Irene Moulitsas’s research while affiliated with The Cyprus Institute and other places

In this work, the combined effects of compressibility and slip in Poiseuille flows of Herschel–Bulkley fluids are investigated. The density is assumed to obey a linear equation of state, and wall slip is assumed to follow Navier’s slip condition with zero slip yield stress. The flow is considered to be weakly compressible so that the transverse velocity component is zero and the pressure is a function of the axial coordinate. Approximate semi-analytical solutions of the steady, creeping, plane and axisymmetric Poiseuille flows are derived and the effects of compressibility, slip, and the Bingham number are discussed. In the case of incompressible flow, it is shown that the velocity may become plug at a finite critical value of the slip parameter which is inversely proportional to the yield stress. In compressible flow with slip, the velocity tends to become plug upstream, which justifies the use of one-dimensional models for viscoplastic flows in long tubes. The case of pressure-dependent slip is also investigated and discussed.

The CYCOFOS-Cyprus Coastal Ocean Forecasting and Observing System has been operational since early 2002, providing daily sea current, temperature, salinity and sea level forecasting data for the next 4 and 10 days to end-users in the Levantine Basin, necessary for operational application in marine safety, particularly concerning oil spills and floating objects predictions. CYCOFOS flow model, similar to most of the coastal and sub-regional operational hydrodynamic forecasting systems of the MONGOOS-Mediterranean Oceanographic Network for Global Ocean Observing System is based on the POM-Princeton Ocean Model. CYCOFOS is nested with the MyOcean Mediterranean regional forecasting data and with SKIRON and ECMWF for surface forcing. The increasing demand for higher and higher resolution data to meet coastal and offshore downstream applications motivated the parallelization of the CYCOFOS POM model. This development was carried out in the frame of the IPcycofos project, funded by the Cyprus Research Promotion Foundation. The parallel processing provides a viable solution to satisfy these demands without sacrificing accuracy or omitting any physical phenomena. Prior to IPcycofos project, there are been several attempts to parallelise the POM, as for example the MP-POM. The existing parallel code models rely on the use of specific outdated hardware architectures and associated software. The objective of the IPcycofos project is to produce an operational parallel version of the CYCOFOS POM code that can replicate the results of the serial version of the POM code used in CYCOFOS. The parallelization of the CYCOFOS POM model use Message Passing Interface-MPI, implemented on commodity computing clusters running open source software and not depending on any specialized vendor hardware. The parallel CYCOFOS POM code constructed in a modular fashion, allowing a fast re-locatable downscaled implementation. The MPI takes advantage of the Cartesian nature of the POM mesh, and use the built-in functionality of MPI routines to split the mesh, using a weighting scheme, along longitude and latitude among the processors. Each server processor work on the model based on domain decomposition techniques. The new parallel CYCOFOS POM code has been benchmarked against the serial POM version of CYCOFOS for speed, accuracy, and resolution and the results are more than satisfactory. With a higher resolution CYCOFOS Levantine model domain the forecasts need much less time than the serial CYCOFOS POM coarser version, both with identical accuracy.

Modeling oceans is computationally expensive. Rising demands for speedier and higher resolution forecasts, better estimations of prediction uncertainty, and need for additional modules further increase the costs of computation. Parallel processing provides a viable solution to satisfy these demands without sacrificing accuracy or omitting any physical phenomena. Our objective is to develop and implement a parallel version of Cyprus Coastal Ocean Forecasting and Observing System (CYCOFOS) hydrodynamic model for the Eastern Mediterranean Levantine Sea using Message Passing Interface (MPI) that runs on commodity computing clusters running open source software. The parallel software is constructed in a modular fashion to make it easy to integrate end-user applications in the future. Parallelizing CYCOFOS also enables us to run multiple simulations using different parameters, and initial and boundary conditions to improve the accuracy of the model forecasts, and reduce uncertainty. The Cyprus Coastal Ocean Forecasting and Observing System (CYCOFOS) was developed within the broad frame of EuroGOOS (European GOOS) and MedGOOS (Mediterranean GOOS), to provide operational oceanographic forecast and monitoring on local and sub-regional scales in the Eastern Mediterranean Basin. The system has been operational since early 2002, consists of several forecasting, observing, and end-user modules, and has been enriched and improved in recent years. The system provides daily forecasting data to end-users, necessary for operational application in marine safety, such as the Mediterranean oil spill and trajectory modeling system. Like many coastal and sub-regional operational hydrodynamic forecasting systems in the Mediterranean, CYCOFOS is based on the Princeton Ocean Model (POM). There have been a number of attempts to parallelize the Princeton Ocean Model, on which the CYCOFOS is based, such as MP-POM. However, existing parallel code models rely on the use of specific outdated hardware architectures and associated software. Additionally, all reported works seem to be one-off attempts with no further development, the emphasis being given to the high performance computing aspect rather than to accurate ocean forecasting and end-user applications. The goal of producing a distributed memory parallel version of POM based on the Message Passing Interface (MPI) paradigm is done in three stages producing three versions of the code. In the first version, we take advantage of the Cartesian nature of the POM mesh, and use the built-in functionality of MPI routines to split the mesh uniformly along longitude and latitude among the processors. This version is the least efficient version because the processors whose meshes contain a lot of land regions will have a lower computational load. Therefore the overall computational load balance is poor and significant portion of the time is spent idling. The objective of this version is to produce a parallel version of the code that can replicate the results of the serial version of the POM code used in CYCOFOS for validation and verification purposes Results from the first parallel version of CYCOFOS will be presented during the conference, and speedup will be discussed. We will also present our parallelization strategies for the second and third versions which will improve the load balancing and speedup by using a weighted distribution of the grids among the processors.

In this work, we derive approximate semi-analytical solutions of the steady, creeping, weakly compressible plane and axisymmetric Poiseuille flows of a Herschel–Bulkley fluid. Since the flow is weakly compressible, the radial velocity component is assumed to be zero and the derivatives of the axial velocity with respect to the axial direction are assumed to be much smaller than those with respect to the radial direction. The axial velocity is then given by an expression similar to that holding for the incompressible flow, the only difference being that the pressure-gradient is a function of the axial coordinate and satisfies a non-linear equation involving the density of the fluid. In the present work, a linear as well as an exponential equation of state, relating the density of the fluid to the pressure, are considered. The pressure distribution along the flow direction is calculated by means of numerical integration and the two-dimensional axial velocity can then be constructed. The effects of compressibility, the equation of state, the Bingham number and the power-law exponent on the solutions are investigated.

We use finite elements and the full-Newton iteration method to solve the steady, two-dimensional flow of a Newtonian planar film issuing from a slit under a pressure difference, in the presence of gravity and surface tension. The simulated film shapes agree with available experimental data within the range of the experimental error. The numerical calculations show that the shape of the film depends strongly on the imposed pressure difference, inertia and gravity, and is rather insensitive to surface tension.