【作者】 邵龙秋；

【导师】 李铁鹰； 张清华；

【作者基本信息】 太原理工大学 ， 控制理论与控制工程， 2009， 硕士

【摘要】 生物免疫系统是一个具有模式识别、分布式检测、记忆能力、自学习能力、多样性等特征的系统。借鉴生物免疫系统,以生物免疫系统具有的诸多特性为基,解决实际工程与应用中的问题,便诞生了人工免疫系统这一极具生命力的新兴学科。由免疫系统自己-非己识别机理衍生而来的阴性选择算法的提出为故障诊断领域提供了新思路和新方法。在阴性选择算法中采用无量纲指标进行旋转机械的故障诊断是一种非常有效的诊断方法。但是到目前为止,可供使用的无量纲指标数目有限,应用在多重并发故障诊断的无量纲指标更是少之又少。近年来,借鉴仿生学的思想,基于生物体系的生物进化、细胞免疫、神经细胞网络等的某些机制,利用抽象描述的数学语言模仿生物体系和人类的智能机制,产生了计算智能。遗传编程是计算智能理论中一种崭新的、重要的方法。遗传编程将遗传算法和计算机程序有机结合,具有在候选解表示上的动态结构特点和搜索寻优机制上全局性的特点,在计算智能中的应用越来越广泛。本文利用遗传编程构建新的无量纲指标,把遗传编程作为一种智能的层式结构优化算法,以现有的无量纲指标为初始参数,通过对原始参数的重新组合和优化,形成新的复合参数。把分类效果作为判断参数优劣程度的准则,获得了具有最佳识别能力的优化指标。仿真结果表明,优化指标对于并发故障有很好的诊断能力。针对本文在无量纲指标免疫检测器的生成过程中因进行约简、聚类等分类处理而丢失部分有用故障特征信息的不足,采用了利用各无量纲指标免疫检测器进行集成诊断的弥补方法。仿真结果表明,该方法具有较高的诊断准确率。以茂名乙烯厂橡胶装置二线挤压脱水机GY6204及膨胀干燥机GY6205为工业应用背景,设计了“橡胶装置二线挤压脱水机GY6204及膨胀干燥机GY6205智能故障诊断系统”。更多还原

【Abstract】 Biology immune system is a system with the ability of pattern recognition, distributed detection, memory capacity, self-learning, diversity, etc. Artificial immune system is a new discipline with strong vitality based on the features of biology immune system to solve the problems encountered in practical application and project. The negative selection algorithm which derived from the mechanism of recognizing self and non-self of immune system provides the new idea and method for the field of fault diagnosis.Applying the non-dimensional parameter to diagnose the faults of rotating machinery in the negative selection algorithm is very effective. But so far, the number of usable non-dimensional parameter is finite, particularly in the application of multiple concurrent faults diagnosis. In recent years, computational intelligence has produced by using abstract mathematics language to imitate biological systems and human intelligence mechanism based on the mechanism of biological evolution, cellular immunity, neuron network, etc of biological systems. The genetic programming is a brand-new and important method of computational intelligence theory. The genetic programming is the organic combination of both the genetic algorithm and computer programs. The genetic programming has the characteristics of dynamic structure on the denotation of candidate solutions and overall significance on searching optimization mechanism. The application of the genetic programming in computational intelligence is becoming more and more widespread. In this paper, the new non-dimensional parameters were constructed using the genetic programming as the intelligent layered structure optimization algorithm to reset and optimize the existing non-dimensional parameters as the original parameters. The optimal parameters which have best discernment were acquired by taking the classification effect as the rule of judging a parameter. The simulation results show that the optimal parameters have very good diagnosis ability for multiple concurrent faults.Aiming at the shortage of losing partial useful failure feature information because of the data processing of reduction and clustering in the process of non-dimensional parameter immune detector generation, this paper adopt a recuperate method of using every kinds of non-dimensional parameter immune detector to the integrated diagnosis. The simulation results show that this method has high veracity ratio in multiple concurrent faults diagnosis.Regarding second-line extrusion dewatering machine GY6204 and expansion dryer GY6205 in rubber device of Maoming ethylene plant as the background of industry application, this paper has designed the“extrusion dewatering machine GY6204 and expansion dryer GY6205 intelligent fault diagnosis system”.更多还原