Parallelization of nozzle flow model in spray-détente instantanée contrôlée technique on parallel virtual machine

• Main Author: Suhari, Nadia Nofri Yeni
• Format: Thesis
• Published: 2017
• Subjects: QA Mathematics
• Online Access: http://eprints.utm.my/id/eprint/81553/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:124848

instant controlled pressure drop or détente instantanée contrôlée (dic) in spray drier has become a new innovative combination processes for powder production that has a potential to overcome the spray drier limitation in terms of high energy consumption and microbiology contamination. this is done by inducing the abrupt pressure dropping in a short time period and predicting the time duration of the thermal treatment at the high temperature in drying processes. due to the presence of spray atomizer in spray-dic technique, the integrated of mass and energy equations in the nozzle flow model is modified by adding the pressure changing parameter in one and two dimensional hyperbolic type of partial differential equation (pde). the parallel algorithms are developed from discretization of finite difference methods (fdm) and dealing with implicit and crank-nicolson schemes with the sequential algorithm as a benchmark. the development of a linear system equation from fdm is solved by alternating group explicit (age) compared with red black gauss seidel (rbgs) and jacobi (jb) as a benchmark methods. the iterative methods are dealt with large sparse matrix size using domain decomposition in grid generation by dividing into the sub-domains and communicate between cpus using single instruction multiple data (simd) in parallel virtual machine (pvm) with c language on linux fedora 21 for intel core-i3. the comparison of numerical analysis and the parallel performance evaluation are conducted for analysis and comparison purpose. the visualization of the multidimensional models, comparison tables and graphs are presented. in conclusion, the nozzle flow model has been validated since the visualization has the same behaviour with the statistical models. therefore, the parallelization of spray-dic model in a large sparse linear system of fdm provide the best performances and has potential to predict the mass of moisture and energy in temperature behaviour during powder drying process compared to the sequential algorithm computation.