Fault Log Text Classification Using Natural Language Processing And Machine Learning For Decision Support

Abstract

In recent years, various industries have been on the quest to derive new knowledge and information from the data they produce. When these data are well utilised, they can create frameworks for improving business processes, product quality, and services. However, more often, data are in unstructured and semi-structured data formats. Because of this, the discovery of critical issues within textual data becomes challenging. In the past few years, the adoption of natural language prepossessing (NLP) and machine learning (ML) techniques are increasingly becoming popular for exploring knowledge within text documents that could help decision-makers and experts to solve business challenges and improve their business processes and systems. This research is being experimented with NLP and ML on the fault log of a UK-based commercial MRO (Maintenance, Repair, and Overhaul) provider in the Aerospace Industry to support decision-making. The first stage systematically leverages text analysis to extract valuable information from many customers' fault notifications, compares its similarity with the expert's maintenance action, and then classifies them into three categories which are Modification, Replacement, and No-fault-found. In the second phase, the extracted features get fed into the machine learner to categorise and predict future faults diagnosis in commercial aircraft’ FQIS (Fuel Quantity Indicating System) to automate troubleshooting, support maintenance operations, and improve decision-making in MRO services.