Prediction Of Residual Fatigue Life Of GFRP Laminated Beams Based On Frequencies

Fiber Reinforced Polymer/Plastic(FRP),in contrast to traditional metallic materials,has many advantages such as high strength,high stiffness,high designability,light weight,corrosion resistance,etc.It is widely used in aerospace engineering,automotive engineering,civil engineering and other important engineering fields.Glass fiber reinforced composites(GFRP)have become one of the widely used FRP materials due to their low cost,such as being used to make large wind turbine blades.In engineering applications,the loading conditions of composite materials are very complicated,especially in the actual service process,FRP structures are usually subjected to repeated effects of alternating loads,and the fatigue problem is very prominent.With the gradual accumulation of fatigue damage in the structure,the stiffness and strength of the structure gradually decrease,which will seriously affect the service safety and service life of the FRP structure.At the same time,the reduction of stiffness during fatigue will cause the decay of the inherent frequency of the structure.Therefore,this paper will investigate the correlation between the remaining fatigue life and frequency of the GFRP beam structure by tracking the frequency change during fatigue,and then further use this correlation to try to quantitatively evaluate the remaining fatigue life of the GFRP structure by the value of frequency change.A combination of experimental studies,simulation and intelligent algorithms or tools is used to develop a method for predicting the remaining fatigue life of GFRP structures based on frequency variation and to validate it both numerically and experimentally.In particular,in the simulation,the element material properties of GFRP structures are used with normally distributed parameters to refine the simulation of the dispersion of material parameters commonly found in FRP composites due to initial defects,impurities,surface damage,etc.To summarize,the main research and conclusions of this paper are as follows.1)Tensile-tensile fatigue tests and modal tests were conducted on GFRP beams with lay-up angles of[45/-45/45/-45/-45/-45/-45/45]_S at different stress levels,and the logarithmic fatigue life was obtained to be basically linear with the fatigue stress level,and the dispersion of the fatigue life of the laminated beams became larger as the stress level decreased.The frequency of the fatigue process decreases with the increase of the number of cyclic loads,and its decay curve is consistent with the decreasing trend of the residual stiffness curve.The higher-order frequencies can be seen greater decrease than the lower-order frequencies.2)The fatigue analysis model of GFRP is constructed,and the model elements are assigned with homogeneous material parameters and random material parameters with normal distribution,and multiple failure modes of the fatigue process are implemented based on Abaqus and Umat subroutines.The simulation results show that the random distribution of material parameters affects the initial location of damage occurrence during fatigue simulation and the subsequent damage expansion path,thus affecting the final simulated fatigue life of the GFRP structure and making the model results show a dispersion similar to the test results.3)Based on the size and material properties of the GFRP beam specimens,fatigue life simulations were carried out for three stress levels,and the test and simulation results were in good agreement,with the relative error of log fatigue life within 3%.Further,the simulated frequencies were obtained by modal analysis of the fatigued GFRP beams at specific stress levels,and after comparing with the measured modal frequencies of the fatigue process in the experiments,it was found that the simulated frequencies of order 1-6 of the simulations were in good agreement with the experimental results.4)A prediction model for predicting the remaining fatigue life of GFRP laminated beams based on residual frequencies is constructed based on three machine learning algorithms:artificial neural network(BPNN),decision tree algorithm,and random forest algorithm.Due to the existence of multiple order frequencies,different combinations of frequencies were used to predict the fatigue life,and the accuracy of the three algorithms in predicting fatigue life was tested and compared using experimental cases and simulation cases.The numerical validation results show that all three algorithms perform similarly in terms of error and goodness of fit,indicating that all three algorithms can theoretically be used to predict fatigue life in the absence of errors.The experimental validation results showed that the combination of first order frequencies was used to predict fatigue life most inaccurately,which may be due to the large measurement error of the first order frequencies and therefore is best not used to predict fatigue life.Comparing the prediction results of the three algorithms in numerical and experimental validation,the artificial neural network overall has better prediction accuracy than the other two algorithms.