Abstract: Based on the nonlinear fatigue damage model and damage accumulative model of materials, the damage elastic modulus of concrete and the damage residual area of steel reinforcement bars and prestressed tendons were introduced into the section stiffness calculation. The stiffness degradation model of the prestressed concrete beam under multilevel amplitude cycle loading was deduced, and genetic algorithm was applied in parameter determination. Then, the neural network method for forecasting the fatigue stiffness degradation was proposed and Matlab program code of the calculation model was generated. Experiment of prestressed concrete beam under multilevel amplitude cycle loading was carried out to establish the parameters, improve the calculation accuracy and verify the calculation results. It is suggested that the stiffness degradation of the prestressed concrete beam under multilevel amplitude cycle loading could be divided into three stages. The model calculation results agree well with experiment results, with a deviation within 5%. The genetic algorithm based model parameter fitting method and the neural network based forecasting method can give accurate results, thus providing useful reference for the estimation of fatigue resistance of prestressed concrete beam.

Key words: bridge engineering, prestressed concrete, multilevel amplitude cycle loading, stiffness degradation, fatigue deflection, damage stiffness forecasting