From Data to Knowledge: A Case for Textual Equipment Reliability Data

Wang, C.; Mandelli, D.; Cogliati, J.

Permalink

https://hdl.handle.net/2142/121856 Copy

Description

Title

From Data to Knowledge: A Case for Textual Equipment Reliability Data

Author(s)

• Wang, C.
• Mandelli, D.
• Cogliati, J.

Issue Date

2023

Keyword(s)

• Knowledge extraction
• Natural Language Processing (NLP)
• Reliability

Abstract

Complex engineering systems such as nuclear power plants (NPPs) are generating and collecting large amount of data equipment reliability (ER) elements that contain information about the status of component, assets, and systems. Some of this information is in textual form where typically, events such as incident reports and maintenance activities are described. The analyses of textual data in current NPPs using natural language processing (NLP) methods have grown in the last decade and only recently the potentials of this kind of analyses have emerged. So far, applications of NLP methods have been limited to mostly classification and prediction with the goal of identifying the nature of the textual element (e.g., safety or non-safety relevant). Here we are targeting a more complex problem: the automatic generation of knowledge out of a textual element in order to assist system engineers in assessing system health. The concept of “knowledge extraction” is very broad, and its definition might vary depending on the application context. Our methods are a blend of rule-based and machine learning (ML) algorithms. In our context, knowledge extraction means that, out of a textual element, we want to identify what are the systems or assets mentioned in it, the type of event that is described (e.g., a component failure or a maintenance activity). In addition, we wat to capture details such as measure quantities, temporal or cause-effect relations between events. In this paper we will present how textual data elements are pre-processed in order to handle typos, acronyms, and abbreviations. Then we will show how ML and rule-based algorithms are employed to decompose and extract knowledge out of either short or long sentences. Few applications will be presented as well.

Type of Resource

text

Language

eng

Handle URL

https://hdl.handle.net/2142/121856

Owning Collections