The problem of on-line optimization of a model IV fluidized catalytic cracking (FCC) unit is analyzed. The "process" was modeled by combining a model IV FCC unit dynamic simulator (McFarlane et at, 1993; Khandalekar, 1993) with a ten-lump yield model (Jacob et at, 1976; Arbel et at, 1995) for the reactor section. The steady-state optimization model consisted of macroscopic steady-state models of the regenerator and reactor, a simplified reactor yield model, and models of the various process constraints. The steady-state optimization problem was solved using successive quadratic programming software (NPSOL, Gill et at, 1986) and the optimum process set points were implemented on the process simulator using a nonlinear constraint controller (Kandalekar and Riggs, 1995). The relative performance of constraint control, off-line optimization, and on-line optimization is compared for different feed characteristics and product pricing structures.