Chemical processes are nonlinear. Processes with extremely high nonlinearities, such as neutralization and high-purity distillation, are very important and need special considerations. The basic problem with such nonlinear processes is that the performance of model-based control is very sensitive to model inaccuracy. It seems that robust control is impossible with pure model based control algorithms. Model predictive control (MPC) has been widely implemented in the chemical industry. However, not very many successful cases of implementing nonlinear models can be found in the literatures. In addition, when such a model is inaccurate, high-frequency oscillation appears across the sensitive region. On the other hand, an accurate model is expensive and frequently impossible since operating data in the sensitive region are scarce. The above factors lead to unacceptable control results. To solve the above problems, we propose a combined global/local control (GLC) in which, when disturbances occur, the global control (GC, MPC in this study), a nonlinear controller, steers the process under control into or near the sensitive region; then, the local control (PI in this study) takes over and finally settles the process at the desired set point. Both simulation and experimental results show that such a combination control is economical. In this study, unlike our previous research, a PI controller was implemented, because PI control can be easily tuned for the sensitive region and a model of moderate accuracy for other non-sensitive regions can be built with much less effort.