Computational models for the simulation and monitoring of developing crystalline deposits originating from dripping process liquors

• Main Author: Dawson, Michael C. H.
• Other Authors: Borman, Duncan J. ; Lesnic, Daniel ; Hammond, Robert B. ; Rhodes, Dominic
• Published: University of Leeds 2014
• Subjects: 621.48
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.638908

The work in this thesis focused on several problems relating to the growth and fouling of crystal mass in industrial environments, due to leakage of salt solutions or process liquor. The work has direct application to the nuclear industry, where the size and morphology of material deposits heavily impact on their associated criticality risk. An absence of clear methods and techniques to either predict or non-invasively monitor the growth of these crystalline deposits proves problematic for industrial specialists. Therefore the main part of the thesis focused on the development and implementation of models such that the growth behaviour of crystalline formations could be evaluated and quantified for varying physical parameters. This was accomplished through both the adaptation of previous geological models, and the development of a coupled multi-physics model such that fluid flow, heat transfer and crystallisation mechanisms could be considered. The models were validated against an experimental dataset provided by the National Nuclear Laboratory, and results were shown to be in good agreement. Through parametric studies it was determined that the characteristic shape of the formation was heavily determined by the initial solution concentration, flow velocity, temperature and the rate of evaporation. A method for the non-invasive monitoring of the deposits was also investigated through the solution of a geometric inverse problem governed by the transient two-dimensional heat equation. A meshless numerical method, namely the method of fundamental solutions was used as a direct solver in a complicated highly non-linear constrained minimisation. The model was shown to perform well when reconstructing simple shapes with highly contaminated input data. Additionally, complex shapes were also captured with a reasonable degree of accuracy and stability.