The Sensitivity Analysis utility is really simple and useful. The Optimizer is a little bit more complicated, but very, very powerful once you understand how it works.
Daniel
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2017-05-10
Last edit: Anonymous 2021-12-15
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
This is within the expected behavior. Probably the newton update estimated an unfeasible value for the HTC and the new try couldn't get a working solution with a starting point equal to zero.
The simultaneous adjust solver calculates uses numerical derivatives to go towards the solution. If the variable is different than zero, a perturbed value of 1.01 times the original value is used. If the variable is zero, the new estimated value is 0.01, which may not work for all cases.
As I've recommended you before, it is better to use the Optimizer tool, where you can define variable bounds very easily. The simultaneous adjust solver could be considered as a simplified optimizer, where it doesn't have information about the variables except for their current values, and the flowsheet model is considered ultimately a "black box".
Daniel
Last edit: SourceForge Support 2021-12-20
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Last edit: Anonymous 2021-12-15
Currently, the tolerance of the global solver is compared against the sum of squared errors: https://github.com/DanWBR/dwsim4/blob/master/DWSIM.FlowsheetSolver/FlowsheetSolver.vb#L1477
Maybe I could get the tolerance setting from each object instead...
Last edit: Anonymous 2021-12-15
<redacted>, could you please attach a sample so I can test the code with the new convergence criteria? Thanks in advance.Last edit: SourceForge Support 2021-12-20
Last edit: Anonymous 2021-12-15
Thanks
<redacted>. It did the job.You can try Update 8 now. Besides the updated convergence code, I've put the tolerance input field in the Adjust editor directly.
I think that, since your model is becoming more elaborate, you could give the Optimizer a shot...
Regards
Daniel
Last edit: SourceForge Support 2021-12-20
Last edit: Anonymous 2021-12-15
I must have forgotten to include some file on the update pack. I'll fix that in a few hours.
<redacted>, please check if Update 9 fixed the issue. If you'd overwritten the file, you may have to redo the connections. I'm sorry for that.Last edit: SourceForge Support 2021-12-20
Last edit: Anonymous 2021-12-15
Glad to know that.
The Sensitivity Analysis utility is really simple and useful. The Optimizer is a little bit more complicated, but very, very powerful once you understand how it works.
Daniel
Last edit: Anonymous 2021-12-15
This is a result of saving the file with Update 8... you could try selecting a new set of units.
Last edit: Anonymous 2021-12-15
Last edit: Anonymous 2021-12-15
Hi
<redacted>,This is within the expected behavior. Probably the newton update estimated an unfeasible value for the HTC and the new try couldn't get a working solution with a starting point equal to zero.
The simultaneous adjust solver calculates uses numerical derivatives to go towards the solution. If the variable is different than zero, a perturbed value of 1.01 times the original value is used. If the variable is zero, the new estimated value is 0.01, which may not work for all cases.
As I've recommended you before, it is better to use the Optimizer tool, where you can define variable bounds very easily. The simultaneous adjust solver could be considered as a simplified optimizer, where it doesn't have information about the variables except for their current values, and the flowsheet model is considered ultimately a "black box".
Daniel
Last edit: SourceForge Support 2021-12-20