一种基于神经网络的免疫识别故障检测模型  

Immune Recognition Fault Detection Model Based on Neural Networks

作　　者：侯胜利[1] 毕宏[2] 毕志蓉[1] 王威[1] 

机构地区：[1]徐州空军学院,徐州221006 [2]徐州医学院江苏省麻醉重点实验室,徐州221002

出　　处：《系统仿真学报》2009年第7期1887-1890,1896,共5页Journal of System Simulation

摘　　要：提出了一种基于神经网络的免疫识别故障检测模型。该模型根据免疫识别原理来构造神经网络检测器,通过训练将被检测对象的故障模式信息存储于分布的检测器中,检测器用于捕获被检测对象的异常模式特征,当检测器与特征样本匹配时则激活该检测器,根据检测器的激活情况来发现故障,并给出了相应的训练算法。通过滚动轴承损伤检测的仿真实验,表明该方法对由轴承损伤冲击造成的信号突变保持了较高的灵敏度和分辨率,对于滚动轴承的监测具有一定的应用价值,并可方便地推广到其他类似的工业应用领域。A new fault detection model of immune recognition was proposed based on neural networks. In this model, neural networks-based detectors were constructed based on the principle of immune recognition. Taking advantage of neural networks training, the information of fault patterns was stored in the distributed neural networks-based detectors. These detectors were used to capture the anomalous pattern features. When a detector was matched up with a feature sample, the detector was stimulated. Through the relevant stimulated detectors, a fault could be found out. The principle and structure of the model were proposed, and its training algorithm was derived. The fault diagnosis scheme using this model was investigated. Defect detection simulations of rolling bearings shows that the proposed model has high resolution for locations of signal singularities, which are caused by impacts between defect and its mating surface in the bearing. This method is contributive for rolling bearing status monitoring and it can be easily extended to other relative industrial application areas.

关 键 词：故障检测 人工免疫系统 免疫识别 神经网络 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP277[自动化与计算机技术—控制科学与工程]