Uma metodologia baseada em grafo de conhecimento para análise de registros de alarmes e eventos industriais
Alarm and event logs make up a voluminous and dormant historical repository of tabular-like data, commonly undervalued or overlooked in manufacturing. Although they are a potentially rich source of relevant information about the monitored plant or process, these records are taken for analysis only a...
Na minha lista:
Autor principal: | |
---|---|
Outros Autores: | |
Formato: | doctoralThesis |
Idioma: | pt_BR |
Publicado em: |
Universidade Federal do Rio Grande do Norte
|
Assuntos: | |
Endereço do item: | https://repositorio.ufrn.br/handle/123456789/45676 |
Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
Resumo: | Alarm and event logs make up a voluminous and dormant historical repository of
tabular-like data, commonly undervalued or overlooked in manufacturing. Although they
are a potentially rich source of relevant information about the monitored plant or process,
these records are taken for analysis only as a last resort, mainly due to the difficulties
imposed by the large volume and low expressiveness of those databases. Such oversight
is no longer acceptable in the contemporary data-oriented scenario, already ubiquitous in
several productive sectors and gaining prominence in traditional manufacturing, especially due to the advent of the Industry 4.0 paradigm. Therefore, it is proposed to transpose
these bases to a more expressive and flexible representation domain, allowing a more
proactive exploration of the episodes reported in the records and, consequently, entailing
more agile incident, anomaly, compliance, and performance analysis tasks. For such,
from the recognition of an ontology, entities, attributes, and associations virtually immersed in the operational context described in the records are mapped into a Knowledge
Graph (KG). The approach uses Exploratory Data Analysis, Natural Language Processing, Network Analysis, Multivariate Analysis, and Composite Indicators techniques to
derive a myriad of aspects, properties, and relations from data, which are incorporated
as hierarchical, temporal, and similarity relationships (edges) between identified entities
(nodes). The visualization of the KG is dynamic and interactive, with different visualization modes and levels of detail. Evaluation scenarios are designed to demonstrate the effectiveness of the approach. |
---|