Obiekt

Tytuł: Design of a smart alarming system using transfer entropy-based causality graphs

Autor:

Pisiecki, Hubert ; Domański, Paweł

Data wydania:

2026

Typ zasobu:

artykuł

Współtwórca:

Korbicz, Józef (1951- ) - red. ; Uciński, Dariusz - red.

Tytuł publikacji grupowej:

AMCS, volume 36 (2026)

Abstract:

This paper addresses the critical challenge of alarm floods in industrial automation systems, where cascading alarms obscure root causes of disturbances. The primary aim is to develop and validate an original smart alarming algorithm that identifies disturbance source by analyzing directional information flow between the system`s process variables. The proposed methodology integrates transfer entropy (TE), an information-theoretic measure of causal relationships, with median ensemble empirical mode decomposition (MEEMD) for signal preprocessing. ; The scope encompasses data normalization, delay estimation via cross-correlation (CC) and mutual information (MI), causality computation using TE, and causal graph construction to visualize disturbance propagation pathways. The algorithm is validated on two systems: a Simulink multi-loop control model and the Tennessee Eastman Process (TEP) - an industrial benchmark. The proposed methodology successfully traces disturbances and validates the method`s ability to identify the root cause. Key innovations include noise-resilient decomposition and dual-method delay estimation, which enhance robustness in high-dimensional systems. The method reduces alarm floods by isolating primary disturbances, even under nonlinear and noisy conditions.

Wydawca:

Zielona Góra: Uniwersytet Zielonogórski

Identyfikator zasobu:

oai:zbc.uz.zgora.pl:96439

DOI:

10.61822/amcs-2026-0020

Strony:

297-313

Źródło:

AMCS, volume 36, number 2 (2026) ; kliknij tutaj, żeby przejść

Jezyk:

eng

Licencja CC BY 4.0:

kliknij tutaj, żeby przejść

Prawa do dysponowania publikacją:

Biblioteka Uniwersytetu Zielonogórskiego

Obiekty Podobne

×

Cytowanie

Styl cytowania:

Ta strona wykorzystuje pliki 'cookies'. Więcej informacji