标题: 智慧型问题诊断系统之构建—以园区通讯厂商品质系统为例
Intelligent Approach for Diagnosis System:A Case Study
作者: 李建德
Chien-Te Li
苏朝墩
Chao-Ton Su
工业工程与管理学系
关键字: 类神经网路;故障诊断;品质系统;Neural Network;Diagnosis;Quality System
公开日期: 1999
摘要: 以往维修人员在面对诊断问题时,仅能依靠经验累积的方式,来找出故障原因,但由于员工的流动频繁,使得企业面临着资深员工培养不易的情况。近年来由于专家系统的发展已有相当的成果,有不少的研究利用专家系统来辅助人员进行问题诊断,也都有不错的成效。但是在专家系统的建构过程中,系统建构者却往往面临着知识撷取困难及系统维修不易等问题。本研究希望运用类神经网路,发展一套通用型之问题诊断机制。首先对于系统做一个分析与整理,将系统中的问题特征,利用量化的技巧加以量化。接着根据问题的特征与相对应之问题发生原因以及可能解决方法,试着建立类神经网路模式。本研究所建立的模式,可以在系统发生特定问题时,能够找出问题的可能发生原因以及解决的办法,提升制造业系统之绩效,节省不必要的浪费。
本研究以园区通信厂商品质系统为例,发展一套品质系统问题诊断机制,在本案例中,对于制程管制部分以及研发管制部分的诊断正确率皆达到90%左右。研究结果显示,本研究所提方法具有相当的可行性。
In the past, when mechanics diagnose the breakdown situations, they only can rely on accumulated experiences and find out the causes. Recently, employee come and go frequently. Enterprises faces the problem of keeping experienced employee. In recent years, due to the development of expert system is quite fruited; in many researches, mechanics diagnose breakdown situations with the assistance of expert system and the outcome is pleasant . However, during the process of building expert system , system builders usually face the problems such as knowledge gathering and system repair. Our institute hopes to apply neural network to build a general model of diagnostic mechanism. First of all, we analyze and gather the system and then quantify the characteristics of the system. Finally, we try to build neural network model via the characteristics of the problems, causes of relative problems and possible resolvable ways. When specific problems occurs, our model can find out possible reasons and resolvable ways , improve the effect of manufacture system and save unnecessary waste.
Our institute uses the quality systems of communication manufactures in science zone as sample and tries to build a quality system diagnostic mechanism. In our case, the diagnostic of making process control and invention control reaches 90 %.Our research reveals that the method we mentioned is quite feasible.
英文摘要…………………………………………………………ii
志谢………………………………………………………………iii
目录………………………………………………………………iv
图目录……………………………………………………………vi
表目录……………………………………………………………vii
第一章 绪论………………………………………………………1
1.1 研究背景与动机………………………………………1
1.2 研究目的………………………………………………2
1.3 研究范围与限制………………………………………2
1.4 研究架构与步骤………………………………………3
第二章 文献探讨…………………………………………………5
2.1 类神经网路……………………………………………5
2.1.1 倒传递网路理论………………………………………7
2.1.2 学习向量量化网路…….……………………………9
2.1.3 机率神经网路……………....………………………10
2.1.4 放射基准机能网路………….………………………11
2.2 类神经网路于问题诊断与分类之应用………………11
第三章 问题诊断系统构建流程…………………………………16
第四章 实例说明…………………………………………………21
4.1 个案说明………………………………………………21
4.2 诊断系统构建…………………………………………22
4.3 比较……………………………………………………37
第五章 结论与建议………………………………………………39
参考文献……………………………………………………………41
附录…………………………………………………………………44
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT880031028
http://hdl.handle.net/11536/65185
显示于类别:Thesis