Two-scale Probabilistic Predictive Method Of Fatigue Life For Metallic Structure

Metallic component is the critical mechanical part in the aircraft structure,the main failure mode of which is fatigue failure.Modern aircraft structural design becomes more sophisticated.Service environment,loading and processing technology are random.Lots of discontinuities exist in metallic component,such as hole and notch.These reasons make fatigue failure of metallic component still critical.At present,fatigue life predicting methods widely used in engineering are too experienced which leads to lower accuracy.It is difficult to handle fatigue failure and reliability.There are two main challenges to predict fatigue life.First,fatigue failure mechanism is very complicated.It is hard to evaluate kinematic properties of defects and accumulation of damage under random loading spectrum.Fatigue life predicting methods in engineering are mostly based on large amounts of test data and can't quantify fatigue damage evolution accurately.Second,scatter of fatigue life is very large mainly because of scatter of defects in material,uncertainty of the process,variable service environment and randomly fluctuant loads.So probabilistic fatigue life predicting method based on continuum damage mechanics(CDM)has great significance for theoretical research and engineering application.The main research contents of the dissertation are listed below.1.Fatigue test is expensive and time consuming.To reduce the cost,uncertainty method based on small size test sample is introduced firstly.The basic formulas are derived.Then influence of sample size and order on convergence is studied.In consideration of different distribution and dependency between random variables in engineering,Iso-probabilistic transformation and Nataf method are applied to polynomial chaos expansion.Numerical examples show that NIPC is more accurate and efficient than Monte Carlo.2.In order to improve p-S-N accuracy,probability is introduced to Basquin stress-life model by regarding model parameters as random variables.A new method to obtain probabilistic properties of model parameters are introduced based on NIPC and Bayes updating.The method uses numerical optimization to obtain samples of model parameters.Then construct NIPC expansion of fatigue life under some stress level for model parameters,based on which the probabilistic properties of model parameters can be obtained.Then fatigue life under another stress level is used to update probabilistic properties by Bayes method.Fatigue tests of aluminium alloy 2024-T3 plant with a hole under four constant amplitude loading are completed.Results show that the model can predict probabilistic fatigue life under constant amplitude loading.3.In order to quantify damage sequential effect due to macro elastic loading spectrum,two scale cyclic stress-strain algorithm is studied based on Eshelby law.Mesoscopic cyclic plastic curve is designed by Armstrong-Frederick kinematic hardening model.Principle of determining plastic model parameters and yield stress is given,which construct numerical algorithm for mesoscopic hysteresis loop under macroscopic elastic loading spectrum.Finally,whole two scale cyclic stress-strain curve is constructed by elastic prediction-plastic correction Newton algorithm.Numerical examples show that hysteresis is reasonable under variable amplitude loading.4.To improve predicting accuracy of fatigue life under random loading,a new two scale damage evolution model and non-intrusive numerical algorithm are proposed based on Lemaitre's damage evolution model.Based on metal tension test and fatigue test,accumulated plastic strain threshold won't be considered.The model can quantity the sequential loading effect on fatigue through hardening effect.An exponential term is introduced to represent nonlinear plastic curve.Mean stress effect is also taken into account by constructing transform relation for exponent of plastic strain with Goodman curve.To improve efficience,non-intrusive calculation program is coded to realize accurate stress-strain computation and successive damage accumulated.5.Based on small size test sample,two scale damage evolution probabilistic model is constructed in the thesis.We take the model parameters as random variables,and apply NIPC finite order expansion to the probabilistic properties of them.Then probabilistic fatigue life can be predicted by coupling sampling model parameters with non-intrusive calculation program.Fatigue test of aluminium alloy 2024-T3 coupon under constant amplitude loading and plate with a hole under random spectrum loading are performed.Test data is used to obtain model parameters and verify validation.Results show that model can predict fatigue life of typical metallic component under complex loading.Finally,probabilistic fatigue life of connection wallboard in C919 is predicted.Results show that macro-micro two scale probabilistic fatigue life prediction model is still applicative for complicated engineering component.