For almost all engineering problems, solutions and calculations need to be reliable and accurate. Errors in calculations can lead to procedures not working as intended, resulting in costly problems, or even serious safety issues. Understanding the limitations of any tools used as part of the design and validation process, including CFD, is essential. Knowing these limitations allows an engineer to estimate the errors and provide reliable conclusions that do not over promise.
There is no doubt that the classic computer science saying of “garbage in, garbage out” applies to Computational Fluid Dynamics (CFD). With some of the more user friendly CFD packages that exist in the current market it is relatively easy to set up a simple model, and produce colourful results that bear little resemblance to the modelled systems behaviour. For this reason some opponents of CFD like to refer to it as “Colourful” Fluid Dynamics. However, when used correctly, with accurate inputs, it can produce incredibly powerful and meaningful results.
Inaccurate CFD results can stem from many sources, each of which can be alleviated or reduced through experience and care. Potential errors include:
- Inadequate mesh resolution in areas of high gradients or fine details. This can lead to excessive diffusion or unrealistic flow fields.
- Using an inappropriate turbulence model. There is a wide range of turbulence models, each of which performs better under different conditions.
- Improperly defined inputs and outputs. Without accurate input and output criteria you are solving a different problem!
- Inadequately located and defined boundary conditions. Putting the boundaries too close can lead to inaccurate flow fields, and not defining the correct boundary velocity profiles in some cases can produce wildly inaccurate results.
- Neglecting small, but important physical processes, such as heat fluxes, turbulent mixing and buoyancy. In many cases these effects are negligible, but sometimes they can have large impacts.
Knowing the potential for errors is the first step to preventing and minimising them. After identification, skilled CFD engineers, such as those at Synergetics, can work to isolate and circumvent them to produce accurate and reliable simulations. This process starts by thoroughly examining the problem, identifying relevant physical processes, and filling any data gaps. Experienced CFD engineers, like those at Synergetics, select the optimal models based on both experience and pertinent academic literature. If published literature is insufficient or there is a lack of agreement, two options are considered.
Firstly, engineers take onsite measurements for existing or similar facilities where possible. This is especially useful when developing models for very complex systems, which have not been modelled in the past. Such onsite measurements allow the CFD model to be validated, and can identify any deficiencies in the model. For example, when performing ground breaking modelling of a multiple hearth furnace, Synergetics collected onsite measurements, and then used the CFD model to replicate those operating conditions, achieving validated errors of under 10% on almost all sensors. This provided both our engineers, and the client with confidence that our model was able to accurately resolve and predict the complex physics occurring within the furnace.
Secondly, CFD engineers consider the sensitivity of the solution to different approaches and, where relevant, perform tests. Sometimes two models will produce near identical results. In other cases, one may predict a more extreme result than another. In this case, if validation data is not available, the solution which minimised risk to the client is taken. Typically this will be the model that predicts the worst case behaviour, presenting a more conservative solution. This approach is especially useful with mesh resolutions, as a sensitivity study will quickly reveal any mesh dependence on the solution.
Using a combination of all these techniques, together with extensive experience, skilled CFD engineers can provide their clients with reliable solutions to their engineering problems. Understanding the limitations in detail enables the identification of both best and worst-case scenarios and gives the client confidence that the solutions can be trusted.
It is important to understand that CFD is not the right tool for every problem. However our staff can help you assess you engineering problems and determine if CFD is the right tool for you, and if it is we can assist you in getting an accurate solution that meets your requirements.