Study On Sampling Inspection Method Of Reticulated Shell Structures In Service

The reticulated shell structure in service will appear structural damage or deformation due to various reasons during the service period,which will lead to the degradation of the bearing capacity of the structure and even the potential safety hazard,and directly affect the safe use of the reticulated shell structure.In order to ensure the normal and safe use of the reticulated shell structure,periodic sampling inspection is always carried out in practice.However,the existing sampling inspection method of the reticulated shell structure have shortcomings in the determination of sample size of test items,the calculation and classification of importance of components and nodes,the determination of sample location of components and nodes,and the estimation of unmeasured defects.In order to solve these problems,this thesis firstly summarizes the current research of sampling inspection method,and then gives the corresponding research results for each problem.To solve the problem of determining the sample size of each test items with the limit of total sampling inspection cost,this thesis firstly gives the confidence interval of the probability distribution characteristic value of the defect or damage based on the initial samples,and then the reduction rate of the variance of statistical indicator(RRVSI)is defined as the reduction rate of variance of bearing capacity with the increase of sample size.Aiming to maximize RRVSI,the sample size determinization method of reticulated shell structures considering mixed uncertainty is proposed.Besides,based on the influence of test items on the uncertainty index of bearing capacity,the importance determinization methods of test items are proposed,which contain scatter plot method,variance method and random forest feature method.To solve the problem of the importance calculation and classification of members,this thesis proposes two different importance calculation and classification methods respectively from the perspectives of local sensitivity and global sensitivity analysis.In the view of local sensitivity,this thesis defines the critical point total strain energy based on strain energy theory,and determines the importance index of member based on the impact of local component damage on the critical point total strain energy.Thus,the classification method of member importance with variable importance index boundary is proposed based on k-medoids method.In the view of global sensitivity,the elementary effect method is introduced into member importance classification,and the elementary effect for member importance calculation is also defined.Then,to divide members are into categories of important and normal members,a two-stage classification method which contains trial calculation and formal calculation is proposed.Besides,the ranking method for important members is proposed based on Technique for order preference by similarity to an ideal solution(TOPSIS)Decision Method.To solve the problem of importance classification of nodes,this thesis firstly defines the nodes whose importance indexes differs greatly from other nodes as outlier nodes.Secondly,the initial classification results of non-outlier nodes will be evaluated based on k-means algorithm.Then the classification results are evaluated based on silhouette coefficient.While the classification quality is poor,the number of importance categories should be modified.In the end,based on engineering need,different importance categories of non-outlier nodes and outlier nodes are combined into the prescribed importance categories.Based on the above steps,a two-stage importance classification method based on k-means algorithm for nodes is proposed.To solve the problem of estimation of the defects of unsampled members,the correlation model of member defects is proposed based on the distance between members,and the standard influence distance is defined to represent the influence range of member defect correlation.Then,an updated Kriging estimation method(UKEM)is proposed based on the Kriging model.UKEM firstly constructs the initial Kriging model with a small number of initial samples.Then,according to certain criteria,the optimal sampling location of new sample in the subsequent sampling process is predicted,and the Kriging model is updated based on the inspection data of the new sample until the convergence index is met or the total number of samples reaches the upper sample size limit.To solve the problem of node positional deviation inspection of reticulated shell structures in service with “deviation aggregation” phenomenon,the node positional deviation aggregation is defined and calculated based on the theory of spatial local autocorrelation,and an improved adaptive web sampling(IAWS)method is proposed,which consists of three parts: calculation of node connection matrix,adaptive web sampling,and supplementary sampling.The node connection matrix expresses the connection relationship between nodes and provide a basis for IAWS,adaptive web sampling focus on the exploration of aggregation areas and can improve the aggregation area sampling accuracy,and the supplementary sampling is designed to improve the overall reckoning accuracy.The research results of this paper can provide necessary theoretical analysis methods for sampling inspection of reticulated shell structures in service,and improve the sampling inspection efficiency and accuracy.