The nonlinear model predictive controller (NAMPC) reported earlier has been extensively tested on a pilot scale distillation column and its performance compared with other controllers such as generic model control (GMC) and proportional-integral (PI). The algorithm has also been extended for dual composition control of the distillation column. With the use of a transport phenomena based analytical model at the supervisory level, the disturbances in the secondary variables generate an automatic feedforward signal which significantly reduces their destabilizing effects. A comparison of NAMPC with GMC for a single input-single output system showed that while the former is somewhat slow because of gradual change in manipulated variable, the latter takes more drastic control action which may result in undershoots and overshoots. Dual composition control of distillation column using NAMPC was stable and smooth for both regulatory and servo problems for which PI was found to be oscillatory and unacceptably poor.