Damage Monitoring Of Resin Matrix Composites Based On Acoustic Emission

With the development of science and technology,composite materials play an increasingly important role in aerospace,automotive,shipbuilding,construction and other industrial fields.Damage monitoring of composite materials during service has become one of the research focuses.By monitoring the damage condition of composite material,the overall performance of the structure can be judged according to the damage condition,so as to avoid the safety problems such as the degradation of the overall structure performance caused by the internal damage and ensure the stability of the overall structure.In the production process of composite materials,there will be impurities,pores and other problems,and in the external impact,fatigue accumulation and other reasons,resulting in the internal composite material can not be observed,such as matrix fracture damage,delamination damage,debonding damage and fiber fracture damage.Based on acoustic emission detection technology,this study takes fiber-reinforced resin matrix composites as the research object,and studies the characteristics of four typical damage and damage mechanism of composites under different working conditions.According to the mechanism and characteristics of four kinds of damage of composite materials,different experimental samples are made to study their characteristics respectively.Based on the characteristics of each damage,the damage mechanism of glass fiber reinforced resin matrix composites under three typical working conditions is studied,and the damage process and characteristics of composites in elastic stage and damage stage are analyzed.A sample library is established according to the damage characteristics of composite materials under different working conditions.Partial least squares discriminant analysis(PLS-DA)and support vector machine algorithm(SVM)are used to establish the identification model to identify the damage stage of composite materials,and the accuracy of the identification model is compared and analyzed by the cross-leave-one verification method.The results show that:1)The main energy peak value of matrix fracture damage signal is concentrated in less than 80 k Hz frequency band;The energy peak of stratified damage is concentrated in the 80-220 k Hz band.The energy peak of debonding damage is concentrated in the 220-350 k Hz band.The energy peak of glass fiber fracture damage mainly concentrated in more than 350 k Hz frequency band;2)The recognition accuracy of the model established by the support vector machine algorithm is higher than that established by the partial least squares discriminant analysis method,and the recognition rate of the feature data extracted by the damage division frequency band is higher than that extracted by the wavelet packet decomposition algorithm.The final recognition rate is 86.7% under tensile condition,86% under bending condition,74% under compression condition,and the average recognition rate is 82.23%.3)In tensile condition,the main damage forms in the non-failure stage are delamination damage and matrix fracture damage,and the main damage forms in the failure stage are debonding damage and glass fiber fracture damage.The frequency characteristics of the two stages are quite different,and the recognition rate of this condition is high.In bending conditions,there are stratified damage and matrix fracture damage in the non-failure stage,and the main damage forms in the failure stage are stratified damage and debonding damage.In the initial stage of experiment and the later stage of sample failure,the frequency distribution is similar,which affects the model recognition results.The frequency characteristics of samples at the junction of two stages are similar,so the identification rate of samples at the junction is low.In the compression condition,the main damage in the effective stage and failure stage is delamination damage and debonding damage.The frequency peak distribution of the two stages is similar,but the frequency band energy is different,and the recognition rate of the compression condition is low.