@andrewlee94 I've started to look into this. What "issues with scaling methods" does addressing the potential division by zero bring up? I've made this change locally, but I can't seem to find any issues.

@MarcusHolly The transform_property_constraints, which is used in many of the old-style scaling methods, breaks as it assumes the constraint has the form property = function.

If we look only at the new Scaling tools this issue goes away, but there is the need to keep backward compatibility and thus the old-style methods would need to be updated.

And by breaking, do you mean that it's silently breaking? From what I can tell, all the tests and flowsheets appear to be running fine after reformulating the rule_mass_frac_phase_comp constraint.

Also, I'm not sure how well this utility function was really working in the first place. When I use report_scaling_factors on flowsheets with the GAC property model, the scaling factor set for eq_mass_frac_phase_comp is None before and after the reformulation. These flowsheets call iscale.calculate_scaling_factors, so it should be calling the transform_property_constraints which calls iscale.constraint_scaling_transform(c, sf). So there may actually some issue with the constraint_scaling_transform.

Oh, no, it was very vocal for me. However, I might have done the following for the reformulation:

def rule_mass_frac_phase_comp(b, p, j):
    return (
        b.mass_frac_phase_comp[p, j]
        * sum(b.flow_mass_phase_comp[p, j] for j in self.params.component_list)) == b.flow_mass_phase_comp[p, j]
    )

Try doing this - if this does trigger a lot of warnings then it means there are two issues that need to be addressed. The first being that the function requires the form property = function() and the second being that it assumes the LHS is the property of interest without confirming it.

The above formulation does not break any flowsheets or tests for me locally either. Perhaps I should create a fresh environment to see if I see the warnings then.