基于模糊kohonen聚类算法的桥梁健康监测数据挖掘模型的建立

发布时间：2018-01-15 23:05

本文关键词：基于模糊kohonen聚类算法的桥梁健康监测数据挖掘模型的建立　出处：《哈尔滨工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：近年来，我国基础设施建设发展迅速，桥梁结构作为交通运输的重要基础设施，安全问题备受人们的关注。为保障桥梁的安全运营，桥梁结构健康监测系统得到了广泛的应用。在桥梁健康监测系统中，包括数量众多的传感器，这些传感器持续、长期、实时的采集各类数据。随着监测时间的不断增长，形成海量数据。本文通过研究基于kohonen算法的数据挖掘方法，对海量监测数据进行分析计算，为桥梁的后期评估以及预警提供基础数据。主要研究内容如下：第一，以三个随机数组为例对算法进行试验，通过三种数组的聚类结果分析算法中存在的不足。针对算法收敛速度和计算准确性两方面的不足提出了算法的改进，通过三个例子对比算法改进前后的收敛速度以及聚类准确性，证明算法的改进具有很好的效果。第二，利用有限元分析软件迈达斯，根据甬江特大桥的设计图纸建立大桥的初始有限元模型，并对大桥进行了初步的分析。分析桥梁初始有限元模型中可能存在的三种误差，确定产生每一种误差的原因。为了获得结构的基准有限元模型，减小各方面的误差，对结构的初始有限元模型进行修正。以结构的动力特性作为目标量对模型进行修正，修正后达到了很好的效果。利用修正后的结构基准有限元模型，模拟桥梁试验荷载，对桥梁存在的静力误差进行验证，，通过结果可以看出，所得的结构基准有限元模型的静力方面误差要小于未进行修正的误差。第三，对健康监测原始数据进行预处理，进行滤波处理。结合kohonen算法与结构的基准有限元模型建立桥梁健康监测的数据挖掘聚类分析模型，确定识别异常数据的异常阀值。最后对得到的聚类模型进行验证，证明聚类分析模型的有效性。
[Abstract]:In recent years, the construction of infrastructure in China has developed rapidly. Bridge structure, as an important infrastructure for transportation, has attracted people's attention in order to ensure the safe operation of bridges. Bridge structural health monitoring system has been widely used. In the bridge health monitoring system, including a large number of sensors, these sensors continue, long-term. Collect all kinds of data in real time. With the continuous growth of monitoring time, the formation of massive data. This paper studies the method of data mining based on kohonen algorithm to analyze and calculate the massive monitoring data. To provide basic data for post-assessment and early warning of bridges. The main contents of this study are as follows: First, take three random arrays as an example to test the algorithm. Through the analysis of three kinds of array clustering results, the shortcomings of the algorithm, aiming at the convergence speed and computational accuracy of the algorithm two aspects of the improvement of the algorithm is put forward. By comparing the convergence rate and clustering accuracy of the improved algorithm with three examples, it is proved that the improved algorithm has a good effect. Secondly, the initial finite element model of the bridge is established according to the design drawings of Yongjiang Bridge by using the finite element analysis software Midas. Three possible errors in the initial finite element model of the bridge are analyzed, and the causes of each error are determined. In order to obtain the benchmark finite element model of the structure, the bridge is analyzed preliminarily. The initial finite element model of the structure is modified by reducing the errors in all aspects. The dynamic characteristics of the structure are taken as the target to modify the model. The modified finite element model is used to simulate the experimental load of the bridge, and the static error of the bridge is verified, which can be seen from the results. The static error of the structural benchmark finite element model is smaller than that of uncorrected finite element model. Thirdly, preprocessing the original data of health monitoring and filtering. Combining kohonen algorithm with the benchmark finite element model of structure, the data mining cluster analysis model of bridge health monitoring is established. Finally, the clustering model is verified to prove the validity of the clustering analysis model.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U446

【参考文献】