Application and Research of Data Mining in Fiber Optic Fault Diagnosis
|School||North China Electric Power University|
|Course||Applied Computer Technology|
|Keywords||Fault Diagnosis Fiber failure Data Mining BP neural network|
With the rapid development of computer and network technology , people's social activities are increasingly dependent on the communication network . This need to provide and support these data interaction network platform has a higher reliability and stability . The communications carrier optical fiber communication capacity , long distance relay performance of confidentiality , adaptability advantages , but in the event of failure , caused by the loss but also can not imagine . The impact fiber failure , fiber failure diagnosis is the difficulty of network fault management and key , so the the fiber fault management technology scholars widely studied . Currently, data mining applied to fault diagnosis have many success stories . BP neural network is an important topic in the data mining research , select BP neural network technology to predict fiber fault diagnosis function , the complex causal physical quantities , after the appropriate number of training more accurately reflected . The main work is reflected in the following aspects : (1) based on VS2003 platform developed optical fiber protection system automatically switches to achieve real-time monitoring of communication optical fiber optical power value is reported to the system 's main control module , and stored in good data acquisition , data mining after work history table . Master module to compare the reported data with the preset threshold , to achieve automatic control . (2 ) real-time rendering of the curve of changes in the optical power , to visually display the optical power of the optical fiber transmission system changes , the user can determine the stability of the fiber runs through the observation of the power curve of the fiber optical . (3) on the basis of the study BP algorithm , BP algorithm improvements, a sample selection of the actual network model , constructed training samples ; standardization pretreatment of the sample selected ; select a three - layer neural network to determine the network's input output nodes and hidden layer nodes ; determine the appropriate learning rate . Constructing appropriate BP neural network , the optical power value forecast mining , analysis and prediction results . The experiments show that the BP neural network of optical fiber fault diagnosis data mining is feasible .