Designing a new treatment technology on the computer

Significant costs and time can be saved by using computational fluid dynamics (CFD) modelling in the design stage of a technology. In a matter of weeks, numerous design modifications can be tested. Process performance will be optimal and many operational problems can be avoided. Here is a practical case example.
‍

CFD-based design of the novel AAA settler at various plants

Authors: Usman Rehman*, Wim Audenaert*, Ingmar Nopens*, Peter Aichinger°, Bernhard Wett° (*AM-TEAM; °ARA Consult)

The AAA ('Triple A') process is a novel A-stage technology developed by Dr. Bernhard Wett. The system consists of two identical reactors going through the different stages of feeding, aeration (activation) and settling/wasting. The reactors alternate, which means, when one is aerating, the other one is settling and wasting.

Rather than explaining the whole technology here, we refer to the video embedded in this article, in which the inventor explains the technology himself.

As shown in the next figure, the influent feeding structure is at the bottom, while the effluent leaves at the top. A well designed AAA process looks like this:

1. No A-stage sludge is washing out (this avoids a downstream issue)
2. Influent is distributed equally, and the sludge blanket fluidization is homogeneous (maximal performance)
3. It is practical and easy to construct

We helped designing three very different types of A-stage reactors, as shown on this slide:

After diagnosing the mixing and flow distribution in a first design, we started changing the inlet and outlet configurations on the computer. And these are more scenarios than you can practically test. Changing a design after full-scale construction is much more expensive and time taking (emptying tank, finding the problem, stainless steel modifications, ...). This holds for most of the treatment technologies. We could find the optimal designs in less than 1 week, or in a matter of a few days if needed.

Lesson 1: CFD is visual + quantitative. It allows you to analytically compare different designs

The CFD plots on the right show the distribution of a 'virtual tracer' we injected in the inlets (green graphs) and local flow velocities (blue graphs). The simulation allowed us to study how the colour would distribute in the system. A completely mixed system has a rather homogeneous color (same tracer concentration everywhere). You probably notice the dark blue zone in the base case design (top left). After changing the inlet configuration, distribution was much more homogeneous. That change was implemented at the plant, with excellent outcomes. To quantitatively compare two designs, we use our in-house developed tool: 'Rehman-Nopens curves' (right). These two curves summarize the distribution of tracer in both designs. Those curves indicate the tracer concentration distribution in the reactor. You immediately notice that the optimal design has an almost vertical curve, while the orange (initial design) shows large heterogeneities.

Same story for the much larger circular tank. We optimised the design, and could see the Rehman-Nopens curves clearly shift towards optimal shapes. The green curve represents the original design (the top colorful figure shows this uneven tracer distribution). Optimisation drastically improved performance. An additional third scenario only provided marginal gains.

The pilot-scale system was rather optimal. We tried to change the outlet location to further minimise the risk on shortcircuiting, but it only detoriorated performance. The final three designs looked like this:

Lesson 2: CFD models are accurate, if run with craftmanship. Don't let your doubts stand in the way of using this powerful design tool

Yes, we performed model validation. It helped convincing all the parties involved, and as such it helped increasing the value of the outcomes. However, is this always needed? No. The model we used was validated so many times before. Hence, the barrier to use CFD is low in many cases, as validation is not always needed. PS: We had a whole separate discussion on validation in our keynote at IWA Watermatex.

We validated the Italian plant quantitatively with velocity measurements (all measurement locations are the yellow dots in the below figure). We checked local sludge accumulation qualitatively with blanket sampling.

All velocity measurements were shown to be transparent (the following graphs show velocity as function of tank depth). We have quite some full-scale plant validation experience, and we can say that the model was proven to be very accurate. The CFD model both predicted the magnitudes and trends of the velocity curves (blue curves are measured velocities; green curves are simulated velocities). 16 out of 19 points had excellent agreement. The 3 other points were all located right in front of intlets. It's always challenging to measure there.

Qualitative blanket measurements were in agreement with the local sludge accumulation predicted by the model.

On top of this, Dr. Bernhard Wett and his colleagues also did dye testing in his pilot. It gave some extra confidence in a qualitative way.

Conclusions

And this was our last slide. Main take home?

CFD is a mature and practical design tool that simply leads to better designs faster, and with less struggle. Many ongoing plant optimizations could have been avoided with proactive use of CFD. CFD modellers talking the 'technologist language' certainly helps and only speeds up the project.

We hereby want to acknolwedge

• Our simulation team, and specifically Dr. Usman Rehman, Dr. Giacomo Bellandi and ir. Simon Duchi.
• Ansys Fluent, for the good collaboration

Did you like this article? Curious to hear your opinion or questions.

PS: Enjoy the Italian mountains in this picture!