A fatigue driving stress that causes fatigue damage is presented and used to predict residual fatigue life under variable loading. This fatigue driving stress is a function of the applied cyclic stress (), number of loading cycles (n), and the number of cycles to failure (N). It increases with loading cycles until the fatigue strength is reached when fracture occurs. By determining the equivalent number of cycles or life fraction at current load that yields the same fatigue driving stress as the previous loads, the remaining life can be predicted. A new damage model is derived as a function of expended life fraction of applied load, the logarithm life of applied load and the logarithm life of the initial applied load. The derived cumulative damage model indicates that P n i N i 4 1 for low-to-high loading sequence, and P n i N i 5 1, for high-to-low loading. In the case of constant amplitude loading, however, P n i N i ¼ 1. Life predictions from the present damage model and the Miner model are compared with experimental results from literature. The comparison shows that the present model presents a good estimation of the experimental data. Furthermore life prediction using the present model is found to give better agreement with experimental data than the popular Miner model.