Adaptive Neuro-Fuzzy Control of pH and its Derivative in Neutralization Processes: A Comparative Study

Abstract

Models for pH and its derivative with respect to base flow rate, dpH/dF_b, were derived for the neutralization of a strong base-weak acid systems. The systems are then controlled and compared through simulation using PI Control, Fuzzy Logic Control (FLC) and Adaptive Neuro-Fuzzy Control (ANFC). Control of dpH/dF_b is an effort made to utilize other phenomenological information of the pH process in a way to improve its control. The dynamics of pH and dpH/dF_b were first compared for various acid/base concentrations and flow rates ratios. The comparison reveals that the dpH/dF_b  curve is highly non-linear and sensitive than the pH neutralization one.     The dpH/dF_b curve is characterized by the singular equivalence point in a two sensitivity  regions which makes it a good candidate as bench mark test to check the effectiveness of non-linear regulatory control strategies of pH. The control of pH and of dpH/dF_b was then compared for the system with and without time delay. The comparison showed that FLC is superior to PI controller for the system with time delay for both pH and dpH/dF_b control for a variety of load changes. Because of the multi-sensitivity-region characteristics of the strong base-weak acid system, adaptation of the FLC could enhance the controllability of both pH and dpH/dF_b. An adaptive method for fuzzy logic control of pH and dpH/dF_b was designed, incorporating neural network in identifying and positioning of the membership functions in the FLC and for determining membership functions' position in the fuzzy logic adapter (FLA) of the scaling factors. The results satisfactorily show that the flexible learning ability of neural network improves adaptation and control performance for both pH and dpH/dF_b.