渦流資料管理與分析系統之研製

Abstract

資料探勘技術在近年來快速的發展並且被運用來分析大量的資料以發掘出隱藏於資料背後的有用資訊。本論文中我們應用此技術設計並製作了一個渦流資料管理與分析系統。此系統包含了特徵擷取、重點影像擷取、區域及全域資料分類分類、及資料探勘器等主要的模組。特徵擷取模組用來擷取每一時間的流場特徵，這些特徵包括了視覺特徵以及統計特徵。區域資料分類模組將渦流資料切分為多個時間片段，並從每個時間片段中選取出一張重點影像。全域資料分類模組將渦流資料作全域的分類以歸納出主要的資料類別。資料探勘器模組則搜尋資料序列中的常見序列以及特徵間的關聯法則。本系統提供圖形使用者操作介面，方便研究人員操作系統以管理與分析大量的渦流資料。我們希望藉由本系統的研製不僅能夠在知識探勘技術上有進一步探討與應用，也能夠提供流力研究者一個好的分析與管理工具。

Data mining techniques have been applied to analyze large amount of data and discover useful information. In this thesis, we apply these techniques to design and implement a vortex flow data management and analysis system. There are four main modules in this system, including the feature extractor, local cluster, global cluster, and data miner. Feature extractor module extracts visual and statistical features of a flow field at each time step. The extracted features include visual features and statistical features. Local cluster module groups a set of flow field data into temporal segments and selects one frame from each segment as a key frame. Global cluster module groups vortex flow data into clusters and finds the main data types. Data miner module searches frequent patterns in data sequences and association rules among features. This system provides graphical user interface to interact with researches to manage and analyze large amount of vortex flow data. The implementation of this system will benefit both the information scientist in the context of knowledge discovery and at the same time provide an efficient flow data management and analysis tool for researchers in the computational fluid dynamics community.