CNN-LSTM-Based Nonlinear Model Predictive Controller for Temperature Trajectory Tracking in a Batch Reactor.

Area of Science:

• Chemical Engineering
• Process Control
• Artificial Intelligence in Industry

Background:

• Batch reactors (BRs) are versatile for specialty chemicals and food processing, handling complex reactions and varying conditions.
• Effective temperature control is crucial for optimizing polymerization reactions and ensuring safety in BRs.
• Traditional control methods may struggle with the complex dynamics of exothermic reactions in batch processes.

Purpose of the Study:

• To develop and evaluate a CNN-LSTM-based nonlinear model predictive controller (NMPC) for precise temperature profile tracking in batch reactors.
• To optimize coolant flow rate for managing exothermic reactions and enhancing control performance.
• To improve computational efficiency using a heuristic method with sigmoidal weighting functions.

Main Methods:

• Utilized a hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network for predictive modeling.
• Trained the CNN-LSTM model using open-loop experimental data from the batch reactor.
• Implemented a nonlinear model predictive controller (NMPC) integrating the CNN-LSTM model and a heuristic optimization approach.

Main Results:

• The CNN-LSTM-based NMPC demonstrated accurate temperature profile tracking capabilities.
• The controller effectively optimized coolant flow rate to manage exothermic reaction heat.
• The heuristic method improved the computational efficiency of the NMPC model.

Conclusions:

• The developed CNN-LSTM-based NMPC offers a robust and accurate solution for batch reactor temperature control.
• This approach shows significant potential for large-scale industrial applications, particularly in the pharmaceutical sector.
• Implementation can lead to enhanced process efficiency, reduced energy consumption, and improved operational safety.