Parameterized data-driven fuzzy model based optimal control of a semi-batch reactor

Area of Science:

• Chemical Engineering
• Control Systems Engineering
• Computational Intelligence

Background:

• Semi-batch processes exhibit complex nonlinear and time-varying dynamics.
• Accurate modeling is crucial for effective process control and optimization.
• Existing data-driven models may struggle with the inherent complexities of these systems.

Purpose of the Study:

• To propose a novel Parameterized Data-Driven Fuzzy (PDDF) model structure for semi-batch processes.
• To apply the PDDF model for optimal control strategy development.
• To evaluate the performance of the PDDF model against established methods.

Main Methods:

• Developed a PDDF model using orthonormally parameterized inputs, initial states, and process parameters.
• Fuzzy rules were derived from a linear data-driven model, with defuzzification using linear regression.
• Applied the fuzzy model to formulate optimal control problems for single and multi-rate systems.

Main Results:

• The PDDF model accurately captured the nonlinear and time-varying behavior of a multivariable semi-batch reactor.
• Optimal control results using the PDDF model were comparable to those from an exact first principles model.
• Performance was also found to be comparable to or superior to optimization results from a data-driven artificial neural network model.

Conclusions:

• The proposed PDDF modeling approach provides an effective tool for semi-batch process analysis and control.
• PDDF modeling offers a competitive alternative to first principles and artificial neural network models for complex process optimization.
• This approach demonstrates significant potential for improving the efficiency and performance of semi-batch operations.