This paper treats the Fault Estimation (FE) problem using a data-projection method. An on-line fast fault estimation procedure is proposed using input-output data collected on-line and off-line. The model structure of the system (linear, bilinear,…) is supposed to be known but the model parameters values are not needed. No prior model parameter estimation procedure is required. The method was presented in previous work by the authors for sensor FE. This paper extends the procedure for actuator faults. Simulation results of a separately excited DC motor are presented to illustrate the efficiency of the proposed data-projection method for actuator fault detection and estimation.