請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/9487
標題: | 運用傅立葉轉換紅外線光譜儀及輻射形光徑煙流分布重建法定位污染源:重建演算法之評估 Using 2-D Radial Plume Mapping Technique with OP-FTIR for Source Localization: Evaluation of Reconstruction Algorithms |
作者: | Shih-Ying Chang 張世穎 |
指導教授: | 吳章甫 |
關鍵字: | 定位,空氣污染,開徑式傅立葉轉換紅外線光譜儀,輻射形光徑煙流重建法,光學遙測,逸散源, source localization,plume reconstruction,optical remote sensing,OP-FTIR,air pollutant,RPM,CT, |
出版年 : | 2008 |
學位: | 碩士 |
摘要: | 輻射形光徑煙流重建法為使用光學遙測儀器進行污染物分布重建及污染源定位之技術。在此技術中,兩種主要之重建演算法被用來進行污染物分布重建-Sooth basis function minimization (SBFM)和Non-negative least square (NNLS)。本研究之目的在比較此兩種主要重建法對於污染源定位及污染物分布重建之表現,此兩種演算法之相同處為重建出與測量得到之PIC相同的PIC,差異處為SBFM使用預選之基本方程式描述污染物之分布,而NNLS則是直接重建逸散區域污染物之濃度。除此之外,在SBFM重建中,我們使用兩種不同之基本方程式(對稱和非對稱)來描述污染源。
本實驗由兩部份構成,第一部份為電腦模擬實驗,在電腦模擬中,首先產生450個基本分布並使用輻射形光徑煙流重建法重建污染物之分布及污染源。結果顯示SBFM使用非對稱基本方程式(bivariate lognormal distribution)做為基本方程式可以重建出完整之污染物;而當污染源位置接近OP-FTIR時,SBFM使用對稱之基本方程式(bivariate Gaussian distribution)做為基本方程式時可以相當精確地重建出污染源之位置,但此時NNLS無法重建出正確之汙染源。而當污染物遠離OP-FTIR時,使用NNLS進行重建會得到較好的結果而SBFM使用對稱之基本方程式無法定位污染源。 本研究之第二部份為實地實驗,使用開徑式傅立葉轉換紅外線光譜儀及輻射形光徑煙流重建法實地進行人為釋放之污染源定位。實地實驗之結果與電腦模擬之結果相似,當污染源遠離OP-FTIR時,使用NNLS重建演算法可以得到相當精確之污染源評估,而當污染源靠近OP-FTIR時,使用SBFM則可以得到較好的結果。此外,由三種重建方法所重建之污染源可以指出真實污染源之方向,佐以重建之污染源附近之較短之測線,可以加以判斷由何者重建出之污染源最接近真實污染源;當附近之短光徑沒有偵測到污染物時,可以選擇NNLS做為重建演算法,而當附近之短光徑測量到污染源時,則可使用SBFM做為重建演算法。 The OP-FTIR measurement combining the RPM technique is able to reconstruct the plume and thus localize the emission source. In this thesis, both the computational simulation and the field experiment are implemented. Two major kinds of the reconstruction algorithm used in RPM technique are evaluated. The first one is the smooth basis function minimization (SBFM) algorithm and the second one is the non-negative least square(NNLS) algorithm. The two algorithms are both implemented by fitting the reconstructed path integrated concentration (PIC) to the measured PIC. The differences are that the SBFM superimposes a basis function to describe the plum while the NNLS directly estimate the concentration value in the emission domain. In addition, two different kind of basis functions (symmetric and skewed) are used to describe the plume in SBFM reconstruction. In the simulation analysis, 450 test distributions are generated to be localized by the RPM technique with different reconstruction algorithms. The result shows that the SBFM algorithm using the bivariate lognormal distribution as basis function gives the best result in both the aspects of plume reconstruction and source localization. Furthermore, when the plume is near the OP-FTIR, the SBFM reconstruction using bivariate Gaussian distribution as basis function may yield better result in the aspect of the source reconstruction comparing to the NNLS reconstruction. However, when the plume is far from the OP-FTIR, the NNLS reconstruction is able to localize the emission source more accurately than the SBFM using bivariate Gaussian distribution as basis function. In the field experiment, four experiments with four pairs of different source locations are conducted to be localized by the RPM technique. The result shows that the reconstructed source locations by the three methods are able to point out the correct direction towards the real source. Furthermore, judging by the peripheral short monitoring lines, the reconstructed source location that is closest to the real source location can be chosen and gives the best estimation of the emission source location. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/9487 |
全文授權: | 同意授權(全球公開) |
顯示於系所單位: | 環境衛生研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-97-1.pdf | 2.46 MB | Adobe PDF | 檢視/開啟 |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。