Hybrid Datafication of Maintenance Logs from AI-Assisted Human Tags

Published

January 15, 2018

Author(s)

Thurston B. Sexton, Michael P. Brundage, Michael L. Hoffman, Katherine C. Morris

Abstract

One of the main challenges of applying AI to some datasets derives from the datasets themselves being unstructured, unclear, and ambiguous. Furthermore, the insights that are to be gained reflect the quality of the data itself; if the data is skewed, so will be the insights. This problem is not unique to AI technology. People looking back at logs of past events often struggle to understand what was recorded, and to put together a timeline amongst a range of actors. AI technology can help humans sort the data out, but it does not provide the same insight often found in the background knowledge of human participants. This contextual weakness has made unstructured data hard to process. In our work, we have studied typical manufacturing maintenance logs to explore whether and how we can apply AI technologies to gain more insight from this—often vast and under-used—datasource. Our approach combines AI techniques for NLP, machine learning, and statistical processing with human understanding to quickly develop structured semantics reflecting unique datasets.

Proceedings Title

2017 IEEE Big Data

Conference Dates

December 11-14, 2017

Conference Location

Boston, MA

Conference Title

2nd Symposium on Data Analytics for Advanced Manufacturing

Pub Type

Conferences

Download Paper

https://doi.org/10.1109/BigData.2017.8258120

Keywords

maintenance, natural language processing, machine learning, prognostics, health management, smart manufacturing

Citation

Sexton, T. , Brundage, M. , Hoffman, M. and Morris, K. (2018), Hybrid Datafication of Maintenance Logs from AI-Assisted Human Tags, 2017 IEEE Big Data, Boston, MA, [online], https://doi.org/10.1109/BigData.2017.8258120 (Accessed April 19, 2024)