Computational Fluid Dynamics (CFD) is the study of fluid dynamics using a computer. The mathematical equations that govern the flow of fluids are of a type that is extremely difficult to solve (Nonlinear Partial Differential Equations). Because of this mathematicians, physicists and mechanical engineers who studied fluid flow without access to modern computers had to make simplifying assumptions in order to solve these equations. Many important discoveries were made by early fluid dynamicists - such as how to make airplanes fly - but the number of problems that could be studied in this way was extremely limited.
Modern computers and sophisticated computer programs are able to solve the fluid equations for almost any type of flow. Every year new problems that were once unsolvable get solved as computers and researchers become more advanced. By using computers mechanical engineers can explore the flow of fluids in regions that are experimentally difficult to observe, and they can perform numerical experiments that are often more flexible and more informative than physical experiments.
Consider the water cooling a the nuclear rods of a nuclear reactor. If the nuclear rods are not cooled the reactor will melt-down and many people will die. The flow of water in the reactor core is difficult to measure experimentally since the core is highly radioactive and extremely hot. To study possible designs for cooling the core, moreover, it would be absurd to build a large number of reactors with different designs and test each one experimentally. Instead it makes much more sense to simulate the system on a computer. One can then simulate many different cooling designs without building many reactors, which is both expensive, and in this case dangerous.
Even in the study of fluid flow in a safe environment, such as the flow of air over a wing or around the space shuttle, computer simulation provides significant benefit.
Suppose one is designing a sailboat. There are two fundamental fluid flow problems: how the hull of the boat moves through the water and how the wind fills the sails. One could, of course, design a model sailboat of a shape that one thought was pretty good, and then test it in an experimental tank with water and wind. One could measure the air and water flows at certain points and try to determine areas where the sails seemed to be spilling wind ( and thus reducing their effectiveness ) or where the hull was moving through the water with unnecessary friction. There are two obvious problems.
The first problem is that with an experiment one only takes measurements at certain points in the flow field so one does not know what is happening everywhere in the flow. Obviously one can use many measuring devices and take more measurements, but this adds to the cost of the experiment, and usually the measuring devices themselves affect the flow so adding more can actually make your results inaccurate. In a computer simulation one solves for the entire flow field and thus one has access to information about the flow field at every point in the flow.
The second problem is that if one decides that the design one just tested is not perfect one must build another model of a new design and test further. Model building ( or prototyping as it is usually called in engineering ) is time consuming and expensive. If one wants to find the "best" design one might want to experiment with dozens of different designs. This would be extremely expensive if one had to build dozens of physical models, but it is very inexpensive (although possibly time consuming ) to test dozens of designs on a computer.
Computational fluid dynamics is a critical part of the research and development in academia and industries such as aerospace, automotive, naval architecture, power plant design, biotechnology, aquaculture, environmental engineering and many more.