In recent years computer science theories have been applied to manufacturing improving products quality, fault detection and process monitoring. However, there is a lack of research in the identification of causal relationships among data. These associations of cause-effect are important since they allow root causes to be analysed, they highlight the most influential process variables and they embed a typical human reasoning model that is largely applied in manufacturing. Compared to knowledge-based approach, data driven causal discovery (DCD) enables causal modeling without overloading expert operators and scales faster. However, DCD is challenging to be applied especially in small-medium enterprises where machines raw data are stored without the support of specialized data analyst team. In this work, we aim to automatically reconstruct the causal interaction model of the production flow from raw data. We use PCMCI, a constraint-based causal discovery algorithm, that handles both linear and nonlinear relationships in time series. We validate our method on a synthetic realization that emulates manufacturing features and on real data with domain expert support. The obtained results confirm that PCMCI is able to recognize more than 50% of causal relationships without any false positives. The application of the PCMCI method in an ultra-processed food manufacturer allows to propose a novel causal interaction model integrating data-driven and expert's knowledge.

Causal interaction modeling on ultra-processed food manufacturing

Menegozzo, G;Dall'Alba, D;Fiorini, P
2020

Abstract

In recent years computer science theories have been applied to manufacturing improving products quality, fault detection and process monitoring. However, there is a lack of research in the identification of causal relationships among data. These associations of cause-effect are important since they allow root causes to be analysed, they highlight the most influential process variables and they embed a typical human reasoning model that is largely applied in manufacturing. Compared to knowledge-based approach, data driven causal discovery (DCD) enables causal modeling without overloading expert operators and scales faster. However, DCD is challenging to be applied especially in small-medium enterprises where machines raw data are stored without the support of specialized data analyst team. In this work, we aim to automatically reconstruct the causal interaction model of the production flow from raw data. We use PCMCI, a constraint-based causal discovery algorithm, that handles both linear and nonlinear relationships in time series. We validate our method on a synthetic realization that emulates manufacturing features and on real data with domain expert support. The obtained results confirm that PCMCI is able to recognize more than 50% of causal relationships without any false positives. The application of the PCMCI method in an ultra-processed food manufacturer allows to propose a novel causal interaction model integrating data-driven and expert's knowledge.
978-1-7281-6904-0
Causal inference, Machine learning, process control
File in questo prodotto:
Non ci sono file associati a questo prodotto.

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/11562/1074691
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? 1
social impact