This study describes a general framework for coupling and optimizing multip
le models with multiplier-free reduced Hessian successive quadratic program
ming. This tailored approach enables the use of existing process simulators
/models simultaneously with the optimizer, and this leads to efficient solu
tion of process optimization problems. In this paper, a unified strategy is
proposed to combine the model information and then decompose the sensitivi
ty matrices that arise from the connections of the streams. With the propos
ed decomposition approach, the need for storing a large constraint matrix i
s avoided and individual models that only pass their Newton corrections and
sensitivity information are left. The solution avoids problems due to mode
l failure and can save a lot of time because models are not converged at in
termediate optimization iterations. The resulting tailored optimization app
roach is illustrated on the dynamic optimization of a batch reactor/column
system selected from Yi and Luyben (Comput. Chem. Eng. 1997, 12, 25).