Skoči na glavni sadržaj

Izvorni znanstveni članak

Causal AI Modelling of Chemical Manufacturing Plants

Želimir Kurtanjek orcid id orcid.org/0000-0001-5453-6255 ; University of Zagreb, Faculty of Food Technology and Biotechnology, Pierottijeva 6, 10 000 Zagreb, Croatia, retired


Puni tekst: engleski pdf 568 Kb

str. 15-21

preuzimanja: 0

citiraj


Sažetak

The concept of “Industry 5.0” is driving significant changes in the production of chemical products and energy, promoting a shift towards a decarbonized and circular economy. Digitalization, robotics, communications, and artificial intelligence (AI) play crucial roles in fostering the development of necessary technological innovations and enhancing intelligent process control. The application of machine deep learning (ML) yields robust, field-neutral solutions for regression prediction objectives, but it is limited in its capacity to address innovative questions that involve causation and counterfactual analysis. This paper presents a proposed application of Bayesian networks (BN) for structural causal modeling (SCM) in the context of manufacturing plants. A critical feature of SCM modeling is its capacity to integrate extensive prior structural knowledge derived from fundamental chemical engineering principles with structures inferred from experimental data obtained from manufacturing plants. The acquired SCM facilitates the forecasting of causal relationships, the simulation of intervention strategies, and the generation of counterfactual responses essential for process innovations and intelligent process management. The SCM model is presented as a tool for examining causality and control in the intricate Tennessee-Eastman process.

Ključne riječi

Bayes network, causality, DAG, ATE, Markov blanket, Tennessee-Eastman-Process

Hrčak ID:

327428

URI

https://hrcak.srce.hr/327428

Datum izdavanja:

3.2.2025.

Posjeta: 0 *