考虑传感器故障与环境影响的结构损伤识别方法研究

In present damage detection methods, only the effect of loading and environment on damage is considered for the measured responses. Due to the fact that the sensor is deteriorated and is even in fault with the variation of time and environment, this seriously influences the diagnosis ratio and increases the detection difficulty. On the basis of the principle of hardware redundancy, this project employs the probability and data fusion theory to investigate the following problems. Firstly, the techniques of response novelty and distinguishing them are studied for the responses caused by the environment, sensor fault and damage, and thus the detection and distinguishing methods between the sensor fault and damage are obtained. Secondly, the data processing technique is investigated for a large volume of measured data under the case of environmental effect and sensor deterioration. Finally, a damage detection method is developed by using the swarm intelligent optimization technique, which is applicable to the incomplete measured data under the environmental effect and sensor deterioration. Through this project, some technical bottleneck problems can be solved in the community of structural health monitoring, such as misclassifying the sensor fault and damage,the damage detection method for incomplete measured data under the environmental effect and sensor deterioration. Consequently some independent core techniques are also achieved, which can provide technical support for the large structural health monitoring systems.

现行结构损伤检测方法在分析实测响应时，仅考虑环境、荷载对损伤的影响，而没有考虑传感器由于时间与环境的变化导致其性能退化乃至发生故障，这严重地影响了损伤诊断率也增加了检测的难度。本项目拟采用硬件冗余性的原理，结合概率统计与数据融合理论，首先研究传感器失效、环境、损伤引起的结构响应异常与区分技术研究，获得检测、分离故障传感器与损伤的技术；其次研究考虑传感器性能退化与实测环境干扰的海量数据处理技术；最后，运用群体智能优化技术，研究考虑环境干扰与传感器性能退化下不完备数据的结构损伤检测方法。通过研究，可部分解决传感器故障与损伤混淆、错误报警，和考虑环境影响及传感器性能退化/故障下不完备数据损伤检测方法等健康监测领域的技术瓶径,获得具有自主知识产权的核心技术,为大型结构健康监测系统真正发挥防灾减灾作用提供技术支持。

现行结构损伤检测方法在分析实测响应时仅考虑环境、荷载对损伤的影响，而没考虑传感器由于时间与环境的变化导致其性能退化乃至发生故障，这严重地影响了损伤诊断率也增加了检测难度。本项目在此基础上首先研究了考虑传感器故障的结构响应异常时刻判别和异常来源区分技术，即提出了自适应一致性数据融合技术结合小波分解和独立分量分析技术用于检测含故障传感器信号的响应异常发生时刻；同时提出了敏感指标用于分离故障传感器。其次研究了考虑传感器性能退化的不完备数据的完备化处理技术，即提出了改进多尺度主成分分析（IMSPCA）的数据恢复方法和自适应一致性数据替换方法分别恢复性能退化传感器采集的数据和故障传感器采集的数据。第三，进行了考虑环境干扰与传感器性能退化乃至失效下的线性／非线性结构损伤识别方法研究，提出了传感器采集的加速度信号为适应度函数的结构物理参数辨识以及损伤识别方法，并应用于Bouc-Wen恢复力非线性模型的参数辨识和损伤识别。此外，为了提高识别精度和计算效率，本项目同时提出了基于模态应变能与impsco的两阶段损伤识别方法。实验室试验和数值试验验证了以上方法的有效性与应用范围，部分在实际工程中得到了应用。通过研究，可部分解决传感器性能退化与损伤混淆、错误报警，和考虑环境影响及传感器性能退化下不完备数据损伤检测方法等健康监测领域的瓶颈。出版专著1部，发表核心杂志论文12篇，会议论文5篇，其中SCI收录5篇，EI收录8篇。承办1次国际会议。申请发明专利3项，授权1项，获软件著作权1项。培养博士生2名，硕士生4名，其中仅有1名硕士在读。