Development of a damage detection system for structures using strain memorizing sensors

使用应变记忆传感器开发结构损伤检测系统

基本信息

批准号：
12555135
负责人：
SATO Tadanobu
金额：
$ 5.89万
依托单位：
KYOTO UNIVERSITY
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (B)
财政年份：
2000
资助国家：
日本
起止时间：
2000 至 2002
项目状态：
已结题

项目摘要

Based on a recent review of system identification, we developed several promising methodologies corresponding to the system identification such as an adaptive H infinity and Kalman filter, a neural network with forgetting capability of network structure and an adaptive Monte Carlo filter. Those algorithms possess efficient time-marching capabilities to identify non-linear and non-stationary structural systems. To improve convergence of non-stationary change of structural parameters we also developed a hybrid algorithm combining the Kalman filter with the Monte Carlo filter and the Genetic algorithm was also introduced into the Monte Cairo filter. Both parametric and non-parametric system identification technologies were also developed suitable to deal with highly nonlinear and time varying systems. Those algorithms were implemented and executed on a high speed laptop computer synchronized with input data through AD converters from an observation system, a structural identification system, which enables a real time identification of non-stationary change of structural parameters.Simple and inexpensive passive sensors that can monitor the peak strain or displacement were developed. The memorized data can be wirelessly retrieved by introducing an LC circuit into the sensor enabled wireless reading of the data. A prototype sensor was fabricated and tested.A portable equipment to be used for structural system identifications was developed by combining a wireless data acquisition system with the structural identification system. Efficiency of the developed system was demonstrated by micro-tremor observations of a three stories frame structure.

基于最近对系统辨识的回顾，我们开发了几种与系统辨识相对应的有前途的方法，例如自适应 H 无穷大和卡尔曼滤波器、具有网络结构遗忘能力的神经网络和自适应蒙特卡罗滤波器。这些算法具有有效的时间推进能力来识别非线性和非平稳结构系统。为了提高结构参数非平稳变化的收敛性，我们还开发了卡尔曼滤波器与蒙特卡罗滤波器相结合的混合算法，并将遗传算法引入了蒙特开罗滤波器。参数和非参数系统识别技术也被开发出来，适合处理高度非线性和时变系统。这些算法在高速笔记本电脑上实现和执行，并通过来自观测系统（结构识别系统）的 AD 转换器与输入数据同步，从而能够实时识别结构参数的非平稳变化。简单且廉价的无源传感器可以监测峰值应变或位移的开发。通过将 LC 电路引入传感器，可以无线检索存储的数据，从而实现数据的无线读取。制作并测试了原型传感器。将无线数据采集系统与结构识别系统相结合，开发了一种用于结构系统识别的便携式设备。通过对三层框架结构的微震观测证明了所开发系统的效率。