Deep Learning Approach for Industrial Process Improvement

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216305%3A26210%2F19%3APU134371" target="_blank" >RIV/00216305:26210/19:PU134371 - isvavai.cz</a>

Výsledek na webu

<a href="https://www.aidic.it/cet/19/76/082.pdf" target="_blank" >https://www.aidic.it/cet/19/76/082.pdf</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.3303/CET1976082" target="_blank" >10.3303/CET1976082</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Deep Learning Approach for Industrial Process Improvement

Popis výsledku v původním jazyce

The full operation of an industrial processing facility with artificial intelligence has been the holy grail of The full operation of an industrial processing facility with artificial intelligence has been the holy grail of Industry 4.0. One of the inherent difficulties is the enumerate and complex nature of processing information within an industrial plant. Hence, such data should be processed efficiently. This paper demonstrates the effectiveness of a deep auto-encoder neural network for the dimensionality reduction of industrial processing data. The deep auto-encoder neural network functions to intake all possible processing data from the processing system by sending it into an encoder neural network. Subsequently, the encoder condenses the data into highly compressed encoded variables. The network is trained in an unsupervised manner, where a decoder neural network simultaneously attempts to revert the encoded variables to their original form. Such a deep learning approach allows data to be highly compressed into lower dimensions. The coded variables retain critical information of the processing system, allowing reconstruction of the full process data. Auto-encoder neural networks are also able to provide noise removal for encoded data. For application, the encoded variable can be utilized as an effective dimension-reduced variable that can be used for plant-wide optimization. This paper also discusses the further applications of encoded variables for industrial process improvements using the Industrial Internet of Things (IIoT) technologies.

Název v anglickém jazyce

Deep Learning Approach for Industrial Process Improvement

Popis výsledku anglicky

The full operation of an industrial processing facility with artificial intelligence has been the holy grail of The full operation of an industrial processing facility with artificial intelligence has been the holy grail of Industry 4.0. One of the inherent difficulties is the enumerate and complex nature of processing information within an industrial plant. Hence, such data should be processed efficiently. This paper demonstrates the effectiveness of a deep auto-encoder neural network for the dimensionality reduction of industrial processing data. The deep auto-encoder neural network functions to intake all possible processing data from the processing system by sending it into an encoder neural network. Subsequently, the encoder condenses the data into highly compressed encoded variables. The network is trained in an unsupervised manner, where a decoder neural network simultaneously attempts to revert the encoded variables to their original form. Such a deep learning approach allows data to be highly compressed into lower dimensions. The coded variables retain critical information of the processing system, allowing reconstruction of the full process data. Auto-encoder neural networks are also able to provide noise removal for encoded data. For application, the encoded variable can be utilized as an effective dimension-reduced variable that can be used for plant-wide optimization. This paper also discusses the further applications of encoded variables for industrial process improvements using the Industrial Internet of Things (IIoT) technologies.

Druh

J_SC - Článek v periodiku v databázi SCOPUS

CEP obor

—

OECD FORD obor

20402 - Chemical process engineering

Projekt

<a href="/cs/project/EF16_026%2F0008413" target="_blank" >EF16_026/0008413: Strategické partnerství pro environmentální technologie a produkci energie</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2019

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Chemical Engineering Transactions

ISSN

2283-9216

e-ISSN

—

Svazek periodika

76

Číslo periodika v rámci svazku

1

Stát vydavatele periodika

IT - Italská republika

Počet stran výsledku

6

Strana od-do

487-492

Kód UT WoS článku

—

EID výsledku v databázi Scopus

2-s2.0-85076315617