Optimization Design Of Structure Health Monitoring System In WSN-based SHM

In recent years, combination of wireless sensor network (WSN) and structuralhealth monitoring (SHM) has attracted increasing attention. Relevant research showsthat, in most of the WSN applications the major energy intensive operation istransmitting a large amount of data for a long distance. Compare to transmission costthe computation cost can be ignored, so previous research focused on reducingtransmission energy. In SHM, it tends to create far more amount of data than otherWSN applications. It is necessary to process raw data in-network, through reduce theamount of data need to be transmitted, energy can be saved. However, in-networkprocessing needs uniform analysis on sample data from multiple nodes. The significantlarge amount of data, complexity of the algorithm and the limited computing power ofthe wireless sensor nodes make the computation energy become a problem worth to payclose attention to.In SHM, the major in-network processing method is extract modal parameters thatdamage detection needed through modal analysis. Existing research works focus on howto realize modal analysis in WSN. Though the computation energy in SHM is worth topay close attention to, there is few research works consider to reduce systemcomputation energy. So, this paper will start from the angle of reducing computationcost in in-network processing, discuss optimization design for damage detection method,optimize clustering and then implement. Our work includes three parts:(1) This paper proposed an energy-efficient damage detection method TS-DD. Thispaper takes the AR-ARX damage detection algorithm as prototype, by using randomdecrement technique, extracts free vibration response from sample data belongs tosingle node, to reduce the amount of data that AR-ARX needs to analyze. Thenconsidering free vibration response from different nodes contain redundancy, by usingPCA, reduce data dimension from multiple nodes through cluster based communicationstructure, to save the computation cost waste on analyzing redundancy. Experimentresult shows, compare to AR-ARX, adopting TS-DD can reduce system energy by44.06%.(2) This paper proposed an optimized clustering algorithm. Most existingclustering research focuses on transmission cost, and always takes the residual energyand the communication range between nodes as basis. However, in SHM, computationcost is affected by sensor node placement, which leads to the inapplicability of existingclustering algorithms. This paper analyzes the relationship between computation costand clustering, proposes the idea of saving system computation cost through optimizingclustering for the first time, and implements this idea through genetic algorithm.Simulation result shows, compared to AR-ARX, adopting TS-DD combine with optimized clustering can reduce system energy by49.1%.(3) This paper designed a cluster-based wireless SHM system, treating the damagedetection function as a sub-system which can realize in-network damage detection. Weselect the hardware which can support the function of system, and implement thedamage detection sub-system by TS-DD and optimized clustering. We have finishedexperiment and simulation in PC and Imote2, and measured the system energy.