AspenPlus interface with Python

[BaileyLadd01]

@YouMayCallMeJesus I was wondering if your code had any nice ways to detect if Aspen throws up an error. So far I have been using an error detection node from the aspen tree, but it doesn't seem to work as intended. I would appreciate any help with figuring out how to set up an effective error detection system for my project.

[YouMayCallMeJesus]

Hey Vincent. The only way I have found is using that node. This only deals with normal simulation errors and not with program wide crashed. Sometimes the whole Aspen just freezes. I used to detect this by taking some stream, changing it's value, then reading the value you just changed. If you get a value then the reading part works, if you get the value that you wrote then you know that both reading and writing works. Now if this takes more then X seconds I then know that neither of them are working. In this case I was unable to actually do anything. So I automatically killed the Aspen program, os.kill (I think?) and restarted it from a safe backup. Op do 26 sep. 2024 14:02 schreef Vincent Bailey Ladd < ***@***.***>:

…

@YouMayCallMeJesus <https://github.com/YouMayCallMeJesus> I was wondering if your code had any nice ways to detect if Aspen throws up an error. So far I have been using an error detection node from the aspen tree, but it doesn't seem to work as intended. I would appreciate any help with figuring out how to set up an effective error detection system for my project. — Reply to this email directly, view it on GitHub <#2>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AUUV43MF35ZDUSCH37LEAOTZYPZTRAVCNFSM6AAAAABO43IR4CVHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZXGIZTSMJRGQ> . You are receiving this because you were mentioned.Message ID: ***@***.*** com>

[BaileyLadd01]

Hi @YouMayCallMeJesus

Unfortunately for my case I wouldn't be able to try and access a stream to detect an error. Typically my errors are infinite convergence errors. What made you choose to use os.kill() instead of aspen.Close() which uses the COM object?

In my experience of closing aspen down using the task manager (which I think is similar to os.kill() ) the aspen file typically opens as a .bkp after instead of a .apw file as intended.

[YouMayCallMeJesus]

Hey Vincent, To be honest I didn't try the Aspen.Close since I was unable to close the application without using the task manager when using it manually. But you are right it might work. I found the same thing with the bkp file being the only thing available (unless you regularly save it) but that was not a big deal in my application. I was just scrapping data to train some RL. If it crashed then I just wanted it to detect it and then restart. If I lost a bit of progress that didn't matter too much. Hmm. Isn't there a convergence node? I am not sure but I though there was something like that. Good luck with your work Richard Op vr 27 sep. 2024 10:13 schreef

…

Hi @YouMayCallMeJesus <https://github.com/YouMayCallMeJesus> Unfortunately for my case I wouldn't be able to try and access a stream to detect an error. Typically my errors are infinite convergence errors. What made you choose to use os.kill() instead of aspen.Close() which uses the COM object? In my experience of closing aspen down using the task manager (which I think is similar to os.kill() ) the aspen file typically opens as a .bkp after instead of a .apw file as intended. — Reply to this email directly, view it on GitHub <#2 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AUUV43IPPGLTFJEOA6WSNCLZYUHSDAVCNFSM6AAAAABO43IR4CVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTANZXGIZTSNY> . You are receiving this because you were mentioned.Message ID: <YouMayCallMeJesus/AspenPlus-Python-Interface/repo-discussions/2/comments/10772397 @github.com>

[BaileyLadd01]

Hey Richard,

I have been using Aspen.Close(), but it doesn't seem to close down the dispatched aspen in the task manager. I had some code that worked in ensuring .bkp files were never opened, but it also opened an extra instance of aspen every time (eating up my RAM).

I have been trying to develop some code that will detect a crash, and store the simulation set up that caused the problem. The stored results are then skipped (along with all the previous successful results), meaning that even if it crashes, the code restarts that particular instance of aspen (as I am multiprocessing across 16 simulations), and continues. But that is still being developed.

I haven't found a convergence node, but I know that in the .his output file after each simulation there is mention to if the simulation converged. So it stands to reason that there is a convergence node that could be accessed. Do you know where it is located on the tree?

Thank you for your help, you have some very comprehensive code for accessing all of aspen plus in python!

Kind regards,
Vincent

[YouMayCallMeJesus]

Ahh ok. So you are using multiple aspen programs open at once or are you using some kind of Aspen internal multiprocessing? I have only ever run multiple aspen programs in parallel such that they had different names in the task manager. Then when I detect a crash via the runtime limit I just killed it and restarted it. About the inputs being remembered when it crashes: in my testing the crashes seem pretty random. Repeating a simulation that crashed for me did not result in another crash. I found that adding some delays like: stream.temp=5 wait(1sec) sim.run Reduced the number of crashes... But it comes at a cost of making everything slower. I decided to just not use it in the end. Convergence node: Sadly the Aspen team did not add everything to the tree that one would wish for. I have first attempted to use the text files, to read them and then extract data but it was too much of a pain and I gave up. I just thought about a bit of a hacked way to detect the convergence. If you find the convergence time distribution for your type of problem, you could define a maximum amount of time you are willing to wait for a result. Then you just use: self.AspenSimulation.Stop() to stop the simulation and continue to the next one(and remembering which one took too long. This is almost as if you had detected a convergence. It might be that these difficult simulations are the most interesting but that depends on your application. Thanks a lot man! If you get any nice results I would be happy to reference your GitHub to show people some more advanced applications. Cheers mate Richard Op za 28 sep. 2024 16:00 schreef Vincent Bailey Ladd < ***@***.***>:

…

Hey Richard, I have been using Aspen.Close(), but it doesn't seem to close down the dispatched aspen in the task manager. I had some code that worked in ensuring .bkp files were never opened, but it also opened an extra instance of aspen every time (eating up my RAM). I have been trying to develop some code that will detect a crash, and store the simulation set up that caused the problem. The stored results are then skipped (along with all the previous successful results), meaning that even if it crashes, the code restarts that particular instance of aspen (as I am multiprocessing across 16 simulations), and continues. But that is still being developed. I haven't found a convergence node, but I know that in the .his output file after each simulation there is mention to if the simulation converged. So it stands to reason that there is a convergence node that could be accessed. Do you know where it is located on the tree? Thank you for your help, you have some very comprehensive code for accessing all of aspen plus in python! Kind regards, Vincent — Reply to this email directly, view it on GitHub <#2 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AUUV43PBNU6WYQRJYJ5RB6DZY2Y6VAVCNFSM6AAAAABO43IR4CVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTANZYGM4DSMI> . You are receiving this because you were mentioned.Message ID: <YouMayCallMeJesus/AspenPlus-Python-Interface/repo-discussions/2/comments/10783891 @github.com>

[BaileyLadd01]

Hi Richard,

I am opening multiple aspen applications, I don't think the aspen devs would have built in any way to run multiple simulations.

The set up of my work is as follows:

16 amine capture plant simulations (difference being flowrates, they are bespoke so each needed their own simulation file).

Code to cycle through every possible packing type, vendor, material, and dimension (both stipper and absorber column). (If an error is detected in this stage, the code would note down the configuration that caused the error then skip to the next one after restarting aspen).

Evolutionary learning method to optimise tuning factors.

I didn't realise your code has the capability to open multiple aspen instances. But it would make sense since it is so comprehensive.

Could you signpost me to your method? I'd be intrigued to see how you got round some of the pit falls, it took me about 2 weeks to solve the constant RCP errors.

I am planning on presenting my method at a conference in November (if all goes well), and after that I will publish the results. I will of course, reference you and any others who have attempted to automate aspen. After that I will make my github repo public for people to use and build upon.

Hope all is well!

Vincent