Application And Research Of Data Mining In Fiber Optic Fault Diagnosis

With the rapid development of computer and network techniques, people's social activities become more dependent on communication network. This requires the network platform which provides and supports datas exchange to be more dependable and stable. The carrier of communications is optical-fiber, it has many advantages, such as communication capacity, long-distance relay, security performance, adaptability, etc; but in the event of failure, the loss is beyond imagination. As the impact of fiber failure is relatively large, the fault diagnosis of fiber-optic is also difficult and critical in the Network Fault Management, so fiber-optic fault management technology is widely studied by foreign scholars.Currently, there are many successful data mining application examples at home and abroad in fault diagnosis. BP neural network is an important issue of data mining. Selecting BP neural network technology to realize fault diagnosis capabilities of fiber, some physical quantity with complex causal relationship can be accurately reflected after the training of appropriate times.The main job of this paper is as following: (1) Develope an automatic switching protection system of optical-fiber based on VS2003 platform, it achieves real-time monitoring of optical-fiber communication, then sends optical-fiber power value to the main control module of system and stored it into history table. Main control module compares fiber-optical power value with presupposed threshold, then realizes autocontrol. (2) By means of real-time drawing optical-fiber power curves, it shows changes in transmission system intuitively. Users can estimate optical-fiber's stability by observing optical-fiber power curves. (3) Based on the research of BP algorithm, selecte samples of practical network model and structure training samples; preprocess selected samples formally; select three neural networks, confirm I/O and hidden layer nodes amount; determine suitable learning rate. After structuring suitable BP neural network, mining and forecast fiber-optical power value, then analyse unpredictable results.