基于dempster-shafer证据理论的结构损伤识别方法结构工程专业论文.docxVIP

哈尔滨 哈尔滨工业大学工学硕士学位论文 Abst Abstract - - PAGE IV- - - PAGE III - Abstract Structural health monitoring is becoming an important tool to study the performance and damage evolution of infrastructure in service. Damage detection is one of key issues in study on health monitoring. In the past decades, many damage detection methods have been developed by researchers. However, ambient noise often decreases the accuracy of damage detection. The methods including wavelet transform, artifical neural networks, data mining and information fusion developed in recent years provide the effective potential to improve the accuracy of structural damage detection. This thesis mainly focuses on developing damage detection methods based on information fusion technique and consists of three parts as follows: The characteristics of some typical damage identification methods are analyzed. The methods include frequency response function method, mode shape method, ? index method and wavelet transform method. Numerical study on a beam with damage is carried out. The measurement locations, damage extent and noise level are selected as factors to research the characteristics of damage detection methods. For measurement location, 8 measure points and 12 measure points in a distribution of identical interval distance are adopted. Additionally, noise level and damage extent are 0%, 2%, 5%, 10%, 15%, 20% and 10%, 20%, 40%, 60% in this numerical study, respectively. Considering the damage evolution and variety of noise at different period of collection time, a damage detection method based on information fusion technique is proposed, referred to multi-period damage detection method based on information fusion technology. Firstly, the collected data is divided into several samples. Damage detection is individually conducted on each sample. After that, data infusion is applied to the damage detection results obtained from the above samples. Numerical studies on a 6m b