1

PAT 2.0 AWS capping at 50 runs

I am trying to run ~2000 models through PAT. I have set it up using the Algorithmic method DOE and it works for small batches(~50 models), but fails at anything larger. I have the server and worker setup to be c3.8xlarge giving me plently of cores for the simulations. The issue is it just doesn't seem to que up the runs when there are greater than 50 design alternatives. What can I do to fix this? Below are the two error messages I found in the logs:

Error 1:

/opt/openstudio/server/lib/analysis_library/doe.rb failed with voidEval failed: Packet[cmd=2130771970,len=<nil>, con='<nil>', status=error:'Unknown variable/method'(127)]

Error 2:

Error in contr.poly(nlevels[i], scores = factor.names[[i]]) : 
orthogonal polynomials cannot be represented accurately enough for 1530 degrees of freedom
Alex Bennett's avatar
327
Alex Bennett
asked 2017-10-06 09:17:51 -0500
Julien Marrec's avatar
29.7k
Julien Marrec
updated 2017-10-09 02:56:14 -0500
edit flag offensive 0 remove flag close merge delete

Comments

What specific version of PAT are you using? Try upgrading to OpenStudio 2.3.0 and see if you have the same issue. You also tagged this as analysis spreadsheet. Are you using that or PAT?

David Goldwasser's avatar David Goldwasser (2017-10-06 09:44:28 -0500) edit

I am using PAT 2.2.0, so I will try updating. Did the latest updates address issues like the one seen above? And sorry was just tagging similar/relevant items.

Alex Bennett's avatar Alex Bennett (2017-10-06 09:50:18 -0500) edit

Same issue with the new version, any ideas?

Alex Bennett's avatar Alex Bennett (2017-10-06 13:31:19 -0500) edit
add a comment see more comments

1 Answer

5

The R package used to generate the full factorial design is the DoE.Base package. It calls contr.poly to help generate the design of experiments. here is the line thats stopping your analysis.

This type of analysis requires at least 2 variables to work (with unique values).

BrianLBall's avatar
2k
BrianLBall
answered 2017-10-13 09:49:41 -0500
edit flag offensive 0 remove flag delete link

Comments

Thanks, the issue lay in the non unique values for my second variable. It runs fine now.

Alex Bennett's avatar Alex Bennett (2017-10-13 10:25:22 -0500) edit
add a comment see more comments