00:04
jaguart joined
00:08
jaguart left
01:40
Manifest0 left
06:04
jaguart joined
06:08
jaguart left
06:21
kjp left
06:22
kjp joined
06:24
kjp left,
kjp joined
|
|||
winfredraj | Hello all | 07:05 | |
do we have anyone working on Modelica with raku? | |||
antononcube | @winfredraj Interesting question -- I recall there are a few packages for/about Modelica. | 07:31 | |
raku.land/github:albastev/Grammar::Modelica | 07:32 | ||
winfredraj | I did have see this now, last entry was 2018 | 07:44 | |
I have just started trying stuff of Maple Sim for some simulation which I then want to interface with the PLC (in simulation mode) using FMU or functional mockups | 07:46 | ||
was wondering if I could use Raku for the glue code | |||
I will have a look to see if I can figure something up | 07:50 | ||
antononcube | Anyone still using Maple!? | 07:53 | |
I tried and tried to use Modelica directly, or through SystemModeler. But ultimately I always end up using Mathematica / Wolfram Language (WL). The tight integration WL and SystemModeler helps, but at least in my use cases, SystemModeler (and Modelica) are just another representation language, useful to communicate with others, but that much for the actual modeling. | 07:56 | ||
Now, with Raku I do facilitate the creation of System Dynamics models, but the code that the corresponding Raku package generates is not Modelica, it is WL. Although, Modelica can be one or the translation targets. | 07:58 | ||
@winfredraj Or in other words -- I use same/similar tools as you but of different camp. | 07:59 | ||
winfredraj | wow | 08:01 | |
I have not zeroed in on Maple as the price is a bit on the higher side. SystemModeler seems resaonably priced but Maple does have some web handling modules which I may have to model myself with SystemModeler | 08:02 | ||
Maplesim also proven to work or has already been tested with B&R PLCs | 08:03 | ||
But if I can generate the FMU using SystemModeler, that should not matter as the PLC can import co simulation FMUs | 08:04 | ||
what sort of Models do you work on, I thought you were into LLMs | 08:05 | ||
antononcube | FMU == Florida Memorial University ? | ||
winfredraj | Functional Mockups | ||
antononcube | Well, I do tell people that LLMs can be used instead of SystemModeler. | ||
For example, with LLMs I can generate System Dynamics (SD) models diagram (in Mermaid-JS.) From those diagrams I can generate SD equations. And vice-versa -- from the equations I can generate the Mermaid-JS diagrams. Etc. | 08:07 | ||
winfredraj | ok, so you give the data and "teach" the model instead of using the old school route of writing and formulating the models | ||
? | 08:08 | ||
antononcube | No. | ||
I have not tried that, although it might be possible. | 08:09 | ||
I think I can find / bring a few examples about the SD models generation I do with LLMs. | 08:10 | ||
BTW, those LLM generations work out of the box in, say, less than 20% of the cases. For simpler models, I would sagy, less than 50%. | 08:11 | ||
But the LLM generations are good initial versions of the models, and / or save time spelling them out. | 08:12 | ||
winfredraj | Where can I see some examples of the SD models you are generating? | 08:13 | |
antononcube | Here is the first -- not a good one mind you, just an example: «LLM over RUSI's "The Attritional Art of War" article», community.wolfram.com/groups/-/m/t/3149566 | 08:14 | |
See the section "System Dynamics model". Basically, the LLM gave a metamodel, a model for the model, not the model itself. | 08:15 | ||
But, LLMs can be forced to produce an actual SD model with better prompting. | 08:16 | ||
winfredraj | looks cool, have you done any physical or engineering models as well? | 08:17 | |
anything in WOlfram would be possible? | 08:18 | ||
if yes, I should be able to articulate my thoughts in writing to create models of motors, circuits etc etc (what is available in Wolfram - and they have a bunch of stuff) | 08:19 | ||
right? | 08:20 | ||
antononcube | Yes, SystemModeler is "littered" with these kind of examples. | 08:21 | |
winfredraj | wow cool | ||
antononcube | Here is my other example : www.youtube.com/watch?v=852vMS_6Qaw&t=1349s | 08:22 | |
winfredraj | so the idea is too make the process of Making the model simpler by basically "speaking it into existence" | 08:23 | |
Ill listen to this in a bit | |||
antononcube | That is one way of doing it (of thinking of doing it.) But, it really depends of the problem area and the level of understanding it by the modeler. | 08:25 | |
I am much more interested of facilitating fast creation of SD models by people who know its elements. | 08:27 | ||
But gradual building of model is of course one the typical, even primary use cases. | 08:28 | ||
10:04
jaguart joined
10:09
jaguart left
11:36
Manifest0 joined
19:20
Tirifto left
19:49
Tirifto joined
19:56
Tirifto left
19:59
Tirifto joined
23:55
Manifest0 left
|