What is the best way to carry the characteristics of the fluid using FloWorks?

[lucas_delago]

[ FlexSim 18.1.1 ]

The main issue here is that some flow processes in the operation depends on the characteristics of the passant fluids.

Let’s think in a toy example:

• The Source feeds the system with water, which is storage in a homogenization tank and then the water flow over a splitter
• The chemical concentration of the incoming water varies over time
• In the tank, the concentration of “new” water and “old” one merge
• The water that outcome of the tank has the average concentration
• Finally, the outflow percent of the splitter depends of the current (“inside” splitter) water concentration

Back to the question, what is the best way to map/modeling this kind of properties in the model?

[arunTTT2P]

@Mischa Spelt

[mischa_spelt · Accepted solution]

Short answer: The best approach is not FlexSim. The second best approach is not FloWorks but FlexSim with time slicing.

Long answer:

No matter how cool we think FloWorks is, and how happy I'd be to tell you it's the perfect solution for anything fluid related, your situation has two characteristics that make me feel like FloWorks is fundamentally not the way to go here.

1. The problem is not linear

This is quite a complex issue, mainly because the concentration does not change linearly which means that you cannot use any built-in functionality such as kinetic tracked variables. In fact, the current concentration is something like

where C(t) is the current concentration, C_in is the concentration of the inflow, t₀ is some fixed starting point (presumably the last point at which the inflow concentration changed) and C₀, V₀ are the concentration and volume of the tank content at that instance; r_in is the inflow rate.

This suggests that any linear approach, like FloWorks uses and FlexSim provides with the kinetic level tracked variable, is bound to fail.

2. The problem is continuous

In fact, the fact that the outflow rate of the tank is also changing continuously based on the concentration in the tank, which in turn changes the concentration in the tank, suggests that any discrete event approach is probably not suitable and you will end up with something like time slicing. In which case, if you are going to discretize this anyhow, you are probably better off finding a CFD package with some proper discrete element calculations.

So what then?

Maybe there are simplifications that you could make, that will linearize the problem, or allow you to run a different model that you can post-process to get the proper statistics. I'd have to put a bit more thought into that and you'd probably have to give me a more extensive description of your goals.

[mischa_spelt]

Thanks, I noticed it. Maybe @Harald E, @dj.moens, @steven.hamoen have something to add?