Research On Health Monitoring Method Of Single-layer Latticed Shells Based On Model Updating Technique

Recently,computer intelligence algorithms have been gradually applied in the health monitoring of large-span spatial structures.There are two main points of health monitoring:Firstly,data acquisition,which including ensuring instrumentation and monitoring methods.It involves determining the sensor layout scheme.Secondly,analyze the collected data and identify structural damage.Today’s research on both will involve intelligent algorithms.Compared with the general structure,single-layer reticulated shell structures have a large number of members and joints,while own complex modal vibration.At the same time,environmental factors will also affect the accuracy of modal parameters and the effectiveness of damage indicators.Therefore,the requirements of the health monitoring intelligent optimization algorithm are higher.Based on the above,this paper focuses on a group-evolution algorithm based on statistical-learning theory: the estimation distribution algorithm(EDA).The algorithm uses statistical learning methods to establish a probability model describing the distribution from the macro perspective of the group.By establishing a probability model to describe the distribution information of the candidate solution in the search space.New populations are then produced by random sampling.Iterates over and over again to achieve the population’s evolution.In this research,in the main function of EDA,Bayesian neural network is called to establish the probability model,a function of applicability based on modal kinetic energy is constructed.Call the appropriate sub-function to implement learning and sampling.Then,the sensor optimization arrangement of the structure of one-dimensional simple support beam,two-dimensional truss and single-layer cylinder shell structure are realized by means of model updating.This is the first time that a sensor optimization arrangement has been implemented using a distribution estimation algorithm.After testing by the modal confidence criteria,the optimization results of this method are correct.In the main function of EDA,Gaussian model is called to establish the probability model,a function of applicability based on damage-sensitivity is constructed.Call the appropriate sub-function to implement learning and sampling.Then,the damage recognition of the structure of one-dimensional simple support beam,two-dimensional truss and singlelayer cylinder shell structure are realized by means of model updating.And to a certain extent to achieve the degree of damage identification,the average accuracy of the degree of damage recognition is more than 90%.What’s more,a model test of the single-layer latticed shell was carried out.Model updating technique was applied to the results to identify the position and degree of the damage.The results show that the method can identify and locate the damage of the mesh rod.But the overall accuracy is lower than the numerical test results.The recognition accuracy average is 86.10%.The rod with low recognition accuracy is located near the support and damage rod.The structure form involved are from simple to complex,gradually apply distribution estimation algorithm to the intelligent monitoring of single-layer cylinder shell.The research of this paper further expands the application of computer intelligent algorithm in the field of intelligent monitoring of civil engineering,which is of great significance to the promotion of intelligent monitoring and safety-early-warning technology in large public buildings.