连通管式光电液位挠度传感器故障时间定位与数据重构

发布时间：2018-03-29 18:17

本文选题：故障时间定位　切入点：卡尔曼滤波　出处：《重庆理工大学》2015年硕士论文

【摘要】：桥梁结构安全问题一直都被学术界和工程界高度重视,系统故障与传感器故障的区分成为了近些年来研究热点。传统的传感器故障诊断多是针对系统中的传感器做故障空间定位,然而对于桥梁大型监测系统,传感器故障发生到故障解决往往需要相当长的一段时间,为保证系统在此期间的正常运作,对传感器做故障时间定位以及故障数据重构显得意义深远。本文主要针对桥梁结构健康监测系统中的光电液位挠度传感器,完成以下几项研究:(1)卡尔曼滤波是目前应用最广泛的最优估计理论,本文将卡尔曼滤波应用到数据预处理,有效的抑制了数据中的噪声误差,分析表明,滤波之后的数据相关性增强,这给后续研究高效精准的故障定位方法提供了较好的支持。同时卡尔曼滤波是单步递推估计,这也为实时在线的高精度故障诊断提供了可能。(2)提出了基于滑动时间窗相关性分析的故障时间定位方法。方法依据组内传感器之间较强的相关性,采用改进的相关度模型,基于滑动时间窗做相关性分析,以相关度量化值对故障进行判定,从而对故障进行时间定位分析。提出了基于数据标准化残差分析的故障时间定位方法。方法依据组内传感器之间只存在幅值上的显著差异,变化趋势高度一致,基于数据标准化作残差分析,以残差偏离量化值对故障进行判定,从而对故障进行时间定位。运用两种时间定位方法,分别对工程中常见的四种故障类型做仿真模拟,分析两种定位方法的有效性与精确度。实验验证,相关法对精度下降故障表现出明显的优势,而残差法对常值故障、固定偏差、漂移故障的定位性能均优于相关法。两种方法结合使用,可达到更好的故障时间定位效果。(3)提出了自适应残差法的故障数据重构方法。方法依据传感器数据之间保持一致的变化趋势,基于标准化的残差具有趋0性,以残差值最小为目标估计故障传感器数据,实现故障数据重构。与经典的数据重构方法——RBF神经网络、多元回归分析、最小二乘法,对重构效果作对比分析,并对重构残差作量化对比分析。实验验证,针对本研究,本文所提的自适应残差的重构效果是最优的。
[Abstract]:The safety of bridge structure has always been attached great importance by the academic and engineering circles. The distinction between system fault and sensor fault has become a hot topic in recent years. The traditional sensor fault diagnosis is mostly aimed at the sensor in the system fault space location, but for the bridge large-scale monitoring system, Sensor failures often take quite a long time to resolve, and in order to ensure the normal operation of the system during this period, It is of great significance to locate the fault time and reconstruct the fault data for the sensor. This paper mainly focuses on the photoelectric liquid level deflection sensor in the bridge structure health monitoring system. Kalman filtering is the most widely used optimal estimation theory at present. In this paper, Kalman filter is applied to data preprocessing, which can effectively suppress the noise error in the data. The correlation of the data after filtering is enhanced, which provides a good support for the subsequent research on the efficient and accurate fault location method. At the same time, the Kalman filter is a single step recursive estimation. It also provides the possibility for real-time on-line high precision fault diagnosis. A fault time location method based on sliding time window correlation analysis is proposed. According to the strong correlation between sensors in the group, an improved correlation model is adopted. Based on the sliding time window, the correlation analysis is done, and the quantitative value of correlation degree is used to judge the fault. A fault time location method based on data standardized residuals analysis is put forward. The method is based on the significant difference in amplitude between sensors in the group, and the variation trend is highly consistent. Based on the data standardization, the residual error is analyzed, and the fault is determined by the quantization value of the residual deviation, and then the fault is located in time. The four common fault types in engineering are simulated by using two time localization methods, respectively. The effectiveness and accuracy of the two localization methods are analyzed. The experimental results show that the correlation method has obvious advantages on the precision decline fault, while the residual method has fixed deviation for the constant fault. The performance of drift fault location is better than that of correlation method. A fault data reconstruction method based on adaptive residuals method is proposed. The method is based on the uniform change trend of sensor data, and the standardized residual error has a tendency to zero. Taking the minimum residual value as the target to estimate the fault sensor data, the reconstruction of the fault data is realized, which is compared with the classical data reconstruction methods, such as RBF neural network, multivariate regression analysis and least square method. The experimental results show that the adaptive residuals proposed in this paper are the best.
【学位授予单位】：重庆理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U446

【引证文献】