【作者】 张伟刚；

【导师】 谭建豪；

【作者基本信息】 湖南大学 ， 模式识别与智能系统， 2007， 硕士

【摘要】 生物免疫系统是一种高度并行的自适应信息学习系统，它能自适应地识别和排除侵入机体的抗原性异物，并且具有学习、记忆和自适应调节能力，能够维护机体内环境的稳定。近年来，人们逐渐意识到生物免疫机制对开发新的计算智能的启示意义，人工免疫算法是受生物免疫系统启示而设计的新型算法，已经用于机器学习、异常和故障诊断、机器人行为仿真、机器人控制、网络入侵检测等领域，表现出较卓越的性能和效率，它已经成为计算智能大家庭的卓有成效的新成员。本文系统的评述了人工免疫系统的国内外研究现状和发展前景，对当前的故障诊断技术展开了全面而又深入的分析；同时详细分析了生物疫系统的基本工作原理，并总结出了发展人工免疫系统可以借鉴的主要机理。为了实现在对大量冗余信息的约减过程中，既能保证具有较高的数据压缩率，同时又要保持原有数据结构的完整性，提出了基于人工免疫网络模型的数据约减算法。为故障诊断中的数据约减和聚类分析提供了一种有效可行的方法。本文在分析Forrest的否定选择算法优缺点的基础上，结合设备异常检测的实际需要，提出了改进该算法的检测器生成算法，使之不但能够检测出系统或设备已知模式的故障，还能够检测出未知模式的异常。通过把这个方法应用于变压器故障检测中，取得了很好的效果，证明了这个方法的有效性。在深入学习借鉴免疫系统的克隆变异机理及已有的人工免疫系统成果，提出了具有对故障样本的连续学习功能的自适应故障诊断方法。实现了故障诊断系统在对新的故障样本学习的过程中，保持对已学知识的记忆。通过对标准样本的分类识别验证了该方法的有效性。更多还原

【Abstract】 Biological immune system is a highly parallel adaptive information learning system, which can identify and remove the antigenic eye winkers in vading the body. This system can learn, remember and adjust adaptively to keep the stabilization inside The body. During recent years, people begin to realize the revelatory significance of the biological immune mechanism to intelligent algorithm. Artificial immune algorithm(AIA) is this kind of new algorithm which is inspired by the biological immune system. This kind of algorithm has been used in many fields, such as machinery study, unconventionality and malfunction diagnosis, simulation of the behavior of robots, control of robots, inbreak detection of networks and etc. It has been a new effective member of the family of intelligent algorithms.The recent research results of AIS are systematically overviewed in this paper and the directions for further study are also provided. This paper also analyses the present fault diagnosis techniques deeply. At the same time, the basic working mechanism of the biological immune system are also analysed in detail, and this paper summarize the main mechanism that the AIS can draw lessons from.During reducing redundant information,to realize higher data compression rate and maintain better data architecture, a new data reduction algorithm is proposed based on the artificial immune network model. As a result, a new effective data reduction and clustering approach is provided for the fault diagnosis of equipment.On the basis of the advantages and disadvantages of the Forrest’s algorithm,this paper study the negatives election algorithm and an improved algorithm for generating the detector set is proposed combining the actual needs of the abnormal detection. Through applying this method to transformer fault diagnosis, this paper has made very good result and has proved the validity of this method.Inspired by the clone and mutation mechanism of immune system and research achievements of artificial immune system, a new self-adaptable fault diagnosis approach with continuous learning is investigated. During learning new knowledge the approach can maintain memory for learned knowledge. The validity of the approach is demonstrated by recognizing standard sample.更多还原