a novel approach for the problem of computing a small perimeter simple polygon which separates a set of M red points from a set of N blue points in the plane (i.e. all the points of the same color are inside or on the border of the polygon and all the points of the other color are outside or on the border of the polygon). Our approach is based on generating non-self-intersecting spanning trees of the points located inside the polygon and then convexifying these spanning trees in order to obtain a minimal separating polygon (i.e. a polygon whose perimeter cannot be decreased further by convexification operations and which completely contains the original spanning tree). By generating multiple spanning trees we are able to obtain separating polygons with various perimeter lengths. We consider both the case when the separating polygon may only contain the original points as vertices, as well as the case when new points can be used as vertices of the polygon. In the second case each spanning tree is convexified to (approximately) the same polygon irrespective of the order in which the convexification operations are performed.
Keywords: Adaptive algorithms, Polygons, Trees.

Ionut Brandusoiu, Gavril Toderean
Technical University of Cluj-Napoca, Cluj-Napoca, Romania (e-mail: ionut.brandusoiu@gmail.com, toderean@pro3soft.ro)

Abstract: Nowadays, organizations are facing several challenges resulting from competition and market trends. Customer churn is a real issue for organizations in various industries, especially in the telecommunications sector with a churn rate of approximately 30%, placing this industry in the top of the list. Because higher expenses are involved when trying to attract a new customer than trying to retain an existing one, this is an important problem that needs an accurate resolution. This paper presents an advanced methodology for predicting customers churn in mobile telecommunication industry by applying data mining techniques on a data set consisting of call detail records (CRD). The data mining algorithms considered and compared in this paper are Multi-Layer Perceptron and Radial Basis Function neural networks.
Keywords: Data mining, churn prediction, neural networks, multi-layer perceptron networks, radial basis function networks.

Monica-Adela Enache, Aurel Campeanu, Sorin Enache, Ion Vlad
University of Craiova, Craiova, 200440 Romania (e-mail: menache@em.ucv.ro)

Abstract: This paper presents the results of a study regarding dynamic regimes of asynchronous motors supplied by static converters which operate by the principle of orientation by field. Two concrete cases have been considered: control by the rotor flux with the implementation of an inverter voltage source and the control by the rotor flux with the implementation of a voltage inverter controlled in current. For this, there are detailed the equations of asynchronous machine controlled by the principle of the orientation by the rotor flux and the Simulink models of the motor and of the two types of converters used. There are also presented the simulations obtained with the help of these models for the cases when the values of the rotor resistance and of the rotor inertia moment are modified. The paper ends with conclusions obtained by finalizing the research and with references.
Keywords: asynchronous motor, vector control, modelling, simulation, testing.

Camelia Maican
Automation and Electronics Department, University of Craiova, Romania (e-mail: camelia@automation.ucv.ro)

Abstract: In this paper is study the actuator's faults detection at the steam superheaters using an electronic simulator and data acquisition board dSpace. The electronic simulator reproduces the real function of the superheater in normal condition and in fault condition. The possibility of the fault exists - blocking of the actuator on the flow injection of the second injector by a switch. In case of fault, the value of the second flow injection is set in a potentiometer. This command can be used to study the function in the manual regime of the simulator.
Keywords: fault detection, electronic simulator, steam temperature, steam superheater, data acquisition.

Elvira Popescu*, Florin Leon**
* Department of Computers and Information Technology, University of Craiova, Craiova, Romania (e-mail: popescu_elvira@software.ucv.ro) ** Department of Computer Science and Engineering, Technical University "Gheorghe Asachi" of Iasi, Iasi, Romania (e-mail: fleon@cs.tuiasi.ro)

Abstract: Learning style is one of the students' characteristics which play an important role in learning, referring to the individual manner in which they approach a learning task. Many studies have investigated the relationship between students' learning styles and their behavior in learning management systems or other traditional educational systems. With the increased adoption of Web 2.0 tools in instructional settings, it is interesting to also explore the influence of learning styles on students' usage patterns of these social media tools. Hence, this paper focuses on students' behavior in a social learning environment which integrates four Web 2.0 tools (wiki, blog, microblogging tool, social bookmarking tool); students' learning styles are categorized according to Felder-Silverman model. The analysis is based on typical machine learning algorithms for classification, association rule induction and feature selection. The investigation includes 3 scenarios: i) the analysis of the number of student actions with each social media tool); ii) the dominant tool corresponding to a learning style; and iii) the temporal evolution of the number of actions and their category. Results show that learning styles have a limited influence on the students' level of interaction with each of the four social media tools considered.
Keywords: social media, Web 2.0 tools, social learning environment, learning styles, behavioral patterns, machine learning, classification, association rule induction, feature selection.

Ion Marian Popescu*
* Department of Automation, Energy and Environment, University "Constantin Brâncu?i" of Tîrgu Jiu, Tîrgu Jiu, Romania (e-mail: marian@utgjiu.ro) Postdoctoral Researcher at University of Craiova, Craiova, Romania (e-mail: pmarian@automation.ucv.ro)

Abstract: In this paper is presented implementation of a real process control system "rotary inverted pendulum", and the possibility of extending advanced control algorithms in the thermoenergetic processes. Implemented algorithms are based on a state feedback control structure and design was created using LQR (Linear Quadratic Regulator), and Lyapunov direct method, thus making a comparison between the two methods. Rotary inversed pendulum is an underactuated process, nonlinear process, unstable, and has characteristics that we find in a complex process such as thermoenergetic group.
Keywords: LQR, Lyapunov, state feedback, power plant.

Razvan Prejbeanu*
*RELOC S.A., 109, Bd. DECEBAL, 200746, CRAIOVA, ROMANIA (e-mail: razvan.prejbeanu@ relocsa.ro)

Abstract: This paper presents new control principles of the induction motors supplied from static converters and determines the coefficients that are needed to control the electrical drives through the analysis of the output voltage deformation and the system stability. The new proposed method is verified by simulation in Matlab/Simulink which is used to draw the diagrams of the stator current, of the rotor speed and of the electromagnetic torque. This method is possibly to be validated experimentally using the tests stand, which allows finding the necessary parameters for dimensioning the electrical drives and the accepted value of the stator current harmonics.
Keywords: converter, harmonics, space vector, asynchronous machine, voltage phasor.

Claudia Alice State*
* Doctoral School of Control Engineering and Computers, University of Craiova, A.I.Cuza, 13, Craiova, RO-200585, Romania (e-mail: cldstate@ yahoo.com)

Abstract: This paper deals with the results of an investigation that was conducted in order to determine an improved mathematical model, for the lateral-directional dynamic of the GAM (Generic Aerodynamic Model) Admire theoretical aircraft. The considered system is of fourth order, expressing the lateral-directional flight dynamic of the mentioned model. In the stability analysis of the pilot-aircraft system, the human operator is supposed to control the airplane along its longitudinal and vertical axis of rotation, by means of the control surfaces - ailerons and rudder. The dynamic stability of the theoretical GAM-Admire airplane was determined by studying the linearized equations of motion around the equilibrium point. The stationary point was obtained by solving the associated homogeneous nonlinear system and the analytical research was completed by numerical computations. Several modifications of the original GAM-Admire model were conducted in order to eliminate the initial instability of the pilot-aircraft system. The closed-loop control system has been simulated using MATLAB/Simulink software package and the results obtained during the simulations under this approach confirm that the automatic control system has a good behavior and the stability is also achieved in the presence of saturation.
Keywords: Lateral-directional stability, Admire, Control system, Linearization, Feedback stabilization, Hurwitz polynomial.

Anca Loredana Udristoiu*, Stefan Udristoiu*
* University of Craiova, Faculty of Automation, Computers and Electronics, Romania (e-mail: aion@software.ucv.ro)

Abstract: This paper proposes a method for predicting student scholar performance based on the utilization of precision variable rough set theory. The study presents how to deal with an incomplete e-learning system in order to generate interesting rules capable of classification. The variable precision rough set theory is a powerful approach that permits the searching for patterns in an incomplete e-learning database. The results of experiments are very promising and show that the methods based on variable precision rough set theory are very useful for predicting the results of the student during a course activity.
Keywords: Variable precision rough sets, Data mining, Electronic learning.