Research On Fault Diagnosis Algorithms Of Heterogeneous Wireless Networks Based On Deep Learning

With the advent of the big data era and the development of new network technologies,the mobile communications industry is undergoing tremendous changes.Facing the surge in mobile data traffic and the demand for various services,heterogeneous wireless networks have gradually become one of the important methods to improve system capacity.However,due to the large scale and high complexity of heterogeneous wireless networks,when faults happen or will happen,how to predict and locate them has become a great challenge.Traditional fault detection and diagnosis algorithms not only consume a lot of human and material resources,but also are difficult to establish an accurate mapping relationship between network symptoms and fault categories.Therefore,this paper combines the method of big data and deep learning to deeply study the intelligent fault diagnosis algorithms of heterogeneous wireless networks.Based on the introduction of existing network fault diagnosis methods,this paper mainly researches the following three aspects:(1)A fault detection and diagnosis algorithm based on convolutional neural network is proposed.Based on the analysis of the causes of complex heterogeneous wireless network failures,the algorithm first selects the optimal feature combination as the input parameters of the fault detection phase by the method of combining Relief F and mutual information to reduce the network characteristic parameters and relieve the burden of network computing and transmission resources.Then,the suspected faulty cells in the network are preliminarily screened by calculating the similarity of the time series data distribution.Finally,the fault cause of the suspected faulty cells is located through a fault diagnosis model based on convolutional neural network.Simulation results show that the algorithm has significantly improved the accuracy and delay of fault diagnosis.(2)A fault diagnosis algorithm based on improved generative adversarial network(GAN)is proposed.In order to solve the problems of high cost of manually adding labels to network data and the large convergence fluctuation of GAN,this paper first proposes a semi-supervised fault diagnosis model by improving the loss function of the generator network and the output layer of the discriminator network.Considering that the complexity of the model's discriminator network is too high to affect the speed and performance of model convergence,the model is further optimized.An algorithm combining generation adversarial network and convolutional neural network is proposed,in which the GAN is responsible for generating various fault data,and then use this data to train the convolutional neural network to complete network fault diagnosis.Simulation results show that the algorithm can also accurately diagnose network faults with only a small amount of labeled data.(3)A fault diagnosis and verification platform based on deep learning is designed and implemented.First,based on OPNET,a heterogeneous wireless network scenario with cross coverage of macro base stations and micro base stations is built and running system collects corresponding network data.Then write a program to screen the optimal feature combination,build a fault diagnosis model of improved generative adversarial network and convolutional neural network,and realize the generation of network data and the diagnosis of network faults.Finally,the effectiveness of the fault diagnosis algorithm combining generative adversarial network and convolutional neural network is verified.