A new fatigue life prediction framework provides an improved life prediction under statistically and spectrally similar irregular variable-amplitude loading for a notched beam model. It enables the cumulative damage rule to account for the load sequence effects by modifying the probability density function of the stress-amplitude history through (1) identification of overloads based on the rainflow-counting algorithm; (2) analytical characterization of the overload retardation effects; and (3) correction to the damage rule using overload amplitude rate characterization. The fatigue lives estimated from experimentally acquired and synthetically generated load-time histories are compared to the ones generated from simulations that qualitatively reproduce the fatigue lives in physical experiments. The notable improvement in prediction accuracy outperforms the Palmgren-Miner's rule and power-spectrum-based life estimation. The demonstrated application to the field acceleration data substantiates its use for in-service structural health monitoring and damage prognosis. This framework does not require a priory knowledge of the applied load, and it can be applied to other engineered structures with known structural and defect properties.