【作者】 谭大鹏； 李具中； 李培玉； 潘晓弘；

【Author】 TAN Da-peng1, LI Jv-zhong2, LI Pei-yu1, PAN Xiao-hong1 (1. Mechanical and Energy College, Zhejiang University, Hangzhou 310027, China 2. Second Steel Factory, Wuhan Iron and Steel Corporation, Wuhan 430080, China)

【机构】 浙江大学机械与能源工程学院； 武汉钢铁集团公司第二炼钢厂；

【摘要】 嵌入式故障诊断技术要求针对性强、占用空间小、可配置、实时性好,但是传统的故障诊断技术需要大量的数据资源作为支持,不能满足以上要求。针对此问题,在对当前故障诊断技术的广泛调研的基础上,提出了一种基于C语言集成产品系统(CLIPS)与人工神经网络技术(ANN)技术的嵌入式故障诊断专家系统。对故障诊断专家系统及其相关技术进行了深入研究,在Linux环境下搭建了专家系统开发环境以及相应的设计工具链;基于构件化设计思想,使系统具备较好的可配置性与可扩展性,同时可以对数据处理构件以及图形用户界面构件进行裁减,以满足嵌入式系统的基本运行要求;利用CLIPS提供的应用程序接口设计实现了相对精简高效的规则库与知识库,并通过ANN信息融合算法完成判断与推理。工业应用系统实例证明:其在ARM-Linux嵌入式环境下运行稳定,上下文切换迅速,能够对一些常见故障进行有效识别。更多还原

【Abstract】 Embedded fault diagnosis technology requires high pertinency, small storage space, good configuration ability and real-time performance. Traditional fault diagnosis needs large quantity data source support, that can not satisfy the demand above. Aiming at this problem, a kind of embedded fault diagnosis expert system (E-FDES) based on the C language integration product system (CLIPS) and artificial neural network (ANN) technology was brought forward by wide investigation for current fault diagnosis technology. FDES and its relative technology was discussed, the developing environment and design tool chain for E-FDES was established in Linux environment. Using modularization theory principle, system re-configuration and expansion ability were ensured, and the data process components and graphic user interface components could be reduced according to special demand to meet the embedded system running requirement in the same time. By the application program interface of CLIPS, knowledge base and rule base were realize, that decreased the developing time of system inference engine and reduced source occupation space effectively. Industry application instance proves that this system runs stably in ARM-Linux embedded environment, switch smoothly and can recognize some common faults for the devices monitored.更多还原