Fatigue failure of aero-engine turbine blades often occurs under complex random loads during service. To solve this problem, an improved Manson-Coffin model considering average stress effects was established through theoretical analysis, numerical simulation and local stress-strain method. The nonlinear influence of average stress on fatigue life s considered in this model. Fatigue test data of GH4133 superalloy, Al7075-T651 aluminum alloy and carbon steel were selected for model validation. The prediction results were compared with those obtained from traditional Manson-Coffin model and SWT model. The results show that 85% of the relative errors between predicted and experimental fatigue lives fall within ±40% range when using the improved model. Both the mean value and standard deviation of error factors are smaller than other models, indicating better prediction concentration. Compared with SWT model and traditional Manson-Coffin model, the relative errors are reduced by 78% and 99% respectively. The improved Manson-Coffin model demonstrates higher accuracy and reliability than the other two models. It is concluded that the improved Manson-Coffin model demonstrates higher accurate and reliable than the other two models.