What is the most difficult part of simulation? It's not about getting a result, nor is it about getting a very complex result, but knowing whether your calculations are correct and understanding the magnitude of the error in your results.

Our textbooks and teachers have always taught us the "correct" simulation methods, but they haven't told us how large the error in the "correct" simulation methods is.

If you don't know how to use simulation software, there are seniors to guide you. If you can't handle complex cases, there are simulation experts to help you. But who can tell you how large the error is?

Without knowing the error, you can't confidently guarantee the reliability of the simulation results, and you can't make decisions based on the simulation results. So, how useful are the results of the simulation?

This kind of difference is often not random; it is always either too low or too high, representing the range of systematic error in this simulation.

Simulation has a long calculation process, during which the data fluctuates chaotically like stock market curves. When you see these fluctuation ranges getting smaller, we say the simulation has converged, and the fluctuation range at this time is the fluctuation error of the simulation.

For example: In the first 100 steps before convergence, the speed fluctuates between 99 and 101, so the fluctuation error range of the speed is 2%.

A common method is to compare the calculation results of different models; the difference is the error range.

This method is unreliable. Even if you keep the grid and boundaries unchanged and only change the model, the resulting changes may not necessarily be model errors. Perhaps the differences in the model are smoothed out by the grid, or the model amplifies the influence of the boundary, resulting in differences. This method seems reasonable but often yields unreasonable error ranges.

A simple method is to look up literature. You may not find an accurate model in the literature, but you can compare these "not very accurate" models.

There is no need to refine the grid and check if the results change; this method of grid independence analysis may not capture the correlation of the grid.

During the simulation process, you will use different forms of grids. Even if the total number of grids remains unchanged, the changes in grid forms lead to more realistic errors.

The above methods may only uncover relatively small error ranges. What if there is a particularly large error? How can you quickly find the maximum error range?

Find the maximum error using the differences between presence and absence:

Maximum error of turbulence model < Difference between turbulence and laminar flow

7. Random Comparison - Normal Error

In fact, estimating the error range is quite simple. In the same simulation, during the debugging process, as you make changes, you may find that certain numbers always fluctuate within a certain range; this fluctuation range is approximately the error range.

However, many newcomers to the industry are scared to death during this process. During the debugging and simulation process, they find that changing one setting or adjusting a little grid can result in data increasing or decreasing exponentially. The simulation results are inconsistent with the experiments and also inconsistent with others' simulations, with errors that are astonishingly large.

The occurrence of this situation has nothing to do with basic errors, systematic errors, fluctuation errors, model errors, grid errors, or any other errors; it is simply a mistake, and a very basic one at that.

Only when your calculation results are relatively stable and roughly the same as the experiments and others' results can you qualify to discuss errors; otherwise, you can only discuss mistakes.

Error: When you have eliminated basic mistakes and are close to data from various sources, this seemingly ambiguous difference is the error range. 

Errors will not ruin your design; mistakes will.

Statement: The content of the above article is organized from the internet. If there are any copyright issues, please contact us immediately.