Research On Fault Diagnosis And Scheduling Optimizaton Scheme In Telecom Network

With the continuous development of Telecom networks such as 5G,more and more users are enjoying various services in the network,such as online education,e-commerce,etc.As the scale of the network continues to expand,and its structure becomes more and more intricate,the possibility of network fault increases gradually.When the network fails,if the administrator can’t diagnose the fault in time and schedule it reasonably,the network may be paralyzed.For this reason,it is very meaningful to diagnose the fault of network and to optimize the scheduling.Aiming at this topic,this thesis introduces a systematic solution of fault diagnosis and scheduling optimization in Telecom network,which includes the following steps:1)Fault diagnosis system of the whole Telecom network is built and the network data is preprocessed.Then,because the accuracy of fault diagnosis by conventional methods is not high and the fault diagnosis model is easy to fall into over-fitting state when analyzing large fault networks,a diagnosis scheme of fault area based on Boosting model is designed.In this scheme,the eXtreme Gradient Boosting(XGboost)model is introduced to construct the Gradient Boosting Decision Tree(GBDT)and realize the preliminary diagnosis of the fault area.Then,for the sake of obtaining higher accuracy of fault diagnosis and decreasing the consumption of time,the fault diagnosis model based on Light Gradient Boosting Machine(LightGBM)is proposed.By dividing features into blocks and setting dynamic thresholds to select large gradient samples,the computational complexity is reduced.The accuracy of the model is improved by combining grid search with features and using histogram algorithm to find the optimal segmentation points.The simulation results based on public data confirm that:1)the accuracy of fault diagnosis is improved from 70%to 99%.2)Compared with XGBoost model,the time complexity is reduced by 70%.2)In order to locate fault devices in the fault area,a diagnosis scheme of devices based on Graph Neural Network(GNN)is proposed.Taking the whole device network in the fault area as the research object,the state of some devices in the area is clustered,and the clustering results provide prior information for the construction of fault diagnosis model.By introducing the embedded representation of graph network,the network information of the whole device is aggregated and compressed,and the aggregated information is used to represent the whole device network.By establishing the GNN diagnosis model,the fault diagnosis of all devices can be realized,and the fault root device can be located.The simulation confirms that the fault diagnosis model built by Graph Convolutional Network(GCN)layer or Graph Attention Network(GAT)layer has low complexity,and the complexity is only O(n).In addition,the model built by GAT layer has higher fault diagnosis accuracy,which exceeds 93%.According to the location result of the fault devices,the fault devices can be marked accurately.3)In order to reduce the effect of fault propagation and improve the overall performance of Telecom network,firstly,the directed acyclic subgraph centered on the fault devices is extracted.Combined with the hierarchical structure diagram,the propagation model is built and the fault propagation link is predicted.Then,the construction of fault network model is introduced,and a scheduling optimization algorithm with selfadaptive and self-organizing ability is proposed.The alternative nodes of the fault devices can be independently selected by voting.Aiming at the problem of excessive traffic allocation of neighboring nodes,the maximum residual traffic redistribution method is recursively adopted to deal with it.Through simulation analysis,compared with traditional scheduling methods,the scheduling algorithm in this thesis can achieve higher network stability at lower network cost.In addition,the scheduling algorithm is suitable for device networks of different scale.