Nonlinear Model-Based Inferential Control of Moisture Content of Spray Dried Coconut Milk

Area of Science:

• Chemical Engineering
• Process Control
• Food Science

Background:

• Powder moisture content is critical for product stability and shelf life.
• Online measurement of moisture content is challenging, necessitating inferential control.
• Spray drying is a common method for producing powders like coconut milk powder.

Purpose of the Study:

• To design a nonlinear model-based inferential controller for coconut milk powder moisture content.
• To integrate fundamental (Reaction Engineering Approach - REA) and empirical (NARX, Neural Network) models for spray drying dynamics.
• To achieve accurate and stable control of powder moisture content.

Main Methods:

• Developed a one-dimensional spray dryer model incorporating the REA model.
• Utilized Nonlinear Autoregressive with Exogenous Input (NARX) and Neural Network models for predictive control.
• Implemented nonlinear model-based inferential control strategies.

Main Results:

• The neural network estimator demonstrated high accuracy with minimal offset (<0.0003 kg/kg).
• The inferential control system provided stable responses across a wide range of operating conditions.
• The system exhibited acceptable disturbance rejection capabilities.

Conclusions:

• Nonlinear model-based inferential control is effective for managing coconut milk powder moisture content.
• Integration of REA, NARX, and neural networks enables robust process control.
• Controller performance is influenced by the applied tuning rules.