Mechanical Equipment Fault Diagnosis Based On Multilevel Hierarchical Fusion In Wireless Sensor Networks

As a brand-new technology for information acquisition and processing system,wireless sensor networks (WSNs) are characterized by self-organization, self-adaption,flexible deployment, extensibility, and inherent intelligent-processing capability. It canbe employed to build the wireless distributed system for mechanical equipment faultdiagnosis, avoiding the shortcomings of complex wiring, high-cost deployment, poormaintainability and expansibility of wired monitoring system in some applications, suchas realization of the fault monitoring of sealed, moving or rotating parts and the faultdetection of vibration, deformation or crack expanding of key components in theprocess of using, which has great application value in engineering.However, when WSNs are applied in the applications of mechanical equipmentfault diagnosis, whose network bandwidth is extremely limited, the informationprocessing model of sense and send can not meet the needs of the applications. But howto realize complex data monitoring and information reports of the mechanical fault isthe most important problem need to be solved by the sensor node with limited capacityand energy in WSNs. WSNs must use the information processing model of sense andprocess and fuse data on each node, to reduce the redundant data of network and thenetwork traffic, to save node energy and to improve the efficiency of informationcollection and system operation speed. So a mechanical fault diagnosis method based onmultilevel hierarchical fusion in WSNs is proposed in this paper, which is of greatsignificance to realize the application of mechanical equipment fault diagnosis based onWSNs.According to the characteristics of the application of mechanical equipment faultdiagnosis, The performance requirements of mechanical equipment fault diagnosissystem based on WSNs are analyzed in details. The architecture of multilevelhierarchical information fusion in WSNs for fault diagnosis is built. The informationfusion in WSNs is divided into three levels, data level fusion, feature level fusion anddecision level fusion. The design principle of the hardware platform of wireless sensornetworks and the wireless sensor networks communication protocol stack formechanical fault diagnosis are discussed. As the first choice for wireless sensornetworks, the IEEE802.15.4communication protocol is analyzed. Also the spectrumcharacteristics and mechanism of spread spectrum adopted by the protocol are analyzed. Then the features of ZigBee protocol based on which is introduced.The mechanical fault diagnosis method based on multilevel hierarchicalinformation fusion in WSNs is discussed, and a fault diagnosis framework of multilevelhierarchical information fusion in WSNs is presented. Hybrid domain featureinformation of the raw vibration signals is extracted by terminal nodes for the data levelfusion. The feature information is processed by cluster head nodes through using radicalbasis function neural network to get the pattern recognition results for the feature levelfusion. Using DS evidence theory, the recognition results are processed by the gatewayto assess the mechanical equipment operational condition for the decision level fusion.The multi-level information fusion is adopted to assign the computing tasks to nodes ineach layer，which can effectively balance the computing load of the sensor node.Through the transmission of a small amount of feature information instead of a largenumber of the original vibration signals, reducing the amount of network datatransmission, saving network bandwidth, and reducing the network energyconsumption.A set of complete system of multilevel hierarchical information fusion in WSNs formechanical equipment fault diagnosis is designed. Based on the dual-core WSNs-G2.2node, a high efficiency and low complexity embedded signal processing unit isdeveloped to make the node has the ability of real time vibration data fusion, and a massstorage system with the ability of data management is used to achieve effectivemanagement and real time storage of a large amount of vibration data, the data fusionsoftware of nodes in each layer are designed and then a hierarchical networkarchitecture of mechanical equipment fault diagnosis based on WSNs is constructed.The transmission formats of command packets and data packets of multilevel fusion arestandardized and illustrated. A host computer monitoring platform is developed.Through the experiments we find that this method can be effectively applied tomechanical equipment fault diagnosis based on WSNs.At the end of the thesis, the summarization of the article and expectation of therelative technology development are presented.