Study On Monitoring Of The Bridge Health Based On Local PC Cluster

With the increasing of the bridge accident, more and more attention is attached to the bridge health monitoring by the government and the engineering board. The real-time capability is acquired because of the randomicity of the bridge collapse, thus, the traditional monitoring system of bridge health needs to transfer lots of data to the remote data center from the bridge local. So, it can't be evaluated about the bridge structure health in time because the Real-time capability is not enough. The PC cluster is applied to the bridge health monitoring to improve the real-time capacity. It include storing the monitoring data at the bridge locale and evaluating the structure in time. The main work and the innovation in this paper are the following:1) The traditional bridge monitoring system have the following characteristic: the real-time capability is not high, the multiple PC's hypogyny is only used for gathering the data and the using rate is very low. Aiming at these characteristic, the necessity and the feasibility is analyzed to apply the field PC cluster in the bridge health monitoring system.2) Aiming at the analysis of the traditional bridge health monitoring architecture, It is bring out that a real-time and economical PC cluster architecture is used in bridge health monitoring.3) Aiming at the characteristic of the bridge health monitoring based on PC cluster, several solutions about time synchronization of distributed data gathering is brought forward and discussed in research.4) Aiming at the characteristic of the bridge health monitoring based on PC cluster, we also discuss the storage slice of the distributed data.5) The neural network algorithm is used often in the bridge health evaluation, so as to adapt it to the parallel environment of the bridge health monitoring PC cluster, the amelioration is come true in this paper. The idiographic amelioration is given according to the following steps: Firstly, we analyzed the various factors that affect the accelerating rate of the parallel neural network and we get the relation between accelerating rate and the number of connecting weights, we also gain the relation between accelerating rate and the number of neural nodes. Based on the analysis, we propose a new fitness function and use GA to predigest the neural network structure. Then we transform the neural node into the graph and take the graph's maximum path as the goal function to judge the good or the...