In this paper, an actuator fault-tolerant control (FTC) strategy based on set separation is presented. The proposed scheme employs a standard configuration consisting of a bank of observers which match the different fault situations that can occur in the plant. Each of these observers has an associated estimation error with a distinctive behaviour when a estimator matches the current fault situation of the plant. With this information from each observer, a fault diagnosis and isolation (FDI) module is able to reconfigure the control loop by selecting the appropriate stabilising controller from a bank of precomputed control laws, each of them related to one of the considered fault models. The control law consists of a reference feedforward term and a feedback gain multiplying the state estimate provided by the matching observer. A novel feature of the proposed scheme resides in the decision criteria of the FDI, which is based on the computation of sets towards which the output estimation errors related to each fault scenario and for each control configuration converge. Conditions for the design of the FDI module and for fault tolerant closed-loop stability are given, and the effectiveness of the approach is illustrated by means of a numerical example