Automated Prioritization of Maintenance Work Orders: A Sentiment-Analysis Approach for MSC
Summary:
This feasibility study implemented Sentiment Artificial Intelligence (SAI) in an industrial maintenance scenario. This is one of the first implementations of SAI in an industrial maintenance scenario. It examined how sentiment analysis (NLP) can be used to objectively rank MSC maintenance work orders by urgency, replacing subjective human interpretation. The work quantified data quality limits, mapped domain-specific language challenges, and tested sentiment-based priority scoring as a pathway toward a more consistent, automated PdM decision pipeline.
Highlights:
- MSC’s SAMM work-order text contains >2,200 out-of-vocabulary terms; domain–specific lexicon creation is mandatory for accuracy.
- Repetitive failure descriptions produce low sentiment variation → key bottleneck for ranking granularity.
- Sentiment polarity + criticality + severity → proposed composite ranking logic (negativity = urgency).
- Sentiment models misinterpret technical abbreviations → maritime lexicon + SME reinforcement needed.
- Feasibility demonstrated; data quality + semantic structure limit production-readiness.
Duration: September 2023 – September 2024.
Associated Members:
