The Research On The Fault Diagnosis Of The Node Of Industry Wireless Network

With the Industry wireless network technology's developing and maturing, the reliability has become the key points for it to be widely used. Not only as the basis and carrier of the Industry wireless network, the network node's fault and failure will affect the normal operation of the network, but also as the base of the process control system and the monitoring of mechanical equipment, if not immediately, the fault will threat to the People's live and property. In the another aspect, because the harsh environment of the industrial field and the node's RF components and sensor components need to directly contact with the industrial field, which will increase the probability of the node's fault, it's necessary for the real-time fault diagnosis and monitoring. With the requirements of reliability and security increasing, the fault diagnosis have developed into a science, for the method including fault diagnosis based on analytical model, fault diagnosis based on signal processing method, fault diagnosis method based on artificial intelligence. But, at present, there are less paper and research about the fault diagnosis of the Industry wireless network's node, even not much about the fault diagnosis of the wireless sensor network's node. This paper takes WIA-PA as example for the Industry wireless network to study the fault diagnosis technology of Industry wireless network's node.Based on the in-depth studying the three main methods for the existing fault diagnosis, after highlighting the advantages of node-redundancy method and the reason for it's widespread application and research, considering that the Industry wireless network don't have the non-spatial correlation characteristics and wanting to improve the shortage of the existing methods'high rate misdiagnosis, poor real-time, strong-person -setting, and the dissatisfying for industry monitoring, the paper proposes a fault diagnosis method for industry wireless network node based on analytical redundancy by taking the large number of field data as realistic basis and the signal singularity of fault as theory basis, making use of the intelligence of artificial neural network and superiority of wavelet in the analysis to local time-frequency characteristics. On the basis, from the methodological point of implementation, the paper drive the mathematical model of method, provide the method process of implementation, study the key point of method. For the first key point,BP neural network technology, according to the shortcoming of the additional momentum and adaptive learning rate method for the BP neural network that do not consider the hierarchical difference of the false optimal between the hidden layer and outputs layer, the paper propose a improving algorithm based on level-different.For the second key point, the best choice of wavelet function,the paper choose the Db wavelet as the tool of diagnosis from the experience and perspective of mathematical properties of wavelet function. In addition, the paper preliminary study the method to the fault which considering the interference to the wireless communication signal of power and link quality which provided by the monitoring component of node, and proposes a fault diagnosis method for industry wireless network node based on analytical redundancy and monitoring component. Finally, the paper verified all the above method through tree simulation experiments. The first set of simulations results show the proposed algorithm based on level-improving have improves the learning rate of BP neural network by 10%. The second set of simulations results proves feasibility of the method of analytical redundancy. The last group of simulations results show the proposed method have lower rate of misdiagnosis, strong real-time and robust.