Fault-tolerant control is an essential component in modern process industries as abnormal situations account for over $20 billion in lost annual revenue in the US alone. Traditionally, control systems rely on centralized control architectures utilizing dedicated wired links to measurement sensors and control actuators to operate a plant at desired conditions and a separate monitoring system for detecting faults. While this paradigm to process operations and control has been successful, modern chemical plants that rely on highly automated processes to maintain robust operations and efficient production are vulnerable to abnormal situations like, for example, actuator faults. Loss of control in a chemical process can lead to the waste of raw materials and energy resources, as well as downtime and production losses but most importantly it may lead to personnel injury or death and/or environmental hazard. This issue has prompted significant research efforts in the integration and application of fault-tolerant control to existing legacy control systems. This dissertation will present a paradigm shift to the existing approach of designing control systems and monitoring systems in that it proposes to design distributed control systems that are stabilizing, robust and optimal, and whose design leads to closed-loop system structures that facilitate fault isolation with the flexibility to not only avert disaster in the case of an abnormal situation but maintain optimal plant operation. To present our method of fault-tolerant control, we will focus on a broad class of non-linear process systems subject to disturbances and persistent control actuator faults. In general terms, the method includes the design of distributed model predictive control laws combined with a fault-detection and isolation approach based on process models and fault-free data that leads to successful detection and isolation of an actuator fault. After isolation of an actuator fault, the fault-tolerant control system estimates the fault magnitude, calculates a new optimal operating point, and ultimately reconfigures the distributed model predictive control system to maintain stability of the process in an optimal manner. Throughout the thesis, detailed examples of large-scale chemical process systems are used to demonstrate the approach.