Integer equation lost by NF

[Francesco Casella]

Please check ​ThermoPower.PowerPlants.ElectricGeneratorGroup.Tests.Test_Generator_SE. When flattening the model with the NF, the equation

steamTurbineStodola.steamState_iso.phase = 0;

does not show up in the flat model, as it does instead with the old FE.

[Per Östlund]

The NF generates the equation

steamTurbineStodola.steamState_iso =
  Test_Generator_SE_total.steamTurbineStodola.Medium.setState_ph(100000.0, 100000.0, 0);

instead, since steamTurbineStodola.steamState_iso isn't a parameter and thus doesn't require the function call to be evaluated. It's not clear whether this is the reason for the failure, but I would guess not since record equations are usually counted correctly.

[Per Östlund]

After 6654660 the NF now evaluates the mentioned function call, and instead generates the record equation (where the rhs still looks like a function call, but is actually a record expression):

steamTurbineStodola.steamState_iso = 
  valveHP.Medium.ThermodynamicState(0, 100000.0, 997.3369503211875, 296.9948190753248, 100000.0);

This doesn't affect the outcome at all, the backend still complains about the model being under-determined. I also tried splitting the equation like the OF does, which made no difference either. So the problem must lie elsewhere, but comparing the flat model from the NF and the OF I can't spot any significant differences.

[Francesco Casella]

I had a look at the output of -d=optdaedump when the NF is used. Until the pre-optimization module evalFunc (simulation) is invoked, the system is balanced. When this function is called, the set of equations is then reduced from 35 to 30 equations, but only 4 variables, namely

21: valveHP.w:VARIABLE(flow=false min = max(-100000.0, max(-1e+060, max(-1e+060, max(-100000.0, max(-1e+060, max(-1e+060, max(-1e+060, -100000.0))))))) max = min(100000.0, min(100000.0, min(100000.0, min(100000.0, min(100000.0, min(100000.0, min(100000.0, 100000.0))))))) start = steamTurbineStodola.wstart unit = "kg/s" )  "Mass flow rate" type: Real  
25: steamTurbineStodola.Pm:VARIABLE(flow=false unit = "W" )  "Mechanical power input" type: Real  
31: steamTurbineStodola.steamState_in.p:VARIABLE(flow=false min = 611.657 max = 100000000.0 start = 5000000.0 unit = "Pa" nominal = 1000000.0 )  "Pressure" type: Real  
32: steamTurbineStodola.steamState_in.T:VARIABLE(flow=false min = 273.15 max = 2273.15 start = 500.0 unit = "K" nominal = 500.0 )  "Temperature" type: Real  

are removed from the set of unknowns. If all the equations which are identical before and after the equations are omitted, what is left is:

Before (9 equations):

1/1 (1): valveHP.w = homotopy(steamTurbineStodola.Kt * sqrt(steamTurbineStodola.steamState_in.p * steamTurbineStodola.steamState_in.d) * ThermoPower.Functions.sqrtReg(1.0 - (1.0 / steamTurbineStodola.PR) ^ 2.0, 0.01), steamTurbineStodola.wnom / (steamTurbineStodola.partialArc_nom * steamTurbineStodola.pnom) * steamTurbineStodola.steamState_in.p)   [dynamic |0|0|0|0|] 
5/5 (1): steamTurbineStodola.Pm = steamTurbineStodola.eta_mech * valveHP.w * (sourceP.h - steamTurbineStodola.hout)   [dynamic |0|0|0|0|] 
6/6 (1): steamTurbineStodola.Pm = (-steamTurbineStodola.tau) * generator_SE.omega   [dynamic |0|0|0|0|] 
7/7 (5): steamTurbineStodola.steamState_in = ThermoPower.PowerPlants.ElectricGeneratorGroup.Tests.Test_Generator_SE.valveHP.Medium.setState_ph(steamTurbineStodola.pin, sourceP.h, 0)   [dynamic |0|0|0|0|] 
13/17 (1): valveHP.w = homotopy(ThermoPower.PowerPlants.ElectricGeneratorGroup.Tests.Test_Generator_SE.valveHP.FlowChar(valveHP.theta_act) * valveHP.Av * valveHP.Y * sqrt(valveHP.rho) * smooth(0, if valveHP.xs >= 0.0 then ThermoPower.Water.ValveVap.sqrtR(valveHP.p * valveHP.xs, valveHP.b * valveHP.dpnom) else 0.0), valveHP.theta_act / valveHP.thetanom * valveHP.wnom / valveHP.dpnom * valveHP.dp)   [dynamic |0|0|0|0|] 

After (5 equations):

4/4 (1): 0.0 = (-steamTurbineStodola.tau) * generator_SE.omega   [dynamic |0|0|0|0|] 
5/5 (2): (steamTurbineStodola.steamState_in.phase, steamTurbineStodola.steamState_in.h) = hermoPower.PowerPlants.ElectricGeneratorGroup.Tests.Test_Generator_SE.valveHP.Medium.setState_ph_eval3(steamTurbineStodola.pin, sourceP.h, 0)   [dynamic |0|0|0|0|] 
11/12 (1): 0.0 = homotopy(ThermoPower.PowerPlants.ElectricGeneratorGroup.Tests.Test_Generator_SE.valveHP.FlowChar(valveHP.theta_act) * valveHP.Av * valveHP.Y * sqrt(valveHP.rho) * smooth(0, if valveHP.xs >= 0.0 then ThermoPower.Water.ValveVap.sqrtR(valveHP.p * valveHP.xs, valveHP.b * valveHP.dpnom) else 0.0), valveHP.theta_act / valveHP.thetanom * valveHP.wnom / valveHP.dpnom * valveHP.dp)   [dynamic |0|0|0|0|] 

Now, I'm not sure where does the setState_ph_eval3() come from, but it seems that there is some issue with functions involving record outputs, losing some pieces along the road.

Note that the the function call to setState_ph, when fed by the old FE, is split by the first back-end function normalInlineFunction into:

20/20 (1): steamTurbineStodola.steamState_in.p = steamTurbineStodola.pin   [dynamic |0|0|0|0|] 
21/21 (1): steamTurbineStodola.steamState_in.T = Modelica.Media.Water.IF97_Utilities.T_props_ph(steamTurbineStodola.pin, valveHP.outlet.h_outflow, Modelica.Media.Water.IF97_Utilities.waterBaseProp_ph(steamTurbineStodola.pin, valveHP.outlet.h_outflow, 0, 0))   [dynamic |0|0|0|0|] 
22/22 (1): steamTurbineStodola.steamState_in.d = Modelica.Media.Water.IF97_Utilities.rho_props_ph(steamTurbineStodola.pin, valveHP.outlet.h_outflow, Modelica.Media.Water.IF97_Utilities.waterBaseProp_ph(steamTurbineStodola.pin, valveHP.outlet.h_outflow, 0, 0))   [dynamic |0|0|0|0|] 
23/23 (1): steamTurbineStodola.steamState_in.h = valveHP.outlet.h_outflow   [dynamic |0|0|0|0|] 
24/24 (1): steamTurbineStodola.steamState_in.phase = 0   [dynamic |0|0|0|0|] 

while the same back-end module fed by the NF returns

31/35 (5): steamTurbineStodola.steamState_in = ThermoPower.PowerPlants.ElectricGeneratorGroup.Tests.Test_Generator_SE.valveHP.Medium.setState_ph(steamTurbineStodola.pin, valveHP.outlet.h_outflow, 0)}}}

I guess there's something wrong with the way the NF processes this function call - maybe the Inline annotation is lost?

[Francesco Casella]

This model now simulates successfully