| Fiber Reinforced Polymer(FRP)is widely used in aerospace,rail transit,civil engineering,energy technology,and other fields due to its excellent mechanical properties.With the full application of composite material structures in various fields,fatigue failure due to the long-term use of composite materials has become increasingly prominent.Therefore,domestic and foreign scholars have conducted in-depth research on the assessment of cumulative damage and fatigue life of composite materials in service.They are committed to developing an efficient,convenient,and reliable prediction method to predict and evaluate the fatigue life of composite materials.Vibration frequency is an excellent non-destructive testing method,sensitive to the small damage generated during composite materials' service.It can predict and locate the size of damage through the change of frequency.The acquisition of frequency is convenient and reliable.It will not affect the mechanical properties or service of the tested structures.The frequency measurement is convenient and reliable and does not affect the structure's overall mechanical properties.It can also be used for online and local monitoring of the structure under test during service.Therefore,in this thesis,we will use frequency as a parameter to investigate the relationship between frequency change and fatigue life and predict the fatigue life of composite structures based on frequency change.Based on the five typical residual stiffness models in the existing literature,this thesis derives five corresponding residual frequency theoretical models and constructs two different artificial intelligence(AI)algorithm models,including support vector machine(SVM)model and artificial neural network(ANN)model.The above five theoretical models and two algorithm models are used to predict composite materials' fatigue life,and the prediction accuracy of the two types of models is compared.In this thesis,specimens of carbon fiber reinforced composite material CFRP and glass fiber reinforced fiber composite material GFRP were produced for the experiment,and the prediction model proposed in this thesis was experimentally verified using experimental data.The results showed that:(1)Among the five theoretical prediction models,the model derived from the Yang model has the best prediction accuracy.Comparing the theoretical models constructed with different mode frequencies,the theoretical model constructed based on the eighth order frequency has the best prediction accuracy.(2)In comparing two different algorithm prediction models,12 different cases weredesigned according to the property of frequency multiple dimensions.The 12 cases are including the single-mode frequency case and the multi-mode frequency combination case.Between SVM and ANN,the SVM algorithm model's prediction accuracy is the best,the minimum log errors are1.34% and 1.84%,respectively.And in different prediction cases,the multi-mode prediction case is better than the single-mode frequency case due to it provides more sample characteristics.The best one is the 12 th case,which includes 2st-mode to 5th-mode frequency,with errors of 1.48%and 1.83%.(3)By comparing the theoretical model with the algorithm models,it is found that the SVM model also has the best prediction accuracy and stability in small sample data cases.The theoretical model is more predictable than the ANN model,mainly due to the high fitting.And ANN model is susceptible to data noise in small sample cases,resulting in poor prediction accuracy.In addition,the finite element method(FEM)based on frequency shifts,has been developed through the ABAQUS software,which mainly includes two parts:(1)Use a multi-step analysis method to establish a static load tensile finite element model based on frequency shifts as a damage parameter.Conduct a regular study on the frequency shifts of element damage during static load through the finite element method;(2)Using Python combined with the development of the ABAQUS subroutine,a fatigue-modal analysis model is constructed.The frequency is shifted with the increase in the number of fatigue cycles.The simulation results have a consistent curve trend with the experiment.Also,the experimental verification of the static load model shows that the frequency value has a shift with the increase of the displacement,which is consistent with the experimental phenomenon.Fatigue life prediction of the simulation case of the fatigue-modal analysis model shows that SVM has the best prediction accuracy,which is 0.64%.In summary,both the theoretical model and the algorithm model based on the frequency value constructed in this thesis can be used for composite material fatigue life prediction.In the theoretical model,the eighth-mode frequency prediction accuracy is the best.And the SVM algorithm has an outstanding prediction effect in small sample data,and the prediction accuracy with a multi-mode frequency combination is the best.Therefore,the multi-mode frequency combination could be used for fatigue life prediction in the algorithm prediction model. |
PDF Full Text Request