基于人工免疫系统的结构故障检测与分类

发布时间：2018-03-02 19:17

本文选题：结构健康监测　切入点：人工免疫　出处：《沈阳建筑大学》2015年硕士论文　论文类型：学位论文

【摘要】：重大土木工程结构基础设施是确保国民经济长期、稳定、可持续发展地物质基础,它们地健康状况和安全性评价更是人民生命和财产安全得重要保障。对一些大型结构进行健康监测与安全评估,及时发现安全隐患,建立安全预警系统是近年来国内外学术研究的热点问题之一。结构健康监测系统通过集成结构中的先进传感元件,实时在线地获取与结构健康状况有关的信息,并与信号信息处理方法以及材料结构力学建模方法相结合,将结构损伤特征参数提取出来,进而识别结构的状态,这样在结构产生不安全因素的早期,就可以对结构体加以控制,从而消除安全隐患或对安全隐患进一步发展加以控制。结构故障检测与分类是结构健康监测中的重要问题,但目前研究成果较少,本文引入人工免疫系统解决结构故障检测与分类问题。主要工作如下：(1)生物免疫系统具有极强的信息处理能力和许多优良特性,非常适合用于异常检测和分类。分析人工免疫系统的生物原型、仿生机理、特征,针对结构健康监测系统自主有效的实时监测建筑物状态的问题,利用人工免疫系统的自治性、主动性、自适应及学习和记忆的仿生机理,基于低能耗、分布式检测的无线传感器网络平台,研究结构健康监测中对结构故障识别和分类的问题,并详细讨论了生物免疫系统与基于人工免疫系统的结构健康监测传感器网络的对应关系。(2)基于对有监督的大型结构健康监测中的故障检测和分类问题进行研究,提出一种基于粒子群优化变异的克隆选择算法。该算法将样本结构模式数据作为抗原刺激抗体集合,抗体集合经过克隆、变异、选择等学习和进化过程来提高记忆细胞质量,以实现对实测数据的故障检测与分类。特别是针对克隆选择算法二进制编码复杂和变异方向不确定的问题,引入了粒子群优化变异。在Benchmark结构模型上的仿真实验结果表明该算法有效地识别故障模式,且提高了结构故障分类的成功率。(3)结构健康监测环境空间跨度大、结构环境复杂和需要长期稳定运行,针对无监督的结构故障检测与分类问题,提出了基于AiNet免疫聚类的结构故障检测与分类算法。首先利用AiNet免疫聚类算法对记忆细胞集合进行更新,得到聚类中心；然后利用所得聚类中心实现结构故障检测与分类。将该方法用于IASC-ASCE SHM任务组提出的Benchmark模型故障检测与分类中,讨论了抑制阈值对聚类结果的影响,计算了分类成功率,验证了该算法的有效性验证了该方法的有效性。
[Abstract]:A major civil engineering infrastructure is to ensure that the national economy long-term, stable material basis of sustainable development, health and safety evaluation of them is an important guarantee of people's life and property safety, health monitoring and safety evaluation of some large structures, timely detection of security risks, establish security early warning system is one of the hot issues in academic study at home and abroad in recent years. The structural health monitoring system through advanced sensor integrated structure, real-time online access related information of the health condition, and combined with the information signal processing method and structure mechanics modeling method, the structure damage characteristic parameter is extracted, and then identify the structural state, resulting in early the unsafe factors in the structure, it can be controlled on the structure, so as to eliminate the hidden trouble of safety or security risks further The development of control. The detection and classification of fault structure is an important problem in structural health monitoring, but there is less research results, this paper introduces the artificial immune system to solve the problem of fault detection and classification structure. The main work is as follows: (1) the biological immune system has strong ability of information processing and many excellent characteristics, very suitable for anomaly detection the classification and analysis. Biological prototype, artificial immune system mechanism, characteristics, for real-time monitoring of building independent and effective structural health monitoring system, the use of artificial immune system, autonomy, initiative, bionic mechanism and adaptive learning and memory, based on low energy consumption, wireless sensor network platform for distributed detection. The structure of fault identification and classification of structural health monitoring, and discussed in detail with the biological immune system based on artificial immune system The relationship between sensor network for structural health monitoring. (2) to study the problem and fault classification of large structural health monitoring supervised in detection based on particle swarm optimization presents a clonal selection algorithm based on variation. The algorithm will sample structure mode data as a set of antigen stimulating antibodies, antibody set through cloning, mutation and the choice of learning and evolution process to improve the quality of memory cells, to realize fault detection and classification of data. Especially for the clonal selection algorithm for binary encoding complexity and variability uncertain problems, introduced particle swarm optimization. The variation in the Benchmark simulation model of the result shows that the algorithm can effectively identify the failure mode, and improve the structure of fault classification success rate. (3) the environment of structural health monitoring of large span, complex structure and the need for long-term stable operation of the environment OK, for the structure and classification of fault unsupervised detection algorithm is proposed AiNet structure fault detection and classification based on immune clustering. Firstly AiNet immune clustering algorithm to update the memory cells set by clustering center; then use the clustering center to realize structure fault detection and classification. This method is used in Benchmark fault model detection and classification of the IASC-ASCE SHM task group is put forward in this paper, discussed the suppression threshold effect on the results of clustering, classification success rate was calculated to verify the effectiveness of the algorithm to verify the effectiveness of the method.

【学位授予单位】：沈阳建筑大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TU317

【相似文献】