Research On WSN Node Fault Diagnosis Method And Design Of Test Platform

With the rapid development of microelectronics technology,communication technology,embedded technology,sensor technology and power technology,the application of Wireless Sensor Network(WSN)technology has been greatly expanded But with the wireless sensor network structure is more and more complex and the degree of automation continues to improve,while the wireless sensor network is mainly working in complex and harsh environments,to bear the impact of many adverse factors,which makes the sensor node is prone to failure and Performance degradation,which affected the performance of the entire network or even lead to network paralysis.Therefore,it is very important to study the sensor node fault and fault diagnosis for the characteristics of wireless sensor network.This paper firstly summarizes the wireless sensor network and its characteristics,and then introduces the sensor fault type and the current domestic and international research status of node fault diagnosis technology,and puts forward a new fault diagnosis algorithm based on analyzing the existing fault diagnosis algorithm.At last,the design of the wireless sensor node fault diagnosis test platform is designed.The design of the test platform includes the design of the sensor node fault information collector and the design of the wireless sensor node.And the sensor node fault information collector is used to collect the current value of the sensor node to obtain the fault diagnosis sample data.As the classification process of the nuclear limit learning machine is susceptible to the influence of the nuclear parameters,the genetic algorithm is used to optimize the nuclear parameters,At the same time,for the wireless sensor network node fault diagnosis input data with the redundancy attribute,put forward Rough Sets attribute reduction algorithm and genetic algorithm optimization kernel limit learning machine(GA-KELM)integrated fault diagnosis algorithm,that is,RS-GA-KELM fault diagnosis algorithm.Firstly,the algorithm is used to set the node fault diagnosis decision table,using the rough set attribute reduction algorithm reduce the decision table to obtain the simplest fault diagnosis decision table and diagnostic rules are extracted from the decision table,and the Neural network model of kernel extreme limit learning is established according to the diagnostic rules,then using the genetic algorithm to optimize the parameters of kernel extreme learning machine,making use of the given sample data for training the neural network obtained the best classification accuracy,finally using the optimal neural network model for fault classification of the input sample data and the fault diagnosis is realized.At the same time,the proposed algorithm for fault diagnosis experiment,compared with GA-SVM and GA-BP algorithm results show that the diagnosis has higher classification accuracy and use less time.Thus it can effectively solve the problem of node fault diagnosis,which has a certain practical value.