Application Of Transfer Entropy Theory In Identification Of Structural Dynamic Nonlinearity And Damage

In this article,the transfer entropy theory is proposed to quantify the difference between the structural dynamic responses under different conditions from the point of view of entropy on the basis of a comprehensive review on structural damage identification methods,so as to identify and evaluate the nonlinear characteristics of the structures.This method is able to partly solve some deficiencies and technical problems in the field of nonlinear damage identification to a certain degree,and can provide reliable theory basis for structural health and safety monitoring in practical engineering.Based on the characteristic of transfer entropy that it can be used to quantify the asymmetric information transmitted between different signals,the differential transfer entropy index is proposed to analyze ultrasonic non-destructive detection signals and evaluate local damage of structures such as beams and columns.By changing the characteristic acoustic parameters of the simulated signals,the propagation characteristics of ultrasonic signals in the damage structure are able to be represented.The results verify the effectiveness and sensitivity of the newly proposed index in identifying the change of ultrasonic parameters in acoustic signals.In addition,the finite element beam model based on multi-physical field coupling is also constructed,and the different damage conditions of the structure are simulated by deducting the elastic modulus of the specified area and setting the internal cavitation defects,which further verifies the effectiveness of the algorithm in practical application.The results show that this damage identification method based on differentiated transfer entropy can provide a more stable and reliable judgment basis,and partly solve the insufficiency of current ultrasonic nondestructive testing methods that the results rely too much on subjective judgmentThe transfer entropy algorithm combined with the alternative data method is then introduced to evaluate the nonlinear features of the original data by evaluating the discreteness degree between the transfer entropy results of the original data and that of its corresponding surrogate data.The applicability and feasibility of the method are verified by using time-history data generated by different nonlinear systems.The research shows that the method of transfer entropy combined with alternative data can be used to identify the occurence of nonlinear features of the system only based on the given real-time data without the dependence on linear data accurately and effectively,and this algorithm has also shown high robustness against noise.Considering that nonlinearity may be influenced by many factors,in order to reduce the impact of experimental or human errors of the data used when validating the effectiveness of the method,a simply supported thin plate subjected to impact loads is studied using analytical method.The analytical solutions of the vibration response based on two different deformation theories are calculated.Appropriate evaluation indexes are defined to quantify the nonlinear degree,the results verify that the proposed method is able to identify the nonlinear characteristics caused by the large deformation effectively,and its corresponding nonlinear transfer entropy evaluation index will not be influenced by the change of structure geometry,it is only related to the nonlinear degree of the structure and shows quite well stability.The identification of dynamic damage based on the nonlinear vibration response of a damaged cylindrical shell under radial load is analysed by using the transfer entropy theory.By changing the defect coefficient of the cylindrical shell,different damage states of the cylindrical shell are able to be simulated.Then,the differential transfer entropy algorithm and the surrogate data transfer entropy algorithm were applied to identify the damage of the cylindrical shell structure for two different cases that whether the original data of the dynamic structure under healthy condition is available.Through the comparison and analysis of the two results,the reliability of the algorithm in identifying the structural damage is verified,and the research results show that the transfer entropy theory can also achieve quite accurate identification results for the minor surface defects,which shows its potential use in practical engineering.New concepts and methods such as bifurcation and chaotic motions are introduced into the analysis for dynamic structures.In view of the nonlinear nature of chaotic systems,the transfer entropy theory is applied to further study the nonlinear dynamic behavior of plate and shell structures under different loading conditions.The results show that the transfer entropy algorithm can identify the chaotic characteristics of the system accurately and effectively.Moreover,through the comparison between the transfer entropy results of different dynamic systems,it can be found that the nonlinear degree of the response of the thin plate structure is lower and the nonlinear characteristic is weaker,while at the same time,the dynamic responses of viscoelastic structures generally show more nonlinearity than those of elastic structuresBased on the theoretical analysis,analytical derivations as well as numerical simulation,this thesis systematically analysis and verifies the reliabilities and advantages of transfer entropy-based evaluation index in the identification of dynamic nonlinearity for beam,plate and cylindrical shells,for the purpose of providing a novel and feasible research method for real-time safety monitoring and damage identification for structures.