Research On Multi-agent Coordination And Fusion Approaches Based Large Structural Health Monitoring

Recently, structural health monitoring (SHM) technology is a research focus in the engineering and academic domain. For the actual large-scale structure monitoring, there are a number of sensors and various ones, which are distributed and dispersed, and the various evaluation methods. For the application, the integration of a wide range of sensors and different evaluation methods must be done. Hence, using the SHM technology on the large complicated practical structures, it is a critical problem that how to effectively manage the distributed sensor network, and coordinate and fusion distributed information and different evaluation methods for the efficiency evaluation of the system. In this dissertation, the multi-agent system (MAS) in Artificial Intelligence (AI) area is adopted to manage the large SHM, in which multi-agent coordination and fusion is intensively studied.The main works and novel researches performed in this dissertation include:1) The theory of the agent and MAS is summarized, and the conflict resolution and coordination strategy of the MAS is studied. Based on the theory, the agent model, the architecture and coordination of MAS based SHM are researched. Since there exits the complex sensor network and various damage assessment methods in a large SHM system, the architecture of MAS based on district monitor agent and central coordination agent is proposed, combining with the distributed and centralized structure. The architecture is the research foundation for large SHM with multi-sensor and multi-damage.2) To autonomously monitor the large structure with multi-sensor and multi-damage, a MAS based SHM is built. For three typical kinds of structure damages including strain distribution change, joint failure and impact load in the large-scale aviation aluminum plate structure, the system is developed with piezoelectric sensor, FBG sensor and strain gauge sensor, and their acquirement system. Multi-agent coordination is adopted to self-organize the sensor network and automatically choose sensing object, choose suitable signal processing method and damage evaluation method to recognize three typical structural states. At last, an effective evaluation is given for the damge state of the large structure.3) To fast and effectively cover each sub-region damage with sensor network in large structure, as well as to coordinate and fuse various damage identification algorithms to give a reliable and effective assessment for all kinds of damage, the multi-region monitoring architecture based on the multi-agent blackboard coordination is proposed for the large-scale structure. Based on the architecture, for the large-scale distributed structural strain distribution and damage classification monitoring, the self-organization network method based on the strain sensor network and the blackboard coordination, and the contract net coordination and fusion method based on the damage classification evaluation is proposed. Also, for large-scale structural distributed acoustic emission source and impact location, the self-organization network method based on piezoelectric sensor network and the blackboard coordination, and the server directory coordination and data fusion method based on the emission source location is proposed. The experimental verification is conducted on a large aviation aluminum plate to validate the efficiencies of the resource allocation and damage location.4) To improve the accuracy and real-time of the large structural impact load location, based on the improved inverse analysis method in time domain with Chebyshev polynomial basis function, for the advantages and disadvantages of the acoustic emission and the inverse analysis methods, a precise impact positioning method based multi-agent blackboard coordination is proposed. A large aviation aluminum plate and T-30carbon fiber laminated composite plate experiments show that the impact positioning method is fast and efficient, with good robustness.5) To improve the accuracy and real-time of the large structural joint failure identification, the identification method based on the classifier choice of the mutual information and the multi-agent decision fusion is studied. Firstly, the indicator and the algorithm of classifier choice are studied, and the algorithm flow of the classifier choice based on the mutual information degree of correlation is proposed. Secondly, the commonly used methods of the classifier combination are studied, and the multi-agent decision fusion based on the agent reasoning model with confidence, communication and coordination is presented. At last, a large aviation aluminum plate and aviation aluminum stiffened plate experiments show that the method can accurately and rapidly identify the damage.6) To accurately assess the structural damage for the SHM system with the sensor failure, the damage monitoring method of piezoelectric sensor network self-diagnosis and self-configuration based on the multi-agent reasoning and collaboration is put forward. Firstly, the research on the sensor self-diagnosis and self-configuration is analysed. Then, the multi-agent active and passive coordination monitoring based on piezoelectric sensor network self-diagnosis and self-configuration is presented to identify the debonding sensor with the active monitoring and the debonding fault factor, and coordinate to self-configurate the normal piezoelectric sensor network to obtain the active and passive evaluation of damage on the basis of the conflict resolution on the shared piezoelectric sensor. At last, an aviation aluminum plate experiment shows that the method is efficient.This work is completed in State Key Laboratory of Mechanics and Control of Mechanical (?)...