Technologies Of Multi-agent Collaboration And Information Fusion For Large-scale Structural Health Monitoring

As an important branch of smart materials and structure, Structural health monitoring (SHM)technology can obtain health state information and make security assessments through functionalelements integrated in the structure. In the process of the health monitoring technology developedfrom the laboratory to the practice, various problems will arise inevitably for the harsh and dynamicenvironment increasing a number of sensors and drivers which can make corresponding signalprocessing and damage diagnosis more complicated in monitoring the large-scale and complexstructures.In order to simplify the design of engineering structural health monitoring system, it is a must tocoordinate and manage the health monitoring system effectively and extract structural damagecharacteristics from the mass monitoring data to conduct security assessments.In this paper, the complex system is decomposed into several subsystems and a large structure ofthe hierarchical multi-agent health monitoring system is built based on task decomposition theory andBDIKR architecture model. Also, systematic study is done due to multi-agent collaboration,coordination mechanisms and multi-source information fusion mechanism. What’s more, the aviationaluminum siding structure is selected to verify the research results and the mobile agent collaborationand information fusion mechanism is achieved based on wireless sensor networks which can optimizethe system. This article consists of the following innovations:1) In view of the large volume of practical engineering structure and the complexity of the healthmonitoring system, the task decomposition theory and the BDI multi-agent system model theory arestudied. Combined with expert knowledge and structure analysis, the hierarchical multi-agent healthmonitoring system is put forward based on BDIKRI model for large structure.Through the learningand reasoning, the overall efficiency of multi-agent system and structural health monitoring accuracyare improved and the multi-agent health monitoring system is built. Meanwhile the diagnosisaccuracy of the damage identification methods mentioned above is validated and the correctness ofthe system structure model is verified.2) Aiming at the communication among heterogeneous agents of the large aluminum sidingstructure health monitoring system, the communication architecture framework is studied based onKQML and Ontology and the KQML communication primitives is designed with structural healthmonitoring domain knowledge. The active and sub-regional blackboard communication model ofmulti-agent system is designed and it effectively solves information exchange of the practical engineering structure health monitoring system and transmission between heterogeneous agents.3) Considering the inefficiencies of contract net collaboration model for large complex system,the contract net collaboration method based on acquaintance and reasoning is improved, but theagents with limited ability and resources can finish complex health monitoring task through efficientcoordination. The experimental verification combined with three kinds of damage identification iscarried out.4) Due to the diagnosis of time-varying, nonlinear and uncertainty for large complex structurecaused by the external environment, a hybrid multi-agent information fusion architecture model basedon BDIKRI architecture is proposed. The redundant data of BDIKRI data monitoring layer iseffectively reduced and the efficiency of data transmission between agents is improved. And then theconflicts of different diagnosis results are resolved. The experiment for information fusion is carriedout.