Study On Optimal Sensor Placement Based On Clustering-Effective Independent Method

Optimal sensor placement is one of key problems of designing structural health monitoring system and,because it has an essential effect on the availability,storage and processing of data collection.Due to cost and field conditions constraints,a given number of sensors are limited,only a minimum number of sensors is placed at the right locations.So how to use a small number of sensors to obtain the most effective information for the measurement points is a key issue for optimal placement of the sensors.In this thesis the problems in the optimal sensor placement algorithm and evaluation method are studied,including the modality observability and damage identification inconsistency of the measurement point data,the concentration of the measurement points,and the evaluation of the sensor optimization scheme.It is impossible to give a quantitative evaluation of the sensor optimization scheme.The evaluation has carried out systematic research,and the research mainly includes the following aspects:(1)There are many evaluation criteria for sensor optimization schemes,but each evaluation criterion can only evaluate the merits of individual aspects and cannot conduct comprehensive evaluation.Based on the five evaluation criteria,this paper establishes a comprehensive evaluation system with five evaluation indicators.The comprehensive evaluation system can be used to quantitatively calculate the comprehensive score of the optimization scheme,which reflects the advantages and disadvantages of the scheme more intuitively and accurately.(2)In actual engineering,the number of sensors is often larger than the number of monitoring modals.At this time,the clustering characteristics of the sensor points will be optimized.To this end,the binary k-means clustering algorithm is introduced to classify the degrees of freedom according to the similarity of vibration characteristics,and divide all degrees of freedom into multiple clusters.The clusters and clusters form disjoint subsets to avoid similar points being selected.,causing problems such as measuring point waste and information redundancy.(3)For the problem of modal observability and damage identification inconsistency,the strain mode is introduced,and the vibration displacement response and strain response equation of the system are used to establish a dual control information matrix containing both structural damage information and modal information.Clustering-effective independent method.(4)Using the proposed cluster-effective independent method to optimize the truss and arch rib displacement sensors of Nanjing Dashengguan Yangtze River Bridge,and compare it with energy method,sequence method,effective independent method and sensitivity coefficient method.To verify its effectiveness.In addition,in order to study the influence of the number of sensors on the optimization algorithm,the relationship between the optimization algorithm and the evaluation index when the number of sensors changes,the validity of the method is fully verified.