Performance improvement for IDEAS.Airflow.Multizone.BaseClasses.TwoWayFlowElement

[kldjonge]

I noticed some discrepencies in the simulation results between mBA_flow and VBA_flow*rho_a1_inflow at low flows when using the operable door model (IDEAS.Airflow.Multizone.DoorDiscretizedOperable). If VAB_flow was exactly zero, mAB_flow wasn't while it should be. So somehow mBA fluctuated around zero while it should have been zero and, consequently, the 'error control' on that flow limited the stepsize a lot.

Upon further investigation, this was fixed by explicitly linking Vflow*rho to the port:

port_a1.m_flow = VAB_flow*rho_a1_inflow;
port_a2.m_flow = VBA_flow*rho_a2_inflow;

For me, this decreased simulation time significantly. Since the door models all actually calculate volume flows at top level, it will most-likely improve the performance in general for all models using this class.

A larger snippet:

  mAB_flow=VAB_flow*rho_a1_inflow;
  mBA_flow=VBA_flow*rho_a2_inflow;
  // Average velocity (using the whole orifice area)
  vAB = VAB_flow/A;
  vBA = VBA_flow/A;

  port_a1.m_flow = VAB_flow*rho_a1_inflow;
  port_a2.m_flow = VBA_flow*rho_a2_inflow;

  // Energy balance (no storage, no heat loss/gain)
  port_a1.h_outflow = inStream(port_b1.h_outflow);
  port_b1.h_outflow = inStream(port_a1.h_outflow);``

[mwetter]

@kldjonge : Can you put the changes on a branch so its clear what exactly is changed, and also explain how mABt_flow is taken into account in port_a1.m_flow. From the limited information, it appears like your changes neglect this contribution, which could explain the faster computing time.

[kldjonge]

I will do so soon. I don't have it before me now, but I can't recall a mABt_flow variable in the model.

[kldjonge]

I see what you mean now, but the change I propose will only effect the discretized door models.

[mwetter]

This model still has peaks in a trajectory that should be smooth, as reported in lbl-srg/modelica-buildings#4360

Setting unbounded=true for the states mExcAB and mExcBA solves the problem. These states are indeed unbounded.