I am modelling a new refrigerant molecule using the contribution method and integer cut with MINLP/RMINLP with Baron/CONOPT solver. The program terminated normally and show that the solution is infeasible but given no reason why this happen. I have tried several ways such as giving an initial guess and setting the lower boundary etc but none of them works.

When you solve the RMINLP with BARON you get with the option you already supply compIIS the following information in the listing file:

A problem may contain several independent IISs. Only one IIS will be
found per run. Alternative IISs may be obtained by using different
values of the CompIIS option in BARON.
Number of equations in the IIS: 2.
Lower: Treduce(1) >= 272
Upper: Treduce(3) <= 0
Number of variables in the IIS: 2.
Lower: Tr(1) >= 0.0001
Lower: Tr(3) >= 0.0001

Since T(3)=0 and TR is strictly positive (.lo=1e-4), the only way to make Treduce feasible for m=3 is to set Tc=0. That collides with T(1) being=272 which can’t be done with Tc=0.

Thank you for the answer. I was still quite confused why T(3) is 0 in this case, as T(2) and T(3) are calculated in similar ways, if T(2) has a value, how can T(3) = 0? is this caused by formulation or another reason?

m 'Calculation of vapour pressure, 1 @ eaporation temperature, 2 @ condensing temperature' /1*3/
T(m) This parameter set is assignment for the convience of vapour pressure calculation
/1 272
2 316/

without the m=3. GAMS is a sparse system and assumes 0 where no data is given. Hence T(3)=0.

The fact is I have removed the pressure calculation part so T(‘3’） should not be a problem at this point. I have also get rid of the unnecessary data point. contribution_table.xlsx (17.8 KB)
However, I still got the infeasibility message with extremely large lower bond as the figure indicate in the attachment. Molecular_design_2.gms (6.77 KB)

here is the infeasibility information from compIIS. I would start by checking equation Bonding2(‘M4’).
You can get this information by using the following code