摘要
为了减少对知识化制造系统生产故障的误诊和漏诊,本文首先提出了一种新的基于多维特征提取的故障监测方法,该方法先对采集的信号进行多维特征提取,再通过归纳学习建立设备的正常状态空间,并以此来判断设备的故障状态。然后,提出一种故障误判概率控制方法。通过多维特征提取和误判概率控制,可以很好地减少对故障的漏判和误判。仿真实验证实了该故障监测方法的有效性。
In order to minimize the probability of false diagnosis in knowledgeable manufacturing systems, a new approach based on inductive learning to fault detection is proposed in this paper. Firstly, multi-dimension features are extracted from sampled signals and compose a normal state space of equipment by inductive learning, according to which an abnormal signal can be detected. A method of controlling the probability of false diagnosis is then proposed to greatly decrease it.The effectiveness of the approach is verified by simulation examples.
出处
《制造业自动化》
北大核心
2006年第2期33-38,共6页
Manufacturing Automation
基金
国家自然科学基金资助项目(60574062
50475075)
江苏省高校"青蓝工程"第二期省级中青年学术带头人培养计划
关键词
知识化制造
归纳学习
多维特征提取
knowledgeable manufacturing
inductive learning
multi-dimension feature extraction
