Research On Intelligent Fault Diagnosis Algorithm In Multi-domain Network

Continued expansion of network applications and business needs to make the computer network more and more large-scale,resulting in network management becomes more complex,high cost of operation and maintenance,and network fault diagnosis is the core of network operation and maintenance issues.It has become an urgent issue for the network operation and maintenance that how to provide high robustness and high reliability of network services,reduce the cost of operation and maintenance and diagnosis the network failure fast and accurately when the complex network system fails.At present,for the diagnosis of network failure,a lot of research work has been carried out from a variety of technical points.However,face of an increasingly large network system,there are still many problems remaining to be solved,such as the high cost,low efficiency and low accuracy of fault diagnosis.In view of the above problems in network fault diagnosis,we must find a feasible network fault diagnosis model.Based on the research and observation of the network operation and maintenance personnel on the network failure,the occurrence of network failure is often accompanied by the state of the network to show a certain state,and form a certain model.So the network fault diagnosis problem is also a pattern recognition problem.With the great progress made in the pattern recognition of artificial intelligence technology in recent years,how to use the network state information and the experience of artificial intelligence to provide a new research idea for network fault diagnosis.In this paper,we focus on the network scene of the sub-domain in the current network,and make a new consideration on the problem from the perspective of centralized fault diagnosis.In particular,the main research work and research results are summarized as follows:1)Aiming at the problem that the training sample dimension is too high in the centralized fault diagnosis model,a fault diagnosis algorithm based on SVM is proposed.The algorithm is designed to find the hyperplane in which the fault samples can be separated in the sample space,and the problem of finding the hyperplane is eventually transformed into the convex quadratic programming problem with maximal interval.Therefore,the algorithm is not sensitive to the sample dimension.In this paper,the effectiveness of SVM fault diagnosis algorithm is proved by simulation and verification of all kinds of fault diagnosis results.2)Aiming at the problem that the fault sample space is getting bigger and bigger,a neural network fault diagnosis algorithm based on PCA-BP is proposed from the point of view of dimension reduction of fault samples.The algorithm first maps the samples to low-dimensional space by PCA technology,finds the main components of the samples,and finally inputs the new samples into the BP network for training of the diagnostic model.In this paper,the accuracy of the algorithm is analyzed by simulation,and the effectiveness of the PCA-BP fault diagnosis algorithm is proved.3)Aiming at the problem that the effective information may be lost in the process of PCA dimensionality reduction,a fault diagnosis algorithm based on deep convolution is proposed.The algorithm solves the advanced features by performing hierarchical convolution on the network feature view,depth learning of the original low-level signals,and then performing diagnostic learning for these advanced features.In this paper,the simulation results are compared with the accuracy of PCA-BP algorithm,and the effectiveness of the algorithm is proved.