Study On The Optical Fiber Structural Health Monitoring System And Its Sensor Network Reliability

The intelligent structural health monitoring system based on the fiber Bragg grating (FBG)sensor network has been widely used in the civil engineering and aerospace fields because of itsoutstanding advantages. In practical engineering, FBG sensors are usually embedded in thestructure, once the sensing or transmission optical fibers are breakage or invalid, it is difficult torenew the invalid point. Therefore, the performance of the structural health monitoring will beaffected. For solving above problems, the optical fiber structural health monitoring system and thereliability of its sensor network are researched in this paper, the method that how to improve thedamage detection precision, fault-tolerant ability and FBG sensor netwok reliability are explored.Firstly, depending on the coupled-mode theory, Transfer matrix, color center model, limitedcladding radius Bragg grating theory, the principle of the performance degradation when the FBGsensor is suffered strain fatigue, high/low temperature, ultraviolet bleaching, chemical corrosionetc. are researched. By virtue of the optical waveguide numerical analysis software OptiFDTD, thesignal characteristic of the main parameters degenerated FBG sensor is analyzed, the resultsprovide a certain reference for judging and assessing the healthy situation of the FBG sensornetwork. Next, with the static load on the typical structure of the plane wing box test pannel andthe dynamic load on the Carbon Fiber composite structure test pannel as subjects seperately, byvirtue of the FBG sensor network signals embeded in the structure, the loading damagelocalization algorithm based on the support vector machine is researched, and the parametersoptimization method of the support vector is studied detailly. The grid search algorithm, geneticalgorithm and particle swarm optimization algorithm are proposed to optimize the mainparameters of the support vector machine, and the performances of the three mathods arecompared in this paper. Depending on the above research method, the reliability method thatpartial FBG sensors are invalid in the structural health monitoring system is researched. Forenhancing the damaging localization precision when the FBG sensors or fiber nodes are invalid inthe network, the model reconstruction which the support vector machine models and parametersare modified dynamically is research. The research results indicate that the proposed modelreconstruction algorithm based on above three damaging localization algorithm can almost keepthe predicting precision when no sensor, one sensor and two sensors are invalid in the structuralhealth monitoring, thus the reliability is improved when there are FBG sensors are invalid in thestructural health monitoring system.Secondly, for improving the reliability of the structural health monitoring, the optimumarrangement of the FBG sensor in the aviation structure is researched. The wing shape aluminum alloy test pannel is taken as research object, by virtue of the correlation coefficient betweendifferent FBG sensor data, the gradual accumulation method and support vector machine are usedto optimize the arrangement placement and numbers of the FBG sensors. First of all, enoughpre-arrangement positions are set in the monitored structure, and certain loads are applied on thedifferent positions of the structure. Thus the correlation coefficient between the loading andmonitoring (that is to say sensor pre-arrangement) positions are obtained. Then, one of position ischoiced as the first sensor arrangement position from all of the monitoring position. Next, thesmallest correlation coefficient relative to the first sensor position is choiced as the second sensorarrangement position. And so on, the smaller correlation coefficient relative to the selectedposition is choiced as the next sensor arrangement position. When one sensor arrangementposition is added to the selected sensor combination, the performance of the monitoring system iscomputed by the support vector machine, and the predicting accuracy is considered as theevaluation standard of selected sensor combination, until the positions and the numbers of theselected sensor meet the monitoring requirement. The research results indicate that the optimalarrangement scheme of the FBG sensor network relative to the initial arrangement mode isobtained, and the monitoring system predicting accuracy can be obtained with fewer FBG sensorsin the optimal arrangement scheme. Simultanely, with the same numbers of FBG sensors, theoptimal arrangement scheme can obtain higher monitoring accuracy.Thirdly, the high reliability structural health monitoring system based on the optimum FBGsensor network topology is researched. The optical switch is introduced in the traditional network,and the sensor network topological model is built depending on the adjacency matrix in the graphtheory, if there are some sensors or transmitted optical fiber invalid, the adjacency matrix of theFBG sensor network topological model is analyzed to ensure the switch sequence of the opticalswitch, thus the transmission path for the survival sensor node can be obtained again, and theinfluence of the invalid point to the sensor network signals are minimize to the smallest.Meanwhile, depending on the related reliability theory, on the basis of model reconstruction, thereliability of traditional and introducing optical switch FBG sensor network are compared andanalyzed separately. The research results indicate that the reliability of the introducing opticalswitch FBG sensor topology is higher than the traditional one obviously. When the failure rate ofthe single component is diffeent, the reliability of the two network topology is also different.When the single component failure rate change between0.001and0.01, if the ratio that thedistance error between the actual value and predicting value less than or equal to40mm countsdivide total sample counts reachs to95%, the failure rate of the introducing optical switch sensornetwork topology is reduced to50%than the traditional one. If the ratio that the distance error between the actual value and predicting value less than or equal to40mm counts divide totalsample counts reachs to90%, the failure rate of the introducing optical switch sensor network isreduced to12.5%at least than the traditional one.Finally, the high reliability FBG sensor network based on multi-agent technique is researched.Not only the multi-agent framework of the structure health monitoring system based on the FBGsensor network is designed, but also the agent’s structure and function, communication mode anddata fusion method between agents are designed, together with the static load on the plane wingbox test pannel as subject，the double agent cooperation is studied. In the first place, depending onthe sensor network optimization method above proposed, the arrangement position and number ofthe FBG sensor network adhered to the structure is determined. Following, the agent fusion resultthat each FBG sensor is in good condition in the agent and partial sensors can’t work as normal orsome sensors signal can’t be acquired in one agent is demonstrated separately. If all of the FBGsensor signals can be acquired correctly in each agent, the weighted average data fusion method isused to fuse the predicting result of the two agents directly for each external loading damageposition. If partial sensor signals can’t be acquired in one agent, at first, the model reconstructionalgorithm above proposed is used to compensate the invalid sensor signals in the faulted agent, onthat basis, the weighted average data fusion method is used to fuse the predicting result of the twoagents for each external loading damage, and the final predicting results are obtained for thestructural health monitoring. The research results indicate that the predicting accuracy of theexternal loading position is not only improved by the multi-agent technique, but also the reliabilityof the structural health monitoring system if partial sensors are invalid in the network is improved.