Ammonia Synthesis

[ermorgan]

Hi Jasper,

I have a quick question (I hope). I'd like to run a parametric study in which I vary the operating pressure of the NH3 synthesis loop. However, the loop contains both a feed compressor, and a recycle compressor. Thus, if I vary only the recycle, then it's not representative of a real plant. Is it possible to link the two compressors with a controller? For example, if the feed is at 180bar, then the recycle knows to be at 180 bar as well. I'd like to do this to see how the compressor power requirements (i.e. cost) change with pressure. It also gives a cost for the net energy if some $/kWh is assumed.

Thanks,

Eric

[jasper]

You can vary one using the study. Then measure that seem feed pressure using a measure unit. Put a pump or compressor (for liquid or vapor) in the other feed line. Couple the two pressures using an information stream using virtual info ports.

[jasper]

Sorry - did not read well. If you are not changing feed pressures but device pressures, you can couple them directly via an information stream on virtual information ports.

[ermorgan]

Hi Jasper,

I've been using the parametric study feature to analyze the sensitivity of parameters. However, it appears to give inconsistent results. For example, if I vary the heat transfer coefficient on my reactor from 0 to 2 W/m^2K, and output the volumetric flow rate through the reactor, the answers seem to make sense. However, if I select the heat transfer coefficient on my reactor to be 1 to 3 W/m^2K and run the same study outputting the volumetric flow rate, the results change. I'd expect the volumetric flow rates for each study to have the same values for corresponding heat transfer coefficients in the 1-2 W/m^2K range. This is not the case. I can remedy the situation by using the same initial values in the range (i.e. 0 to 2 and 0 to 3, rather than 0 to 2 and 1 to 3). This suggests that the parametric study does not "reset" the flow sheet for each queued entry and instead uses the solved value from the previous solved/unsolved entry. Is that how the study works, and if so, can it be augmented to reset the flowsheet for all entries in the queue? Could you explain the general algorithm that's used.

Thanks as always.

Eric

PS I was able to match the outputs for the "Control" paper attached in this thread. I also matched the Luyben book chapter that discusses a cooled ammonia reactor. Dead on.

[jasper]

The initial value for each step in the study is taken from the nearest solved case. As you start from a solved case, modifying a parameter to be both higher and lower typically will result two directions being walked at the same time. The solution dialog indicates for each case which case solves as initial case. For multiple variables being changed, the order of 'nearest' is described here: http://cocosimulator.org/index_help.php ... metric.htm

From what you tell I would also expect the curves to overlap. Unless multiple solutions are possible (e.g. if the solution is underdetermined, for example because pressure inside a recycle loop is not fixed to an absolute value), or in case the sensitivity is so high that both curves are solved with specified precision.

In either case, you should manually be able to reproduce different solutions for the same inputs on the main flow sheet, if you come from different directions, one would think. On the main flowsheet, the current state is used as the initial guess for the next solution. If you do this in two separate documents, converging to the same input conditions from different sides, you could compare the two solutions.