2026 February Update

Major Recent stuff:

What I think is next is two big questions:

  • How do we make solving more reliable?
  • How do we actually link up a model to live data?

How do we make solving more reliable

With building the Evaporator 3 Effect flowsheet, the model would solve, but was so unstable that even a change in flow rate of 1 kg/s would cause it to fail. However, it was not actually infeasible - a series of updates changing the flow by 0.1 kg/s each time was able to slowly lower the flow until it did solve at that point. This shows that the idaes Homotopy Meta-Solver would be able to solve the model. So that's one potential solution (though it would require refactoring so that instead of updating constraints for expressions, we add variables for expressions and fix them.)

Better model formulation

Better Model Scaling

Slack variables for model diagnostics

phasing between model complexity levels

ideas.blending-property-packages (Private)


Backlinks