Research On Damage Detection And Life Prediction Of Composite Laminates Based On Bayesian Updating

The minor damage(including matrix crack and delamination)is easily formed in composite laminates under low-speed impact or cyclic loading,which is not easy to be detected.However,these "insignificant" internal damage evolves at an accelerated rate under load,which in turn causes fiber fracture and dislocation of the composite plate,and ultimately leads to the failure of the plate.It is difficult to estimate the loss caused by material failure if the composite plates are not detected in time during production or operation.Therefore,damage detection and residual life prediction of composite plates under cyclic load have very important research value and application value.In the first part,the fatigue damage of composite plates is analyzed quantitatively.Firstly,the finite element model of laminated plates with two typical defects was established by ABAQUS simulation software to simulate the propagation process of Lamb wave and extract the characteristic index of Lamb wave detection signals.Due to the different sensitivity and correlation between the feature index and the damage,the weighted comprehensive score method was used to screen the feature index and form the optimized feature vector.Then,the optimized feature vector of single path or different path was quantitatively analyzed.Finally,the finite element model was used to calculate the variation range of strain energy release rate as a health index.The superiority of strain energy release rate model in quantitative assessment was verified by comparing with the damage assessment method based on support vector data description.In the second part,a delamination localization imaging method based on Bayesian updating model is proposed to study the fatigue damage of composite plates under cyclic loading.Firstly,a physical model of the direct relationship between the delamination physical information and the sensor signal characteristics is established.Secondly,the optimal eigenvector set is selected which is the same as the finite element model.Then,the posterior distribution of damage location and size was obtained by Bayesian updating method.Finally,the experimental data of NASA and the results of ABAQUS finite element model were used to verify the theory,and the results showed that the Bayesian updating results were basically consistent with the actual damage.In The third part,the Adaptive Parameter Optimization Model-Particle Filter(APOM-PF)method is proposed to predict The fatigue residual service life of composite laminates.Firstly,a dynamic evolution model was constructed based on the strain energy release rate model for the damage state evolution(stiffness reduction,delamination area increase and matrix microcrack density increase)that occurred before the damage prediction,and then the particle filter algorithm was used to estimate the continuous evolution of the particle state.A key feature of the proposed method is the systematic control of various uncertainties,including model parameters,model errors and perceived noise,which is an effective control of parameter variances through the adaptive parameter optimization model.Finally,by setting a threshold value,the Remaining useful life(RUL)can be obtained when any damage parameter of the particle reaches the threshold value.Compared with the particle filter method without parameter optimization,APOM-PF is proved to be superior in controlling the uncertainty in the early stage of prediction,and makes the whole RUL prediction process more efficient and accurate.