- 學位論文
應用集群分析於混料工作之安排
Applying Clustering Analysis to Arrange Material Mixing Operation
摘要
集群分析一直是資料探勘中非常重要的一個方法。當所要考慮分析的資料量過多時，或是目標不夠明確時，可以將所要進行的資料分群，把性質相近的資料歸成一類，讓同一群裡的資料同質性高，不同群的資料同質性低。目前集群方法可以分成多變量統計方法、類神經網路和遺傳演算法等工具，尤其是多變量統計方法最常使用。 在輪胎製造產業中，混料作業是最先的一道製程，所需混料的膠料成分會因輪胎產品種類的不同而有不同的混合配方，而對上百種不同配方之混料工作的安排也就變得十分複雜，且關係到混料工作之效率。因此本研究選取部分膠料資料為研究對象，利用集群分析的方法來進行混料工作之分群，以提升混料機台工作之效率。本研究藉由兩階段集群法：第一階段使用層次集群法（最近法，平均法，華德法）確定分成幾群；第二階段使用K-Means確認最後分群個數。再比較三種層次集群法，用華德法的分群結果較好。因此，針對本研究案例可將配料工作分成2群，並使用兩台混料機台來作業。
並列摘要
The Cluster Analysis can be acclaimed one of the most important methodologies for database establishment and analysis. In case of the necessity of analyzing excessive volume of data or incompetence of validating the research target, such the Cluster Analysis can be applied in processing the data required to be validated for classifying the data with similar attributes into one corresponding category and for creating each individual cluster constituted of data with high similarity attributes and a mechanism that there is a low relevancy among each of all individual clusters of data. Currently, the Cluster Analysis methodology includes such varieties as Multivariate Statistics Analysis, Artificial Neural Network or Genetic Algorithm in which the Multivariate Statistics Analysis has been applied most frequently. In the tire industry, the manufacture materials mixture shall be the preliminary manufacture process. The required mixture of plastic materials and the proprietary mixture composition formula vary with the tire intended to be produced. In case of a tire manufacture necessitating arrangement for hundreds of manufacture material composition formulas, the manufacture process scheduling for producing each individual type of tire would be significantly complicated and can definitely influence the working efficiency of manufacture material mixture substantially. Under the aforesaid concern, the data analysis for some rubber manufacture materials are thus selected as the case study for this research with intention for searching the improvement measures for enhancement of the working performance and efficiency of the raw rubber manufacturing materials mixers by means of applying the aforesaid Cluster Analysis in clustering and classification of raw rubber materials mixture works in progress. The TwoStep Cluster Analysis methodology is adopted by this research in which: in the 1st Phase, hierarchical clustering method (the Nearest Neighbor method and Average Linkage Method the Wards Method) are utilized for determination of the total number of cluster classification categories; and in the 2nd Phase, the K-Means statistics methodology is performed for validating and finalizing the total required number of the cluster analysis categories. Afterwards, the Hierarchical Cluster Analysis is conducted in this research for making comparisons and thus concluding that the Wards Method demonstrates a better Cluster Analysis performance and efficiency. Therefore, it can conclude a feasibility in this research that the required mixture of raw rubber manufacture materials shall be classified into two categories in conjunction with usage of two material mixers for executing the works in progress simultaneously.