Hybrid deep neural network–model predictive control strategy for high-yield magnesium stearate: A sustainable batch reactor temperature control framework

Document Type

Article

Publication Date

8-1-2026

Abstract

A hybrid Deep Neural Network–Model Predictive Control (DNN-MPC) controller is developed to regulate the temperature of a batch reactor for high yield magnesium stearate production using palm stearin as feedstock. In this work, a deep neural network (DNN) soft sensor is employed as the artificial intelligence (AI) component within the model predictive control (MPC) framework to enhance control performance. The DNN architecture consists of three hidden layers with 32 neurons each is chosen as the soft sensor for hybrid controller as it demonstrates the best performance by achieving lowest error and highest accuracy. The trained DNN is integrated into a stable MPC framework to form the proposed hybrid DNN-MPC controller. The proposed DNN-MPC controller significantly outperforms conventional single, hybrid and AI-based controllers, achieving rapid settling time, minimal overshoot, and precise set-point tracking. It demonstrates good disturbance rejection and robust performance under real-world conditions with minimal error. The highest yield and selectivity of magnesium stearate, 68.43% and 56.87% respectively, were achieved with the proposed DNN-MPC controller. This work improves batch reactor temperature control and product yield by implementing a DNN-MPC framework, addressing a gap in the current literature and offering a more practical and effective control strategy.

Keywords

Artificial intelligence, Batch reactor, Deep neural network-model predictive control, Soft sensor, Temperature control

Publication Title

Engineering Applications of Artificial Intelligence

ISSN

0952-1976

DOI

10.1016/j.engappai.2026.114989

Volume

177

Share

COinS