La disponibilité élevée des systèmes technologiques comme l'aérospatial, la défense, la pétrochimie et l'automobile, est un but important des nouveaux développements de la technologie de conception des systèmes sachant que la défaillance onéreuse survient, en général, soudainement. Afin de rendre les stratégies classiques de maintenance plus efficaces et pour prendre en considération l'état et l'environnement évolutifs du produit, un nouveau modèle de pronostic analytique est développé en tant que complément des stratégies de maintenance existantes. Ce nouveau modèle est appliqué aux systèmes mécaniques soumis à la défaillance par fatigue sous charge cyclique répétitive. Sachant que l'effet de fatigue va initier des microfissures qui peuvent se propager soudainement et conduire à la défaillance. Ce modèle est basé sur des lois d'endommagement existantes dans la mécanique de la rupture comme la loi de propagation de fissures de Paris-Erdogan à côté de la loi de cumul de dommage de Palmgren-Miner. A partir d'un seuil prédéfini de dégradation DC, la durée de vie résiduelle (RUL) est estimée à l'aide de ce modèle de pronostic. Les dommages peuvent être cumulés linéairement (Loi de Palmgren-Miner) et aussi non linéairement afin de prendre en compte un comportement plus complexe des chargements et des matériaux. Le modèle de dégradation développé dans ce travail est basé sur une sommation d'une mesure de dommage D à la suite de chaque cycle de chargement. Quand cette mesure devient égale à un seuil prédéfini DC, le système est considéré dans l'état de panne. En plus, l'influence stochastique est incluse dans notre modèle pour le rendre plus précis et réaliste. / The high availability of technological systems like aerospace, defense, petro-chemistry and automobile, is an important goal of earlier recent developments in system design technology knowing that the expensive failure can generally occur suddenly. To make the classical strategies of maintenance more efficient and to take into account the evolving product state and environment, a new analytic prognostic model is developed as a complement of existent maintenance strategies. This new model is applied to mechanical systems that are subject to fatigue failure under repetitive cyclic loading. Knowing that, the fatigue effects will initiate micro-cracks that can propagate suddenly and lead to failure. This model is based on existing damage laws in fracture mechanics, such as the crack propagation law of Paris-Erdogan beside the damage accumulation law of Palmgren-Miner. From a predefined threshold of degradation DC, the Remaining Useful Lifetime (RUL) is estimated by this prognostic model. Damages can be assumed to be accumulated linearly (Palmgren-Miner's law) and also nonlinearly to take into consideration the more complex behavior of loading and materials. The degradation model developed in this work is based on the accumulation of a damage measurement D after each loading cycle. When this measure reaches the predefined threshold DC, the system is considered in wear out state. Furthermore, the stochastic influence is included to make the model more accurate and realistic.

There are three ways to deal with component failure: reactive maintenance, preventive maintenance, and predictive maintenance. Reactive maintenance is to repair only once something breaks. Preventive maintenance is to repair before it breaks, independent of actual wear. Predictive maintenance is performed on the basis of real time operational data, repairing when components cross a certain degradation threshold.  With classification models one can determine the health state of a component. Regression models, on the other hand, allow the user to calculate a more precise estimate of remaining useful life. Previous research on regression models have exclusively used sensory data while classification models have used both sensory data as well as log-data. Research on predictive maintenance using regression models have found most success using SVM regression, decision trees, random forest regression, artificial neural networks and LSTM models.  Companies have more and more data to their disposal about the performance of their machines, but usually in the form of log-data. The goal of this research is to find if it is possible to use log-data for regression models. If this is the case, more sophisticated regression models can be used to apply predictive maintenance more accurately on a broader scale than is currently the case. The project was performed through a case study at a company in the semiconductor industry in the Netherlands, with years of log-data of their product that are gradually degrading over time. After quantifying the log-data and trying all kinds of different regression models in combination with different time scales, the results were unilaterally abysmal and were unable to make any decent prediction.  The reason for this according to several experts in the field of data science is that there was no in depth understanding of the data. They say it is required to have an integral understanding of the log-data and to closely collaborate with field engineers who know the data in and out. If a field engineer can say something about the degradation of a machine using only the log-data, a machine learning model can do it too. If a machine learning model is unable to purposefully overfit on the training data and the results are bad, there is no signal in the dataset and the task is impossible. It does not matter if the data was originally sensory or log-based, the only thing that matters is understanding what the data means and the presence of the degradation signal within.