乏信息下混合不确定性损伤识别的模型修正方法及云模型理论研究

Many uncertainties in damage identification of the engineering structure are often co-existing and interacting with each other. It is difficult to describe and quantify this uncertainty reasonably, especially in the case of the limited test information, which greatly increase the difficulty in structural damage. For this reason, the analytical methods for damage detection considering the mixed uncertainty and limited test data based on modal interval analysis, statistical method and cloud model theory is proposed in this project. Firstly, the uncertainty quantization method of modal parameters based on bootstrap method and the modeling error quantization method based on grey theory and interval mathematics are studied, respectively. Subsequently, this project also studies the effect mechanism of these two kinds of uncertain factors and their coupling on structural damage identification index. According to the researches mentioned above, the damage identification methods based on modal interval-random model correction and cloud model theory are proposed, to form more perfect mechanism and method for solving the disturbances mixed different uncertainties in damage identification. Finally, the feasibility and reliability of the methods are verified by numerical simulation, model test and real bridge test. The research results can solve the uncertain disturbance problem of the actual diagnosis for structural damage. Also, it can provides the basis for the study of structure disaster mechanism, structure health monitoring and safety evaluation.

工程结构中存在的诸多不确定性因素往往具有各自特点且相互影响，在测试信息缺乏的情况下通常难以准确描述与量化，极大增加结构损伤识别难度。为此，本项目基于模态区间分析、概率统计及云模型等理论，对乏信息下混合不确定性损伤识别进行深入研究。首先，分别提出基于自助法的模态参数不确定性量化方法和基于灰理论和区间数学的建模误差量化方法，对测试噪声和建模误差进行量化；其次，探究这两种不确定性因素及其耦合作用对结构损伤识别指标的影响机制；在前述研究基础上，分别提出基于模态区间-随机模型修正和基于云模型理论的损伤识别新方法，形成较完善的解决损伤识别问题中混合不确定性干扰的有效机制及方法。最后，基于数值模拟、模型试验和实桥测试结果验证所提方法的可行性及可靠性。研究成果可在一定程度上解决实际结构损伤识别的不确定性干扰问题，为重大工程结构灾变机理、健康监测和安全评估研究提供科学依据，具有较大的理论意义和工程应用价值。

工程结构中存在的诸多不确定性因素往往具有各自特点且相互影响，在测试信息缺乏的情况下通常难以准确描述与量化，极大增加结构损伤识别难度。为此，本项目首先研究各类型不确定性来源的合理描述方式。针对模态参数识别结果的不确定性量化问题，提出来基于Block-bootstrap和多层次聚类的模态参数自动识别及不确定性量化算法，量化模态自动识别过程存在随机抽样误差和测试误差；同时分别提出了基于云模型理论、改进自助算法和信息扩散理论的结构不确定性物理参数量化及传递分析方法，探究参数不确定性对结构响应的影响机制。在前述研究基础上，对考虑不确定性因素影响的结构损伤识别理论进行了研究，基于云模型理相似性度量，分别提出三种不确定性损伤识别方法。此外，为了进一步提高损伤识别的精度和自动化程度，本项目还提出了两种基于深度学习理论的结构损伤识别方法。实验室试验和数值模拟试验验证了上述方法的有效性及应用范围。研究成果可在一定程度上解决实际结构损伤识别的不确定性干扰问题，为重大工程结构灾变机理、健康监测和安全评估研究提供科学依据，具有较大的理论意义和工程应用价值。