Opened 5 years ago

Last modified 3 years ago

#5574 new defect

Failed to solve linear system of equations because of wrong dynamic state selection

Reported by: ahaumer@… Owned by: Lennart Ochel
Priority: blocker Milestone: 1.19.0
Component: Backend Version: v1.13.2
Keywords: linear system of equations solver Cc: christian@…

Description

Dear OM-team,

I hope I have assigned correctly the solver problem to Backend.
I developed the enclosed small library about DC braking of induction machines for my lectures.
There are 4 variants / layouts:
Y3 and Y2 simulate without complaints.
D2 and D3 fail (but they simulate in Dymola 2020).
I have no idea why the linear solver fails - could be a problem of index reduction?
Please analyze the problem:

  • This should improve OM quality.
  • This would enable me to let students simulate the applications.

In case of any questions, do not hesitate to contact me.

Many thanks & best regards,
Anton Haumer

Attachments (1)

IMC_DCBraking.mo (10.2 KB ) - added by ahaumer@… 5 years ago.

Download all attachments as: .zip

Change History (11)

by ahaumer@…, 5 years ago

Attachment: IMC_DCBraking.mo added

comment:1 by Francesco Casella, 5 years ago

Toni, have you tried with the latest nightly build? There's been a lot of improvements since the last official release.

comment:2 by Christian Kral <dr.christian.kral@…>, 5 years ago

I tested the examples

  • IMC_DCBraking.DCBraking_D2
  • IMC_DCBraking.DCBraking_D3

and they both fail in the actual nightly build version

Connected to OpenModelica 1.14.0~dev-26584-gae23951
Connected to OMSimulator unknown-linux

with the old and the new frontend and the simulation flag -d=newInst. In all cases the compilations finishes and the simulation fails.

comment:3 by Francesco Casella, 5 years ago

Priority: highblocker
Summary: Failed to solve linear system of equationsFailed to solve linear system of equations because of wrong dynamic state selection

I tried the IMC_DCBraking.DCBraking_D2 example. The state selection includes four statically selected states and one dynamic state, selected in a set of cardinality two. Unfortunately, OMC won't tell me who these states are, if I turn on -d=bltdump OMEdit crashes. Why this happens should be investigated, BTW.

I compared the solution with Dymola, which selects the same static states, and one dynamic state selected among imc.i_0_s, imc.idq_sr[1], and imc.idq_sr[2].

At runtime, Dymola selects imc.i_0_s as state, which keeps a value below 1e-14 for the whole simulation.

Conversely, OMC selects imc.idq_sr[2] - I checked that the numerical values of the dynamic state and of imc.idq_sr[2] are the same. imc.i_0_s grows very quickly to 1e14, then the simulation stops.

It looks like

  • there are some issues with the dynamic state selection (one less candidate w.r.t. Dymola)
  • there are some issues with the initial selection of the active dynamics state (the selected one leads to a singularity)

Probably related to #5459 and/or #4143. @Karim, while you try to solve them, would you mind monitoring this one as well?

comment:5 by Francesco Casella, 5 years ago

Milestone: 1.14.01.15.0

Releasing 1.14.0 which is stable and has many improvements w.r.t. 1.13.2.

This issue, previously marked as blocker for 1.14.0, is rescheduled to 1.15.0

comment:6 by Francesco Casella, 5 years ago

Milestone: 1.15.01.16.0

Release 1.15.0 was scrapped, because replaceable support eventually turned out to be more easily implemented in 1.16.0. Hence, all 1.15.0 tickets are rescheduled to 1.16.0

comment:7 by Francesco Casella, 4 years ago

Milestone: 1.16.01.17.0

Retargeted to 1.17.0 after 1.16.0 release

comment:8 by Francesco Casella, 4 years ago

Milestone: 1.17.01.18.0

Retargeted to 1.18.0 after 1.17.0-dev.beta2 release

comment:9 by Francesco Casella, 3 years ago

Milestone: 1.18.0

Ticket retargeted after milestone closed

comment:10 by Francesco Casella, 3 years ago

Milestone: 1.19.0

1.18.0 blocker tickets moved to 1.19.0

Note: See TracTickets for help on using tickets.