Further inStream optimizations

[Francesco Casella]

The simplifyInStream module of the backend currently implements the substitutions listed in ticket:3885#comment:15

• inStream introduces positiveMax(cref, eps) and positiveMax(-cref, eps)
• simplify positiveMax in Backend after removeSimpleEquation and EvaluateParameter and before IndexReduction
• positiveMax(cref, eps) = cref where variable is constant >= 0
• positiveMax(-cref, eps) = -cref where variable is constant <= 0
• positiveMax(cref, eps) = 0 if variable(cref).max <= 0
• positiveMax(-cref, eps) = 0 if variable(cref).min >= 0

The idea of the last two substitutions is that if a fluid never flows out of the stream connector, its contribution is zero. In case of components with only one direction of flow, one can set min=0 at the inlet to activate this optimization.

In fact, when dealing with components without flow reversal, not only the fluid never flows out of the inlet, hence inlet.m_flow.min = 0, but also the fluid never flows in the outlet, hence outlet.m_flow.max = 0. This fact can be stated by setting the max attribute of the outlet port mass flow rate to zero. If this property holds, then

• positiveMax(cref, eps) = cref if variable(cref).min >= 0 test: positiveMax(1, eps) = 1
• positiveMax(-cref, eps) = -cref if variable(cref).max <= 0 test: positiveMax(-(-1), eps) = 1

These additional optimizations would allows to remove unnecessary positiveMax function calls, hence eliminating unnecessary hard nonlinearities from the model equations.

@vitalij, if you are willing to implement this, I can prepare the tests ASAP.

[Vitalij Ruge]

maybe fixed with ​PR2624.

[Vitalij Ruge]

@casella

if you are willing to implement this, I can prepare the tests ASAP.

that would be great!

[Vitalij Ruge]

There was a side effect.

e.g. for BuildingSystems_Fluid_Examples_FlowSystem_Simplified6

tabsNorth2\[8\].port_b.h_outflow = (pmpNorth.m_flow*h)/(pmpNorth.m_flow +0.0)

where we have pmpNorth.m_flow at the place of max(pmpNorth.m_flow,eps).
No other module simplified it to the alias equation

tabsNorth2\[8\].port_b.h_outflow = h

[Vitalij Ruge]

With ​PR2626 do simplify of expressions which were change in simplifyInStream.

[Francesco Casella]

Replying to vitalij:

@casella

if you are willing to implement this, I can prepare the tests ASAP.

that would be great!

See attached updated package, Test12 and Test13. I wrote the expected optdaedump output in the documentation section, unfortunately I cannot test this myself now as the Windows nightly is currently broken.

[Vitalij Ruge]

I get a dismatch for Test13

pipe2.outlet.h_outflow = (max(-pipe3.inlet.m_flow, 1e-07) * sink3.h + pipe1.inlet.m_flow * source.h) / (max(-pipe3.inlet.m_flow, 1e-07) + pipe1.inlet.m_flow) 

where

1: pipe3.inlet.m_flow:VARIABLE(flow=true min = max(-1000000000.0, -1000000000.0) )  type: Real
6: pipe1.inlet.m_flow:VARIABLE(flow=true min = max(0.0, 0.0) )  type: Real

[Francesco Casella]

@vitalij: my fault, what I wrote in the documentation was bogus. The output you reported in comment:9 is indeed correct.

I updated the attached library with the correct expected output, please update the test files on GitHub and close the ticket when you're done.

[Vitalij Ruge]

new tests were added with ​PR1021

[Francesco Casella]

@vitalij, I apologize, there was still a mistake in the documentation of Test13 that I catched just now. I have updated the attchment, do you mind updating it on GitHub?

Thanks!

[Vitalij Ruge]

fixed with ​PR1026

[Francesco Casella]

Thanks @vitalij!