stages of a
CFD simulation
back to computational fluid dynamics
Generation of the 3D Model
Generating a three-dimensional CAD model of the geometry of the fluid domain is the first stage and is usually based on 2D plans or diagrams supplied by the client and then converted into solid 3D suitable for numeric treatment.
The solid model must maintain the initial geometry and the relevant characteristics in order to capture the flow, making it possible to overlook details that are essential on a manufacturing level, but which are considered of negligible importance from the perspective of the physical processes that occur. There are no written rules about what is relevant, therefore even at this initial stage of the process a certain degree of applied experience is required.
The most universally accepted formats of solid exchanges are IGS and Parasolid, although we can also use other common formats available to the client.
Domain shaping
The discretisation of the fluid domain in small cells called elements or finite volumes is the second stage. The form of these may vary (tetrahedric, hexahedric, prismatic, etc.). The size of the domain divided by the required resolution determines the number of elements, limited by the available memory.
The complexity of the physics involved together with the size of the domains gives an outline of the size of the problems and the potential of the calculation required. The density of nodes or elements may vary from some regions to others, with greater numbers accumulating in areas where strong variations of certain variables are expected.
Solving equations
Equations that govern the transfer of mass, quantity of movement, energy, species, etc. are resolved in each element of the mesh generated in the previous stage. Since the equations are in partial derivatives, first they have to be converted into algebraic equations (introducing numerical errors of discretisation and truncation) using the most appropriate numerical schematics. This involves having a series of equations in partial derivatives over a space continuous (x,y,z,t) to a finite system of algebraic equations with discrete independent variables (x[i],y[i],z[i],t[j]). The number of equations to be solved is obviously very high, to the order of one million per time period.
In the intermediate stages of discretisation and truncation numerical errors appear, in addition to rounding errors owing to the use of a finite number of decimals, which must aim for zero in order for the numerical solution to be similar to the real one.
Associated to the quantification of errors are the concepts of verification and validation of the calculations. Checking that all the equations have been solved correctly is known as model verification. This has little to do with physics and is purely a matter of numerical calculation. Model verification, to the contrary, involves determining the suitability of using equations that are really solved as an approximation of the mathematical model of the physical process.
Analysis of the results
Generally speaking, the analysis and post process of the results are attributed less importance than they actually possess.
As soon as the equations have been solved, we have the values of the variables that define the problem in each element of the mesh. Furthermore, if the problem is not stationary, we obtain a series of such data for each time period. As one would expect, we have a large quantity of data from which useful information needs to be extracted.
At SIMPPLE we have the resources, experience and multi-disciplinary skills necessary to know exactly what the client is looking for. And more importantly, how to obtain it. The CFX-Post© post process and visualisation software contains all the tools necessary together with the maximum flexibility to obtain exactly what is required and to show the client the results in the best way. Visualising the flow and related aspects, even in the most complex geometries, is the best way of understanding the process and of going directly to the optimum solution.
