This habilitation thesis presents my research and teaching activities at the Centre de Recherche en Automatique de Nancy (CRAN) and at the Faculty of Science and Technology of the University of Lorraine between 2005 and 2022. These activities are mainly related to the topic of fault-tolerant control of dynamical systems. They are structured under a bi-paradigmatic perspective declined in a data-based paradigm on the one hand and in a model-based paradigm on the other hand. In the data-based paradigm, after an introduction to the basic concepts of J.C. Willems' behavioral approach to systems, we present our contributions via two techniques for synthesizing fault-tolerant controllers based on the plant trajectories, one based on the so-called projection methods and the other based on direct online controller synthesis. We then discuss the important ancillary issue of control authority transfer during control reconfigurations, and propose two innovative solutions to this problem. In the model-based paradigm, the proposed solutions to fault-tolerant control are based on receding-horizon optimal control with implementation on application targets, wind turbines and non-residential buildings, for economic and energy efficiency purposes. We also address the problem of fault diagnosis for bilinear systems and present a new observer-based fault diagnosis algorithm. The last part of this thesis describes our research project in two main directions, one around data-driven intelligent control for energy efficiency in buildings and the other, more theoretical-oriented, around fault-tolerant optimal control based on time series.