Abstract:Polypropylene (PP) is a synthetic resin with exceptional physical and chemical properties, making it suitable for various household and industrial applications. However, as the demand for PP continues to rise, the polypropylene industry faces challenges in improving its quality and efficiency. Developing an accurate mathematical model that describes the continuous polymerization process is crucial for enhancing PP production, as it is useful for controlling product quality and improving production efficiency. While previous studies have focused on modeling propylene polymerization using laboratory-scale and pilot-scale reactors, these models do not effectively apply to industrial-scale processes due to the complexity of the industrial environment. Therefore, a novel kinetic modeling method combining the modeling of the kinetic mechanism of the polymerization reaction with an intelligent optimization algorithm for the industrial-scale continuous gas-phase fluidized bed propylene polymerization process is proposed. To establish the kinetic mechanism model, the improved moments (MOM) method is utilized to derive a set of mass balance equations that account for various fundamental processes such as catalyst activation, chain initiation, chain growth, chain transfer to hydrogen, chain transfer to monomers, chain transfer to co-catalysts, and spontaneous deactivation. However, kinetic parameters serve as the link between the theoretical model and the actual production process, so the kinetic parameters estimation problem need to be transformed into a multi-objective problem by minimizing the least squares difference between the measurement data of the actual industry process and the predicted result of the kinetic model. Therefore, an improved multi-objective mayfly algorithm (IMOMA) is introduced to estimate the polymerization kinetic parameters, which incorporates the simulated binary crossover (SBX) operator and the polynomial mutation (PM) operator into the basic multi-objective mayfly algorithm (MOMA). Finally, by fitting industrial measurement data of the melt index (MI) and the polymerization rate (PR) of an actual polypropylene industrial case study, the proposed algorithm accurately determines the polymerization kinetic parameters. The results demonstrate that the MOM-IMOMA modeling approach outperforms other methods and holds significant promise for industrial-scale polypropylene production.

Control-Oriented Modeling for Industrial Propylene Polymerization Process Based on Physics-Informed Neural Network

Physics-informed Neural Networks to Solve Lumped Kinetic Model for Chromatography Process.

Developing Physics-Informed Neural Networks for Model Predictive Control of Periodic Counter-Current Chromatography

ARX-NNPLS Model Based Optimization Strategy and Its Application in Polymer Grade Transition Process*

Design and Development of Computer Integrated Control System for Spheripol Propylene Polymerization Unit

Kinetic Parameter Estimation of the Polyethylene Process by Bayesian Optimization

Recurrent Neural Network Integrated with Process-Priori-knowledge and Its Application

Simultaneous Optimization and Control for Polypropylene Grade Transition with Two-Layer Hierarchical Structure

Physics informed integral neural network for dynamic modelling of solvent-based post-combustion CO 2 capture process

Physics Informed Integral Neural Network for Dynamic Modelling of Solvent-Based Post-Combustion Co2 Capture Process

A Knowledge-driven Physics-Informed Neural Network model; Pyrolysis and Ablation of Polymers

Polymer chemistry informed neural networks (PCINNs) for data-driven modelling of polymerization processes

Stiff-PINN: Physics-Informed Neural Network for Stiff Chemical Kinetics

Physics-Informed Neural Networks for Dynamic Process Operations with Limited Physical Knowledge and Data

Physics-informed neural nets for control of dynamical systems

Physics-Enhanced Neural Ordinary Differential Equations: Application to Industrial Chemical Reaction Systems

Reduced-PINN: An Integration-Based Physics-Informed Neural Networks for Stiff ODEs

Novel kinetic modeling strategy for industrial-scale UNIPOL polypropylene with Ziegler-Natta catalyst

Nonlinear predictive control with modified closed-loop subspace identification-piecewise linear model for double-loop propylene polymerization process

Physics-Informed Neural Networks with Hard Linear Equality Constraints

Dynamics and Stability Analysis of Gas-Phase Bulk Polymerization of Propylene