Research On Fault Prediction Algorithm Of Rough Sets And Back Propagation Neural Network For Space Information Networks

Space information network is a comprehensive heterogeneous network.It can be used in many applications and it has a very important strategic significance.As the space information network is usually working in poor or hostile environment,it is easily destroyed and cause frequent occurrence of failure.Thus its quality of service declines.But before the failure occurs,which often accompanied by some of the characteristics of the signal or data trends indicate occurrence of failure.If these data can be effectively collected and judgment,that can achieve the effect of the failure prediction.Then which enable network to operate safely,reliably and stably.A fault prediction algorithm based on rough sets and error neural network is proposed.This algorithm is composed of three modules,which are data collected module,data forecasted module and fault prediction module.The data collected module established the discernable matrix decision tables based on the rough sets,and use the distinguish matrix for data reduction,At last,through the objective evaluation of attribute to exclude the subjective factors to get more accurate data for fault diagnosis.The data forecasted module use the gray model to predict the data,But,the predicted data and the actual value has certain deviation,so which uses the back propagation neural networks to analysis the error correction.Then,the predicted value as the input,the actual values as the output to train the networks,which completed that can get the accurate forecast data.Fault prediction module used the normal data and fault data collected by collected module as the input of the neural network,fault type as the output to train the network,then the error precision to meet the requirements to complete the training.Then the predicted data is input to the trained network to achieve the aim of the fault prediction.The network simulation software of NS2 and the simulation platform of Matlab was used to evaluate the proposed algorithm.Simulation results show that the proposed algorithm can be for fault diagnosis,Through the forecast data,and input the data into the fault predicted model,which achieve the purpose of fault prediction.