La prédiction des performances des séparations membranaires barométriques, fortement affectées par la polarisation de concentration, serait une avancée importante pour le dimensionnement et l’optimisation des procédés. Dans ce contexte, les équations couplées de Navier-Stokes et de conservation du soluté adimensionnées sont résolues numériquement dans le cas d’un écoulement stationnaire laminaire en filtration tangentielle. Le canal plan bidimensionnel comporte des parois perméables soumises à des conditions aux limites du type solubilisation-diffusion. Le flux de perméat, le taux de rétention et le débit, la concentration et la chute de pression du rétentat sont déterminés localement. Les simulations soulignent l’influence des perméabilités membranaires au soluté et au solvant sur la polarisation de concentration et la dépendance non-asymptotique du taux de rétention avec la pression appliquée. Le modèle est validé pour des modules plans et spiralés d’osmose inverse et de nanofiltration dense en comparant les calculs à des résultats expérimentaux tirés de la littérature et de nos propres essais pilotes de dessalement. Aussi, une méthode à l’échelle de la paillasse permettant de déterminer les perméabilités au soluté et au solvant par des expériences d’osmose et diffusion est développée et appliquée à des membranes d’osmose inverse et de nanofiltration. La divergence des mécanismes de transfert engendrés sous l’influence de la pression ou sous l’influence d’un gradient osmotique est mise en évidence. Le modèle numérique et la méthode expérimentale sont des outils prometteurs d’applicabilité immédiate dans le domaine des membranes. / The prediction of the performance of pressure-driven membrane separations, deeply affected by concentration polarization, would be an important advance for process design and optimization. In this context, the dimensionless coupled Navier-Stokes and solute conservation equations are solved numerically for a steady laminar cross-flow filtration. The two-dimensional flat channel consists of permeable walls subject to solution-diffusion boundary conditions. The permeate flux, the rejection rate and the retentate’s flow rate, concentration and pressure drop are determined locally. The simulations highlight the influence of the membrane solute and solvent permeabilities on concentration polarization and the non-asymptotic dependence of the rejection rate on the applied pressure. The model is validated for reverse osmosis and tight-nanofiltration plate-and-frame and spiral-wound modules by comparison to experimental results from the literature and from our own pilot desalination tests. Furthermore, a bench-scale method enabling the determination of solute and solvent permeabilities from osmotic-diffusive experiments is developed and applied to reverse osmosis and nanofiltration membranes. The divergence between the transport mechanisms engendered by pressure and by an osmotic gradient is evidenced. The numerical model and the experimental method are new promising tools with immediate applicability in the membrane field.

M. Tech. (Department of Information Technology, Faculty of Applied and Computer Sciences), Vaal University of Technology. / Computer based assessments have become one of the largest growing sectors in both nonacademic and academic establishments. Successful computer based assessments require security against impersonation and fraud and many researchers have proposed the use of Biometric technologies to overcome this issue. Biometric technologies are defined as a computerised method of authenticating an individual (character) based on behavioural and physiological characteristic features. Basic biometric based computer based assessment systems are prone to security threats in the form of fraud and impersonations. In a bid to combat these security problems, keystroke dynamic technique and facial biometric recognition was introduced into the computer based assessment biometric system so as to enhance the authentication ability of the computer based assessment system. The keystroke dynamic technique was measured using latency and pressure while the facial biometrics was measured using principal component analysis (PCA). Experimental performance was carried out quantitatively using MATLAB for simulation and Excel application package for data analysis. System performance was measured using the following evaluation schemes: False Acceptance Rate (FAR), False Rejection Rate (FRR), Equal Error Rate (EER) and Accuracy (AC), for a comparison between the biometric computer based assessment system with and without the keystroke and face recognition alongside other biometric computer based assessment techniques proposed in the literature. Successful implementation of the proposed technique would improve computer based assessment’s reliability, efficiency and effectiveness and if deployed into the society would improve authentication and security whilst reducing fraud and impersonation in our society.