The BASF Antwerp site is one of the most advanced and integrated chemical production sites of the BASF group and worldwide. The chemical plants on site are heavily interconnected, i.e., the product of one plant is the raw material used in the next one and some of the plants use the steam produced from neighboring ones. This high degree of interconnection makes it quite difficult to assess the feasibility let alone the economic optimality of production and maintenance plans for the entire site. This research explores the use of a data-driven digital twin to simulate and assess these plans. Two value chains, i.e., a collection of plants converting raw material to valuable end products have been selected as a proof-of-concept for the entire site. Each plant has been modeled by utilizing simple or multiple regression. Each regression model correlates the final product of a plant with the needed raw materials or utilities (e.g., steam or electricity). All regression models have been found using the software JMP. Additionally, all relevant tanks used to stock raw materials, intermediates and final products have been modelled. This allowed for the visualization and troubleshooting of a particular component excess or shortage. The resulting system of 76 variables has been solved in MATLAB in a multi-period fashion, where a period represents a day. The simulation has been first performed on the training data set, and then on a validation period to verify the models' performance.

Data-driven digital twin of a chemical production site for production and utilities planning

Manenti F.;
2021-01-01

Abstract

The BASF Antwerp site is one of the most advanced and integrated chemical production sites of the BASF group and worldwide. The chemical plants on site are heavily interconnected, i.e., the product of one plant is the raw material used in the next one and some of the plants use the steam produced from neighboring ones. This high degree of interconnection makes it quite difficult to assess the feasibility let alone the economic optimality of production and maintenance plans for the entire site. This research explores the use of a data-driven digital twin to simulate and assess these plans. Two value chains, i.e., a collection of plants converting raw material to valuable end products have been selected as a proof-of-concept for the entire site. Each plant has been modeled by utilizing simple or multiple regression. Each regression model correlates the final product of a plant with the needed raw materials or utilities (e.g., steam or electricity). All regression models have been found using the software JMP. Additionally, all relevant tanks used to stock raw materials, intermediates and final products have been modelled. This allowed for the visualization and troubleshooting of a particular component excess or shortage. The resulting system of 76 variables has been solved in MATLAB in a multi-period fashion, where a period represents a day. The simulation has been first performed on the training data set, and then on a validation period to verify the models' performance.
2021
File in questo prodotto:
File Dimensione Formato  
153.pdf

accesso aperto

: Publisher’s version
Dimensione 1.04 MB
Formato Adobe PDF
1.04 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/1196613
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact