Examine reason for solver failures in Flash workshop example

[andrewlee94]

See IDAES/examples-dev#142 and IDAES/examples-dev#144.

Something strange is happening with the repeated solves of the Flash unit in the first workshop example. From examining the various model states throughout the example, I had the following observations:

1. There are minor differences in the final solution (i.e. last three to four places in the floats) for solutions that otherwise should be the same. Notably, I saw two different model states: the first deterministic solution, and the solution after re-initializing the model from any other state. However, the fixed variables are identical in all cases, so any variation must be due to scaling?
2. Re-initializing the model prior to generating lots works, so that suggests that is the more stable solution. Thus, something is not quite right with the first deterministic solution?

This needs to be investigated further - my first impression is to try scaling the model better and see what happens (which will require the scaling tools).

[andrewlee94]

@carldlaird I know it is unlikely that you will be able to give much insight without looking at the model, but do you have any ideas?

[andrewlee94]

This is definitely still relevant.

[carldlaird]

@andrewlee94 Ping me over email and let's setup a time to discuss.

[dangunter]

Sorry @andrewlee94 I just assumed given the dates that this had been resolved or was in wont-fix state

[andrewlee94]

This is old enough that I am going to assume it is either fixed or no longer matters.