Application Of Cluster Analysis In Damage Diagnosis Of Tower Crane Steel Structure

Tower crane is widely used in various construction fields,and it is a very important construction machinery.However,because of the complex structure and working environment of tower crane,it is a kind of construction machinery with high risk and prone to all kinds of accidents.Many tower crane accidents are caused by damage to the steel structure of the tower.How to quickly and accurately find the damage of the steel structure of the tower,eliminate the danger in time,and ensure the normal operation of the tower crane has become the focus of attention at home and abroad.Cluster analysis has developed rapidly in recent years,and many new clustering methods have emerged.The application of cluster analysis in intelligent damage diagnosis of steel structure of tower crane is studied in this thesis.First of all,the finite element model of the tower crane is established according to the QTZ315 tower crane,and the modal analysis is carried out.According to the vibration characteristics of the tower crane,the dynamic load excitation function of the tower crane is calculated,the dynamic analysis of the tower crane model is carried out,and the dynamic data of different working conditions are extracted.The theoretical research and application of cluster analysis theory are compared,and density peak clustering and density clustering are selected as the methods of damage analysis.According to the characteristics of damage identification and density clustering of tower crane,a method of automatically calculating density clustering radius is proposed.Secondly,this thesis proposes a damage identification method based on AR model and cluster analysis.The residual variance and data mean of the model are taken as damage features,and the damage identification is carried out through the cluster analysis of damage features.The simulation data of different working conditions of the finite element model are extracted and verified,and the feasibility of the method is verified by the data obtained by single-limb experiments.In this method,the early warning of the damage state of part of the tower can be realized by using a single sensor.Finally,a damage identification and location method based on input-output model is proposed.the input-output model includes single-input single-output model and double-input single-output model.The single-input single-output model includes single-input single-output model based on cantilever beam and single-input single-output model based on time series data;double-input single-output model includes double-input single-output model based on cantilever beam and double-input single-output model based on time series data.The damage characteristics,namely damage factors,are extracted by the input-output model based on cantilever beam and the input-output model based on time series data,and the damage factors are located by cluster analysis of damage factors.In order to improve the accuracy of damage identification,the particle swarm optimization algorithm is used to optimize the model coefficients of the double-input single-output model based on time series data to improve the accuracy of damage location.The results show that the damage location of the tower can be located accurately by using double-input and single-output model combined with cluster analysis.