Changes between Version 1 and Version 2 of Ticket #6200, comment 13


Ignore:
Timestamp:
2021-09-14T18:02:42Z (4 years ago)
Author:
Francesco Casella

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #6200, comment 13

    v1 v2  
    33In 99% of the cases I am aware of (mostly thermo-fluid models), the redundant equations of the initialization problem have a very simple form, either {{{der(x) = 0}}} or {{{x = some_value}}}, where {{{x}}} are state or dummy state variables, and they simply show up more than one time in the final equation set.
    44
    5 Assuming the V-nodes are ordered lexicographically, we just iterate over them, and for each of the ones that correspond to states or derivatives, we check if there is more than one arc connecting to an equation node that has no other arcs. In that case, it is very likely that those equation nodes correspond to redundant initial equations, and we can symbolically check if they are consistent or not.
     5Assuming the V-nodes are ordered lexicographically, we just iterate over them, and for each of the ones that correspond to states or derivatives, we check if there is more than one arc connecting to an E-node that has no other arcs. In that case, it is very likely that those equation nodes correspond to redundant initial equations, and we can symbolically check if they are consistent or not, possibly deferring the check to runtime in case the result depends on parameters, as we already do.
    66
    77We could then keep all the other more elaborate algorithms to take care of other potentially occurring cases I'm now aware of.