Abstract:
The use of Neural Network coupled with Fuzzy Logic in control systems is on increase due to its
capabilities of adapting to the plant dynamics variations, dynamic environmental effects or working
in ill-defined environment. In contrast to classical methods, these techniques do not require a
mathematical model for designing the controllers. The accuracy of mathematical models can be
questioned in the case of dynamic, nonlinear and complex plants / processes. Neuro-Fuzzy systems
are based on the experts‟ knowledge and trained judgment of skilled workers, covering the whole
dynamics of the system through learning process and are better suited for dynamic environments.
This work is focused towards the use of adaptive neurofuzzy inference system in control systems
for the estimation and control of the error due to plant model mismatch and process uncertainties in
a Super-saturation Controlled Batch Crystallization process producing crystals with desired
properties (target size distribution). As a result of this research, a neuro-fuzzy controller is proposed
which is capable of compensating uncertainties related to the plant dynamics including stirring
speed, model inaccuracy and measurement errors for the control of temperature trajectories for
super-saturation controlled industrial batch crystallization process. We are using systemic direct
design approach (producing optimal temperature trajectories for super-saturation set point with
time) for crystallization of Potash Alum concentration in water, yielding desired target Crystal Size
Distribution at the end of a batch. The validation results are very encouraging and prove the efficacy
of fuzzy neural networks in the designing of dynamic control systems.
