In this paper, a novel fatigue prognosis methodology integrating sensor data from usage monitoring system is proposed. The key concept is to reconstruct the dynamic responses for the critical spots where the direct sensor measurement is unavailable. The sensor data at limited locations measured from usage monitoring system are used as the basis of the reconstruction. Fatigue prognosis for the critical spot is performed using the reconstructed dynamic responses. The Empirical Mode Decomposition (EMD) method with certain intermittency criteria is employed to decompose the sensor measurements into a series of Intrinsic Mode Functions which represent the dynamic responses under mode coordinates. An extrapolation procedure based on the finite element analysis is used to reconstruct the dynamic responses for the critical spot after the decomposition. The extrapolated dynamic responses for the critical spot is then integrated with a physics-based fatigue crack growth model for fatigue damage prognosis. The proposed methodology is demonstrated using a multiple degree-of-freedom cantilever beam problem.