Research On Data Aggregation And Application In Machinery Fault Diagnosis In Wireless Sensor Network

As a new and highly interdisciplinary research field, WSN is now attracting more and more attention at home and abroad. The basic application of WSN is to gather and transfer the information of monitored region from sensor nodes. How to reduce the volume of data transmission to improve transmission efficiency is the biggest challenge in WSN. Data aggregation becomes an important research topic in WSN since its target of reducing redundant data, reducing communication volume, saving energy and increasing the information collecting efficiency.WSN has broad application prospects in military, environment, security, medical treatment, industrial and agricultural production, household living and etc. for its technical characteristics. WSN could be deployed in large complex devices to build monitoring systems, so that the costs of device inspection will be reduced. Meanwhile, the monitoring system will reduce downtime, improve efficiency, and extend the use of device time since it could identify problems in advance. At present, however, the applications of WSN in the machinery industry are still in initial stage. This thesis gives an in-depth study on data aggregation in WSN. In this thesis, the general machinery fault diagnosis methods are concerned, neural network, virtual force and fuzzy reasoning are combined with data aggregation technologies and the following work are completed:First of all, according to features of the WSN and application requirements of machinery fault diagnosis, a WSN system model are built. Requirements and design principles of different function modules in the model are concerned and the design scheme in hardware system as well as the design idea in communication protocols and operating system are still introduced. As the core of the system, WSN technologies enhance the system's flexibility, maintainability and expandability. Moreover, the system with modular and open structure has good portability.Coverage of a large number of sensor nodes deployed in the monitored region is the first problem in WSN, which directly affect the accuracy of the target monitoring. In complex machinery fault diagnosis applications, rational deployment of sensor nodes could not only cover the monitored region well and perform real-time fault diagnosis but also reduce redundancy and waste of network resources. In this thesis, a coverage enhancement algorithm based on virtual force of heterogeneous nodes is proposed and improved to be effective on coverage efficiency and coverage uniformity.A fusion tree establishment algorithm combining the routing technologies with data aggregation is proposed, which based on analysis of the methods of building fusion tree and takes the graphic central point as the fusion node. In this algorithm, the establishment of routing fusion tree could be divided into two stages: division stage and connection stage. The data will be fused in the center point node, so that the overall volume of transmitting data will be reduced and the network life time will be extended.Fuzzy Inference System(FIS) is introduced into measuring distance between sensor nodes, and FIS model for RSSI(Received Signal Strength Indicator) measuring distances is developed. Many ameliorations in existed physical network protocols are done in terms of dynamic location, which reduce energy cost in WSN, so that a node tracking technology based on data aggregation in WSN is realized.According to the research of theory and experiment, the data aggregation technology based on neural network in WSN proposed in this thesis is applied to machinery fault diagnosis. Circular statistics methods is used to extract the fault feature and then the WSN data aggregation based on PCA neural network and fuzzy neural network is used for the further fault diagnosis, so that the precision and reliability of the information will be improved effectively, and the transmission congestion as well as latency of network will be reduced.