Modelica.Blocks.Math.LinearDependency has strange formulation

[modelica-trac-importer]

Reported by Joe Riel on 1 Jun 2015 14:58 UTC
The relatively new block LinearDependency seems strange. It's defining equation is y = y0*(1 + k1*u1 + k2*u2);. y0 is a parameter; its description says it is an initial value, which makes no sense. It is a scale factor (with a useless default of 0). However, if its value is nonzero, then one cannot produce an output with a non-zero offset. I'm thinking the block would be slightly more useful if the defining equation were y = y0 + k1*u1 + k2*u2.

---

Migrated-From: https://trac.modelica.org/Modelica/ticket/1724

[modelica-trac-importer]

Comment by Joe Riel on 1 Jun 2015 15:32 UTC
A correction. A nonzero value of y0, which is practically necessary to have a useful block, means the offset must be nonzero. With the proposed change, the offset is independent of the gains.

[modelica-trac-importer]

Modified by dietmarw on 1 Jun 2015 20:14 UTC

[modelica-trac-importer]

Modified by beutlich on 18 Aug 2015 12:08 UTC

[modelica-trac-importer]

Comment by beutlich on 31 Aug 2015 09:29 UTC
There is only one component of LinearDependency in the MSL, however the Real output Modelica.Blocks.Examples.RealNetwork1.linearDependency1.y is not used in the example model.

[modelica-trac-importer]

Comment by otter on 28 Sep 2015 11:45 UTC
The proposal would lead to a non-backwards compatible change of the library (which is not possible with MSL 3.2.1 build 4). Therefore, the milestone is moved to 3.2.2.

To avoid a non-backwards compatible change in 3.2.2, I would suggest to just improve the documentation of this block (and the use case for which it was designed).

[modelica-trac-importer]

Modified by otter on 11 Dec 2015 14:20 UTC

[modelica-trac-importer]

Comment by ahaumer on 18 Dec 2015 12:05 UTC
IMHO the implementation is erroneous and will be changed when a conversion script is applicable. This is stated in the documenation of the block.

[modelica-trac-importer]

Comment by beutlich on 20 Dec 2015 20:45 UTC
I reverted the changes from 811c96e of Math.mo with 8bd005c since

• it is not clear if this change went in by design or by side-effect (no commit message was given)
• the comment used the undesired h3 headline
• the formula in the comment was invalid.

Please revise again for MSL 3.2.2. For that reason I changed the milestone back to MSL 3.2.2 which should be reset again once the comment is without errors. Sorry for the extra work.

[modelica-trac-importer]

Comment by ahaumer on 23 Dec 2015 18:28 UTC
Please, do not revert - give a comment what's wrong in your opinion.

Once again: IMHO the implementation is wrong and will be corrected in the next major version (conversion script needed; ticket remains open). This is stated in the documentation (92389bf).

[beutlich]

Point to @JoeRiel.

[beutlich]

My proposal #2886 was rejected by @MartinOtter .

@AHaumer @MartinOtter Please provide an alternative fix!

[JoeRiel]

On Fri, 09 Aug 2019 06:36:14 -0700 Thomas Beutlich ***@***.***> wrote: My proposal #2886 was rejected by @MartinOtter . @AHaumer @MartinOtter Please provide an alternative fix!

Hmm. The original is definitely wrong and completely non intuitive. I understand the reluctance to change a model, as doing so will definitely change any models that use it, but suspect it was never/rarely used given its weirdness. Maybe name the fixed model LinearDependent, or TwoInputLinear, or something like that.

…

-- Joe Riel

[beutlich]

#3088 is my second attempt to fix this issue.