Compétences nécessaires pour intégrer l’utilisation de l’apprentissage automatique en techniques de maintenance industrielle

Abstract

Data science and artificial intelligence is becoming increasingly important in Quebec manufacturing industry. In order to address this business requirement, it is necessary to adopt a culture of data science. At the moment, the industrial maintenance program does not include core competencies in the field of data science or digital transformation. Within technologies that can have a significant impact on the role of the industrial maintenance technician, we can note internet of things, big data, and artificial intelligence since they are at the heart of connected machines. In this study, a centrifugal pump is used to build a proof of concept to test an industrial data analysis approach. The main objective of this study is to introduce machine learning core competencies in the industrial maintenance program and apply anomaly detection algorithm to a maintenance problem. This study aims to : (1) list which machine learning tools can be used as an introduction to anomaly detection, (2) map the necessary core competencies needed to use theses tools, (3) develop an approach to introduce those new competencies in the industrial maintenance program, (4) develop a diagnostic chart to adapt the training program and add the required competencies to the curriculum. This study is based on the development of a complete pipeline applied to a mechanical system. The objective is to collect data from the pump to build a dataset, feed an anomaly detection algorithm with the data and run inferences on the system to determine its condition by using the data. This study proposes a simple and affordable methodology to test a proof of concept and experiment with anomaly detection for an industrial maintenance problem. The result of his work is intended to give the technicians the ability to use the concepts behind machine learning and use associated tools in their future careers. These new competencies could modernize the current industrial maintenance program and answer the business needs for integration of digital technologies and partially relief the need for maintenance technicians. In this study, a dataset that represent working conditions of a pump is collected to perform anomaly detection. The proposed methodology requires to master new machine learning core competencies for the technician to be able to cooperate with data science experts. Recommendations are grouped in a table indicating the core competencies to add to the industrial maintenance program.

Date	31 Mar 2026
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Michel Rioux (Supervisor) & Tony Wong (Co-supervisor)