Stratégies intelligentes de maintenance en production manufacturière dans le contexte de l’économie circulaire

Abstract

Circular economy has become an innovative approach for natural resource management nowadays. This approach is increasingly integrated in industrial settings as it enables meeting environmental requirements. In this context, intelligent maintenance can enhance the performance of production equipment. Therefore, the aim of this dissertation is to propose a predictive intelligent maintenance strategy to determine the condition of components within the next 24 hours. The predictive model obtained will provide probabilities of component failures. Several manufacturing contexts will be considered, where components are arranged in series, parallel, or series-parallel configurations. For each arrangement, we will seek to optimize production, inventory, and preventive maintenance plans while considering recycling during preventive maintenance. The benefits achieved for each configuration will be compared to determine the impact of predictive maintenance in a manufacturing company. We have therefore established two models, the first being a failure prediction model and the second an optimization model. Regarding the prediction model, it utilizes a pipeline consisting of preprocessed data and a classifier model using four algorithms: Logistic Regression (LR), Random Forest (RF), Adaptive Boosting (AdaBoost), and K Nearest Neighbors (KNN). An analysis of performance is conducted to determine the best model capable of classifying the failure of a component type. The results of this first model provide the failure probabilities for each component. These probabilities are then utilized in constructing the optimization model, which employs Mixed-Integer Quadratic Programming (MIQP). The construction of these models required several techniques and tools, including telemetry data collection, supervised machine learning, and the utilization of the LINGO 19.0 solver. In this dissertation, we first present the context of our study in Chapter 1, followed by a literature review in Chapter 2. The implementation drivers of predictive maintenance are addressed in Chapter 3. As for the predictive model, it is discussed in Chapter 4, followed by the definition of various industrial configurations, and finally, an optimization model is presented in Chapter 5, leveraging the results of the predictive model in association with the circular economy.

Date	5 Sept 2023
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Jean-Pierre Kenné (Supervisor)