Machine Learning Applied to the Classification of Technical Inspection Recommendations Regarding the Trend to Increase Criticality.

Context: The problem of prioritizing maintenance activities has become a subject of great interest to the industry in a highly competitive scenario where profitability, productivity, and safety are sought in highly automated plant operations. Problem: The operational unit object of this study faces a high demand for maintenance services, indicated by the technical inspection recommendations (TIR) and the risks arising from not meeting them in a timely manner, with the consequent operational and financial losses. Solution: This article proposes a tool to support the prioritization of equipment maintenance activities through a machine learning model that classifies TIRs according to their criticality, thus providing a second opinion for the classification made by the inspector and, therefore, indicating those that have a greater tendency to escalate in criticality. IS Theory: This work is associated with the Theory of the knowledge-based company, assisting in decision-making and efficiency in the application of resources. Method: To construct the Machine Learning model, algorithms and ensembles used in classification problems and natural language processing methods were used, considering that one of the most relevant attributes of the dataset is a free textual field. Summary of Results: After carrying out several experiments, we arrived at the best performance model, with a recall of 87.08% and an accuracy of 95.19%, values understood as promising for the implementation of the tool in production. Contributions and Impact in the IS area: The main contribution was to identify the feasibility of using machine learning models for the classification of technical inspection recommendations regarding the trend to increase criticality.