Architectures for neural networks as surrogates for dynamic systems in chemical engineering

[Display omitted] •Surrogate models for dynamic systems in chemical engineering.•Evaluation of recursive neural network designs.•First ever surrogate model for a batch distillation from start-up to shut down.•Hyperparameter tuning by Bayesian and Bandit optimization. Surrogate models for dynamic sys...

Full description

Saved in:
Bibliographic Details
Published in:Chemical engineering research & design 2022-01, Vol.177, p.184-199
Main Authors: Esche, Erik, Weigert, Joris, Brand Rihm, Gerardo, Göbel, Jan, Repke, Jens-Uwe
Format: Article
Language:English
Subjects:
Bayesian analysis
Chemical engineering
Chemical industry
Computer architecture
Distillation
Dynamic systems
Dynamical systems
LSTM
Neural networks
Optimization
Recurrent neural networks
Shutdowns
Surrogate models
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:[Display omitted] •Surrogate models for dynamic systems in chemical engineering.•Evaluation of recursive neural network designs.•First ever surrogate model for a batch distillation from start-up to shut down.•Hyperparameter tuning by Bayesian and Bandit optimization. Surrogate models for dynamic systems in chemical engineering are increasingly of interest. Neural networks have already been applied in research, but it remains unclear which types of neural network architectures are actually required for practical systems. The focus here lies on recurrent neural networks of type Jordan, Elman, and LSTM layers. These are investigated for different types of data sets as training basis: batch trajectories, data of a proper excitation of a continuous process, and a typical operation trajectory of a large chemical plant. To ensure a rigorous investigation, hyperparameter tuning by Bayesian and Bandit optimization is included. As a first, a dynamic surrogate model using LSTM layers for a batch distillation system is presented, which is valid from start-up until shutdown. The evaluation shows further need for adjustments in data preparation and objective/loss function compared to the state of the art.
ISSN:0263-8762
1744-3563
DOI:10.1016/j.cherd.2021.10.042