Tunis University

About: Tunis University is a education organization based out in Tunis, Tunisia. It is known for research contribution in the topics: Population & Thin film. The organization has 11745 authors who have published 15400 publications receiving 154900 citations. The organization is also known as: University of Tunis & UT.

A GIS-based statistical model for assessing groundwater susceptibility index in shallow aquifer in Central Tunisia (Sidi Bouzid basin)

Abstract: A comprehensive approach for assessing the shallow aquifer Susceptibility Index (SI) to pollution was proposed by combining the Vulnerability Index (VI) and Quality Index (QI) in Sidi Bouzid basin in Central Tunisia. Hydrochemical investigation showed that nitrate concentrations and total dissolved solid (TDS) values of the Mio-Plio-Quaternary (MPQ) aquifer in the study area were ranging from 14.3 to 111 mg/l and 1218 to 6202 mg/l successively. VI was first estimated using either a generic DRASTIC model or DRASTIC-LU model by adding land use (LU) factor, with preset factor weights; these weights were later adjusted using a single parameter sensitivity analysis (SPSA) or two different statistical methods: canonical analysis of principal coordinates (CAP) and partial least squares (PLS). Compared to the generic models, the weight of the factor impact of vadose zone (I) is equal to 5 remained the highest for all the other models, except for DRASTIC one using a CAP weight adjustment technique where the weight of I is equal to 1. DRASTIC-LU and DRASTIC-LU-CAP models predicted the widest (VILU − min=89, VILU − max=206) and narrowest (VILU − CAP − min=59, VILU − CAP − max=125) VI range, respectively. VI obtained by different weight adjustment techniques significantly correlated with nitrate concentrations with a significant correlation coefficient, higher than 0.50. Based on a model selection criterion, correlation between vulnerability indices and nitrate concentration, DRASTIC-LU-CAP may be recommended as the best model. QI was assessed by simply adding the concentration of some major elements ( $$ {Cl}^{-},{Na}^{+},{NO}_3^{-} $$, $$ {SO}_4^{2-} $$) and electric conductivity (EC) transformed into ordinal classes (1–5). Groundwater SI maps for both drinking and irrigation water generated into a GIS-based map showed that a great part of the study area had a high SI to pollution.

Model Predictive-Sliding Mode Control for Three-Phase Grid-Connected Converters

Abstract: This paper focuses on the implementation of an extended model predictive-sliding mode control (EMP-SMC) for three-phase grid-connected converters (GcCs). The proposed control considers the GcC model in the dq synchronous reference frame to forecast possible future behavior of the grid current. After that, the applied voltage vector is selected so that a predefined cost function is minimized. This function is aimed to reduce the grid current ripples during steady-state operation. Compared to the conventional MPC, the proposed EMP-SMC algorithm uses 19 voltage vectors (7 real voltage vectors and 12 additional virtual voltage vectors) for the prediction process. Accordingly, lower grid current THD and lower switching losses are obtained. However, the increase of voltage vectors number will lead to higher computation time delay that may affect the control performances. To overcome this problem, the proposed control uses a sliding mode-based preselection step that limits the prediction process to only ten voltage vectors at the most and a table-based implementation process that reduces significantly the execution time of the whole control algorithm. Simulation and experimental results are presented in order to show performances and effectiveness of the proposed EMP-SMC algorithm.

Machine learning framework for image classification

Abstract: Hereby in this paper, we are interested to extraction methods and classification in case of image classification and recognition application. We expose the performance of training models on varying classifier algorithms on Caltech 101 images categories. For feature extraction functions we evaluate the use of the classical SURF technique against global color feature extraction. The purpose of our work is to guess the best machine learning framework techniques to recognize the stop sign images. The trained model will be integrated into a robotic system in a future work.