本文選題：決策樹　切入點：聚類分析　出處：《山東大學》2014年碩士論文

【摘要】：產(chǎn)品質(zhì)量是對顧客需求的具體反映,也是顧客滿意的必要因素。為了能在競爭激烈的市場當中占據(jù)有利地位,提高企業(yè)競爭力,現(xiàn)代企業(yè)必須加強質(zhì)量管理。隨著科學技術的不斷發(fā)展與市場的日趨成熟,市場對質(zhì)量的要求不斷提高,現(xiàn)有質(zhì)量管理技術與工具已不能滿足需求。 制造執(zhí)行系統(tǒng)(Manufacturing Execution System,MES)環(huán)境下生產(chǎn)過程中的質(zhì)量數(shù)據(jù)是由車間作業(yè)現(xiàn)場控制收集,系統(tǒng)與工程師進行分析處理并儲存在指定的數(shù)據(jù)庫或數(shù)據(jù)倉庫中。其數(shù)據(jù)內(nèi)容包含加工記錄、實時監(jiān)控信息等,這些數(shù)據(jù)的累積導致分析人員對數(shù)據(jù)的處理能力下降,其潛在價值沒有被進一步挖掘而產(chǎn)生信息的浪費。為了充分利用生產(chǎn)過程中收集的大量數(shù)據(jù),實現(xiàn)全面質(zhì)量管理,應用數(shù)據(jù)挖掘技術(Data Mining,DM)實現(xiàn)智能化加工過程質(zhì)量診斷控制的技術和系統(tǒng)成為眾多專家學者及企業(yè)新的研究熱點。本文在總結(jié)前人研究基礎之上構建MES環(huán)境下制造過程質(zhì)量數(shù)據(jù)挖掘平臺,分析加工質(zhì)量相關數(shù)據(jù),研究具有控制圖模式識別、質(zhì)量預測及診斷功能的質(zhì)量控制系統(tǒng),主要研究內(nèi)容包括以下三方面： (1)針對制造過程數(shù)據(jù)特點,提出一種適用于過程質(zhì)量數(shù)據(jù)分析的不純性度量：Fβ度量類置信度比Fβ-Confidence Proportion,FCP),并建立基于FCP不純性度量的決策樹(Decision Tree, DT)。在構建分類器的基礎上,構建基于FCP決策樹的控制圖模式識別系統(tǒng),針對控制圖數(shù)據(jù)維度較高的現(xiàn)象引用統(tǒng)計量作為控制圖模式識別的統(tǒng)計特征。經(jīng)測試表明該模式識別系統(tǒng)精度高,處理速度快,符合質(zhì)量預警控制的需求。 (2)提出基于FCP決策樹與聚類分析(Cluster Analysis,CA)的質(zhì)量預測方法。同時應用兩種數(shù)據(jù)挖掘算法分析不同種控制圖模式下的制造過程質(zhì)量數(shù)據(jù),在獲取聚類信息及異常質(zhì)量歸類的基礎上研究工序質(zhì)量的異常預測。通過質(zhì)量影響因素的聚類分析,獲取不同模式下的質(zhì)量影響因素組合,利用決策樹分析現(xiàn)有過程質(zhì)量數(shù)據(jù)實現(xiàn)質(zhì)量預測。 (3)將基于案例推理(Case-Based Reasoning, CBR)的知識庫引入質(zhì)量診斷系統(tǒng),實現(xiàn)了診斷知識的管理與自學習功能。通過對數(shù)控加工中心底座銑削制造過程實例中信息數(shù)據(jù)的處理,展示了知識提取與質(zhì)量診斷的全過程；最終建立MES環(huán)境下數(shù)控加工中心質(zhì)量預測診斷系統(tǒng),建立質(zhì)量預測數(shù)據(jù)庫、基于案例推理的知識庫及各子系統(tǒng)模塊的人機交互界面,實現(xiàn)了控制圖模式識別、質(zhì)量預測與診斷、知識庫的自學習與維護功能。
[Abstract]:Product quality is a concrete reflection of customer demand and a necessary factor for customer satisfaction. In order to occupy a favorable position in a highly competitive market and improve the competitiveness of enterprises, Modern enterprises must strengthen quality management. With the development of science and technology and the maturation of the market, the market demand for quality has been improved constantly, the existing quality management technology and tools can not meet the demand. The quality data in production process in manufacturing Execution system mes environment is collected by workshop job field control, analyzed and processed by system and engineer and stored in a specified database or data warehouse. In order to make full use of the large amount of data collected in the production process, the accumulation of such data leads to a decline in the ability of the analyst to process the data, and the potential value of the data is not further mined, resulting in a waste of information. Achieve total quality management, The technology and system of intelligent manufacturing process quality diagnosis and control based on data mining technology (DM) has become a new research hotspot for many experts and enterprises. Based on the summary of previous researches, this paper constructs manufacturing under MES environment. Process quality data mining platform, The quality control system with the function of pattern recognition, quality prediction and diagnosis of control chart is studied by analyzing the relevant data of processing quality. The main research contents include the following three aspects:. 1) according to the characteristics of manufacturing process data, a kind of uncertainty metric: F 尾 -confidence ratio F 尾 is proposed for process quality data analysis, and a decision tree based on FCP impureness metric is established. The control chart pattern recognition system based on FCP decision tree is constructed, and the statistical quantity is used as the statistical feature of the control chart pattern recognition for the phenomenon with high dimension of the control chart data. The test results show that the pattern recognition system has high precision and fast processing speed. In line with the quality of early warning control requirements. (2) A quality prediction method based on FCP decision tree and cluster analysis is proposed, and two kinds of data mining algorithms are used to analyze the quality data of manufacturing process under different control chart models. On the basis of obtaining clustering information and classifying abnormal quality, the abnormal prediction of process quality is studied. Through clustering analysis of quality influencing factors, the combination of quality influencing factors under different models is obtained. Using decision tree to analyze the existing process quality data to achieve quality prediction. The knowledge base based on Case-Based reasoning (CBR) is introduced into the quality diagnosis system, and the management and self-learning function of diagnosis knowledge is realized. The whole process of knowledge extraction and quality diagnosis is demonstrated. Finally, the quality prediction and diagnosis system of NC machining center under MES environment, the quality prediction database, the knowledge base based on case-based reasoning and the man-machine interface of each subsystem module are established. The functions of pattern recognition, quality prediction and diagnosis, self learning and maintenance of knowledge base are realized.
【學位授予單位】：山東大學
【學位級別】：碩士
【學位授予年份】：2014
【分類號】：TP311.13;TB114.2

【參考文獻】

相關期刊論文 前10條

1 李雄飛;李軍;屈成偉;劉麗娟;孫濤;;數(shù)據(jù)挖掘中平衡偏斜訓練集的方法研究[J];計算機研究與發(fā)展;2012年02期

2 馬秀紅,宋建社,董晟飛;數(shù)據(jù)挖掘中決策樹的探討[J];計算機工程與應用;2004年01期

3 侯世旺;同淑榮;;基于小波重構的控制圖并發(fā)異常模式識別研究[J];計算機工程與應用;2008年28期

4 施蕾;唐艷琴;張欣星;;數(shù)據(jù)挖掘中決策樹方法的研究[J];計算機與現(xiàn)代化;2009年10期

5 何文;同淑榮;王克勤;;基于神經(jīng)網(wǎng)絡的質(zhì)量控制圖異常診斷[J];機械制造;2013年05期

6 湯文宇;李玲娟;;CBR方法中的案例表示和案例庫的構造[J];西安郵電學院學報;2006年05期

7 孫林;楊世元;;基于SVM的柔性生產(chǎn)模式下生產(chǎn)過程質(zhì)量智能預測[J];系統(tǒng)工程理論與實踐;2009年06期

8 張根保;數(shù)字化質(zhì)量管理系統(tǒng)及其關鍵技術[J];中國計量學院學報;2005年02期

9 張敏;黎向鋒;左敦穩(wěn);繆宏;;基于主成分分析的BP神經(jīng)網(wǎng)絡內(nèi)螺紋冷擠壓成形質(zhì)量預測[J];中國機械工程;2012年01期

10 趙方方;何楨;;基于SVM的控制圖模式識別方法研究[J];組合機床與自動化加工技術;2009年08期

相關博士學位論文 前2條

1 于秀娟;加工過程的質(zhì)量異常預測與診斷方法研究[D];吉林大學;2011年

2 雷亞國;混合智能技術及其在故障診斷中的應用研究[D];西安交通大學;2007年

，

本文編號：1662730