In this paper, a multi-scale finite element driven framework for fatigue life prediction and parameter optimization of wind turbine blade composites is constructed. For the stress transfer between layers, interface damage evolution and stiffness degradation process, the finite element model is established and the life samples are formed. The dual-branch feature mapping and life prediction module is designed, the parameters, damage variables, strain energy density and load spectrum statistics are uniformly coded, and the adaptive particle search is combined to complete the parameter optimization. Experimental results show that compared with BP regression network, random forest regression and LSTM time series regression model, the absolute percentage error of the test set of the proposed method is 4.87%and the determination coefficient is 0.941. After optimization, the blade mass is reduced by 5.75%, the residual fatigue life is increased by 10.34%, and the constraint satisfaction rate is 96.8%. The results show that the proposed framework can provide computational support for wind turbine blade fatigue assessment and design.
