In the aircraft Maintenance, Repair and Overhaul (MRO) industry the high uncertainty of resources needed for a given operation is a very recurrent problem. This implies that being able to collect and analyse data from past events to forecast future ones provides a company with more stability of outcomes and a competitive advantage in the industry. However, the heterogeneous manner in which data is collected and stored causes it to be often disregarded or at least not used in all its potential. This paper presents the development and evaluation of a BERT-based Natural Language Processing (NLP) model designed to classify aircraft zones from free-text maintenance reports. Using maintenance reports of 372 maintenance projects from a Portuguese MRO, the model achieved an accuracy of 85.69%. While developing the model a sensitivity analysis was performed to examine the impact of learning rate, warm-up period, and dropout probability on performance, allowing considerations for future model development in this area. In addition, the added value of this paper includes recommendations for maintenance management in companies intending to implement such model.
Lopes et al. (Thu,) studied this question.