Research On Fabrication And Reliability Analysis Methods Of All Carbon Fibre Composite Truss

Composite truss is an important structural type in aerospace vehicle for its light weight,large span and flexibility.In order to further improve its load-carrying efficiency and take into account the structural parameter uncertainty cause d by process dispersion,this dissertation performs research works on all carbon fiber composite truss,including fabrication,analysis and test of load-carrying performance,structural reliability method,etc.The main research contents in this dissertation include:The fabrication method with integral assembly is proposed and established for all carbon fiber composite truss.The fabrication method abandons the way of separately molding joints beforehand,which is usually used.It directly connects prefabricated bars on the spot with composite and incorporates the two process,joints molding and truss assembly.With this method,the integral fabrication with high geometric precision is made for all carbon fiber composite truss.A fixing device of rods is designed which is composed of modular triangle plates.The design ensures the geometrical precision of molded truss.A rods connection method with composite reinforcement is invented,which is constituted by three procedures.They are gluing shell,winding tow and covering fabric,successively.The connection method avoids the use of metal connections,guarantees connection strength and achieves a higher load-carrying efficiency.Moreover,a split-type curing device is developed specially for single connection.These three aspects make connection implementation simple and flexible,reducing time and costs of truss fabrication.With the proposed fabrication method,an all carbon fiber composite truss sample is fabricated which has long narrow beam-like form and hollow triangular cross section.Its advantages on both load-carrying efficiency and geometrical precision are validated by comparison with two typical commercial truss.Simulation and validation of load-carrying performance of all composite truss are performed,in order to establish a simulation model which is accurate and efficient.The simulation model combining dense and sparse mesh is established,based on the idea of structural multi-scale finite element and by giving the connection method of interface between different grid models.The simulation model can take into account local action of joints.It greatly improves the simulation efficiency while keeping almost the same precision with a entirely refined simulation model,providing a highly efficient simulation model for further truss reliability analysis.The experimental devices of bending,torsion and vibration specially for truss are established,and accuracy of the simulation model combining dense and sparse mesh is verified by the experimental results.For the test of truss torsional stiffness,a measurement method for twist angle is developed based on displacement transformation.The method is able to precisely measure small twist angle.Also,the results of the simulation model combining dense and sparse mesh are compared with those of the entirely refined simulation model and a beam element model,validating that the simulation model combining dense and sparse mesh can take into account both precision and efficiency.In view of the requirements in initial stage of truss design,the homogenization based equivalent idea is introduced,and then the equivalent analysis methods for truss torsional stiffness and vibration properties are given through the equivalent of a closed thin-walled bar and that of a uniform beam,respectively.In order to reduce the reliability analysis cost of composite truss for single failure mode,the reliability method with local Monte Carlo simulation and based on adaptive Support Vector Machines is established.The reliability method adopts a simulation manner that requires less samples and an efficient method for samples classification.The reliability method effectively reduces computational cost while keeping accurate.In the aspect of simulation manner,local Monte Ca rlo simulation is proposed,which aims at only important region,and the definition of important region is given accordingly.Also,the H-L reliability index is introduced to directly identify safe points.These effectively decrease samples size and classi fication difficulty.To efficiently classify samples,adaptive manner is adopted for the construction of Support Vector Machines,which is used as a classification surrogate.For the construction,an adaptive strategy is proposed which is based on the poin t selection criterion with most likely support vector.The construction method allows a lower cost for samples classification.Accuracy and efficiency of the proposed reliability method are validated through analysis of typical examples and that of the present truss.Based on the same idea as the proposed reliability method,a reliability sensitivity method for single failure mode is established,which is no longer costly.The reliability and reliability sensitivity of single failure mode of all composite t russ are analyzed with the proposed methods.Failure probability and its influence rule of each single failure mode are determined.To efficiently perform system reliability analysis for all composite truss with multiple failure modes,Radial-based Importance Sampling is introduced for its robustness.Efficiency of each stage of Radial-based Importance Sampling is improved by fully excavating prediction information of Kriging.It is then proposed the reliability method with Kriging based Radial-based Importance Sampling(K-RBIS).First stage of K-RBIS establishes an approximation process for optimal radius with failure point search as the center.This process makes quick determination of optimal radius.It defines search domains for failure point based on limit state point.With this definition,ranges of intermediate approximate optimal radiuses can be well controlled.A search method for failure point is proposed based on Kriging prediction in the approximation process,making quick acquisition of a failure point possible.The twice convergence condition is defined for the approximation process,accelerating approaching efficiency while ensuring accuracy.In the second stage of K-RBIS,samples are classified by Kriging surrogate,which is refined adaptively through active learning.And the initial experimental design of the active learning procedure is composed of all computed points in the first stage.As a result,cost of samples classification can be largely reduced.Further more,cost of system reliability analysis can be reduced by improving the whole process of K-RBIS.The improvement is according to the idea that Kriging directly approximates the performance function of single failure mode.Accordingly,the system reliability method with Kriging based Radial-based Importance Sampling(system reliability K-RBIS)is established.Besides,system reliability sensitivity method is established according to the same idea of system reliability K-RBIS,and the evaluation method of contribution of each failure mode to system failure is given upon information produced by system reliability K-RBIS.These two methods both require no computational cost.The accuracy and efficiency of K-RBIS and system reliability K-RBIS are validated through analysis of examples.The system reliability of all composite truss is analyzed comprehensively with these proposed methods for system reliability analysis.The system failure probability of the truss and its influence rule are determined,which provides basis for reliability desig n of truss.