Mechanical Fault Diagnosis System Based On Star Wireless Sensor Networks

Wireless sensor networks can effectively overcome some problems for the traditional cable condition monitoring system in some application of mechanical fault diagnosis, such as the sealed environment or the rotating environment in the mechanical fault diagnosis. But as a result of mechanical vibration signals in the fault diagnosis often need higher sampling frequency, high sampling rate of vibration data, but the wireless sensor network bandwidth and storage capacity is limited, so we can realize that original data real-time transmission and online monitoring is infeasible. An alternative way is conducted vibration signal feature extraction on the acquisition node firstly, we can choose transfer some features instead of the original signal transmission, and then implement fault diagnosis using the signal characteristic value in the PC terminal initially, finally comprehensive diagnosis of each node information accurate diagnosis results are obtained.According to the characteristics of the mechanical equipment fault diagnosis, this paper firstly analyzes the system requirements and determines the system hardware design and selection, including of the dual processor module, data acquisition module, data storage module, power module, wireless communication module design and choice. Then design and select software protocol stack, including of the only assigned ad-hoc network, star network topology, node synchronization acquisition and reliable data transmission mechanism design and choice. And then put forward the suitable mechanical vibration monitoring and reliable wireless sensor network architecture.Due to the limited storage resources of the WSNs mechanical fault diagnosis system, the node calculation traditional cable machinery fault diagnosis method for large amount of calculation can not be used in wireless sensor network node directly. Therefore this article puts forward the mechanical fault diagnosis method based on star wireless sensor network.In the terminal nodes, time and frequency domain feature set characteristics of scattering matrix is adopted to improve the classification ability of sorting, then take the top five eigenvalues characterization of mechanical state. Then transmitting the characteristic value of each node is the highest level computer algorithm of RBF neural network with more sophisticated preliminary fault diagnosis and DS evidence theory of multiple nodes decision fusion.The diagnosis results displayed in the machine fault diagnosis software interface in PC. Using this method can reduce the sensor node load effectively, save network bandwidth effectively, reducing energy consumption effectively, improve the efficiency of mechanical fault diagnosis and the diagnosis accuracy.Designing complete star wireless sensor networks mechanical fault diagnosis system based on dual core architecture G2.3 node hardware platform, adopting higher acquisition performance of IEPE accelerometer to obtain higher signal acquisition accuracy. The protocol stack based on synchronous acquisition should improve the acquisition synchronization precision and the reliability of the network transmission. Designing of embedded software based on scattering matrix feature extraction algorithm to improve the effectiveness of the nodes in the feature extraction.Designing the PC monitoring system of complete functions, friendly interface to improve the efficiency of mechanical fault diagnosis and the accuracy of diagnosis.Proposed the star the mechanical fault diagnosis system of wireless sensor network in this paper is verified by the experiments of can be effectively used in the actual fault diagnosis of mechanical equipment and diagnosis accuracy is 99%.Finally, the research work of this paper is summarized, and future research directions are discussed.