CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Siemens (
-   -   Datacenter Modeling (

ds7202 March 8, 2013 02:34

Datacenter Modeling

For my independent study, I've been attempting to generate a CFD model of one of the datacenters on campus, hopefully to use as an analysis tool to optimize cooling.

I have been pulling my hair out on this! None of the professors in my department have ever done anything like this (including the one overseeing my study), and I cannot find any resources online to assist me.

I've worked the problem out and I know what I need to do. I'm trying right now to just assemble a simple model of a single rack in a room. I have one tile set as a velocity inlet, and a tile in the corner of the room as a pressure outlet.

I have modeled the geometry of the room as a 4x4x4-m cube. I created small cubes on the bottom of the large cube, then used the imprint feature to imprint the outline of the cube faces onto the 'floor' of the room, then subtract the cube to create separate faces for the inlet and outlet. Then, I extruded a sketch to create a rectangular cube to simulate the server rack.

I set the inlet of the rack as a mass-flow inlet. I need to set the outlet of the rack as a fan outlet, but I cannot find an option for this. I went with split-flow outlet (i think that's what it's called?).

I created two separate regions- one for the room, and one for the server rack. This is where I initially thought my problem lied- I was having trouble interfacing the two regions, and the boundaries of those regions. However, I decided to make the problem even simpler, and turned the server itself into just a set of walls. So, the airflow from the inlet will just flow around the wall and leave via the outlet.

When I meshed, I got a wonky mesh surface- the bottom of the 'server rack' looked like it had been eaten away with acid. I used the surface remesher and polyhedral mesher. I played around with different mesh models, only with the same (or worse) result. I went back to the CAD drawing and did a boolean operation to subtract the server rack from the room, with the room being the target. That solved my meshing issue.

Then, I tried to solve. I received a floating point exception error. So, I deleted the rack, and tried to solve again. I receive the same error.

Am I violating the rules of CCM by placing an inlet in the middle of a room and having a pressure outlet in the corner of the room?

I've only ever used this software to run simulations of flow over objects- such as aircraft, spheres, etc. Those seem pretty straightforward compared to this- they have a velocity inlet in the freestream, a symmetry plane, and the rest of the C.V. surfaces are also freestream.

I'm sort of laughing, because the CD-Adapco rep who gave the presentation at my university when we first acquired this software told me "yeah, we do datacenter modeling all the time!" Yet, I haven't seen any help topics on the subject, nor have I seen any examples. Their support have been pretty unhelpful too.

If anybody could point me in the right direction, that would be wonderful. It's half semester, and I'm at my wits end here- if I can't get this figured out, I'm not going to have a good time, haha.


ds7202 March 8, 2013 02:42

I should also note that Star-CCM+ is the only software I have access to. CoolSim and other softwares developed solely for this seem to be out of the question, however, if anybody knows how I may access a student (hint:free) license for one of these, that would be awesome too.

cwl April 29, 2013 08:52

Briefly - i've solved (as an exercise for myself) a problem that looks like the one you've described: two boxes - RACK (air in, heater air out) and Fancoil (air in, cooled air out).

Main idea: better use not split-flow, but Mass Flow Inlet - for both intake and outtake of RACK and cooling device, just with positive and negative mass flow boundary condition.

// PM or Skype ( if you need help - i rarely read forum %>

All times are GMT -4. The time now is 21:55.