Sensing Technology Of The Smart Environment Based On Distributed Optical Fiber Sensing

With the development of society and the advancement of technology, both the country and individuals have paid more and more attention to environmental security problems of personal and public property, society wealth and common instruments. To monitor safety problems of and to defense a site with fixed geographical location and a broader square such as the border, airport, military base, headquarters, communication center, traffic network, energy pipeline, lines of communication, etc., a Perimeter Security System is deployed with the ability to sense the whole environment intelligently. The complexities of such a system lie in the difficulties to (1) achieve an ultra-long-distance (50Km) 24×7 real-time monitoring in all-weather conditions, (2) achieve a smart awareness of the environment, i.e., intelligently process of environmental information collection and threatening activity recognition, and (3) adapt to environmental changes, i.e., realize an incremental learning function to deal with the new conditions or new damage activities. To overcome these three difficulties, one needs to consider many existing technologies, invent some new ones and properly integrate them into a complicated system, which involves so many branches of science and information technologies like sensing technology, electronics, digital signal processing, pattern recognition, software/hardware design and development, system integration, and computer networks. As a result, neither existing products from foreign companies nor Lab prototypes domestically developed has ever solved all of the three difficulties, especially the latter two, environment awareness and incremental system learning.This work is part of a National 863 project, in which a buried distributed fiber is used as the sensing system to collect the environment information, i.e., to sense micro vibration of soil in the surrounding area. Based on this sensing mechanism, we focus on intelligent processing and smart recognition of micro vibration signals. Specifically, the research work in this thesis mainly includes: 1. Analysis of the principles of information collected by a distributed fiber-sensing: we introduce a Mach-Zehnder sensing model of the interferometer based and distributed fiber used for environment information collection. Vibration signal, attenuation of external disturbance and locating of vibration source have been discussed.2. Analysis of the micro vibration signal: Taking the characteristics of real-time micro vibration signal into account, we developed an envelope curve and graphic morphology based algorithm to get vibration fragments, and we adopted an algorithm in the stationary wavelet threshold shrinkage class for the difficult noise filtering task.3. Pattern classification of micro vibration and smart awareness of environment: We applied both statistical analysis by Gram-Charlier series, and time-frequency analysis by wavelet packet decomposition to get a feature vector of each vibration fragment. Then we conducted principal component analysis and independent component analysis on the feature vector. A kernel space hierarchy clustering based SVM tree algorithm is presented in the classification phase. After that, the ideas of time series analysis are also introduced and applied to post-process the result of SVM tree and smart awareness of environment is realized. Finally, two levels incremental learning are proposed in classification phase, one for enlarged sample and the other for new class threatening, both with an appropriate scheme.4. Establishment of a general intelligent environment awareness and safety monitoring architecture: Take the above requirements and corresponding techniques into account, we established a general architecture with workflow"signal segmentation - signal de-noising - feature extraction - signal classification - activities analysis - human-computer interaction - adaptive incremental leaning".5. Implementation of a long-distance oil pipeline safety monitoring system: Based on the developed architecture, we have implemented a long-distance oil pipeline safety monitoring system and deployed it in real application by China Petroleum Pipeline Bureau.The main innovations of this thesis include: 1. It is the first complete and stable architecture presented in paper which deals with on-line analysis and intelligent recognition of real-time micro vibration signal.2. Based on the architecture presented in this thesis, we realized an integrated"physical sensing - signal processing - recognition model"system on the embedded computing platform under the hardware/software loosely coupling principles. As an ultra-long-distance vibration signal processing and intelligent environment recognition system, it has better performance compared with other systems reported by both domestic and international organizations.3. The long-distance oil pipeline safety monitoring system based on the architecture presented in this thesis has been put into operation by China Petroleum Pipeline Bureau. The results show that it has advantages of high precision and high intelligence. The system has excellent scalability and can be cascaded to meet various coverage levels from pipeline, region to nation. This work fills the domestic and international gap in oil/gas pipeline safety monitoring field.