Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/1489
標題: 以車輛遙測來篩選高污染車輛的效果分析
The Effect Analysis of the High Emission Vehicles by Remote Sensor Detecting
作者: 王舜平
Wang, Shuen-Ping
關鍵字: Remote sensing
遙測
High Emitter Profile Model
Reproduced
LPG vehicles
Artificial Neural Work Model
高污染車輛預測模式
重覆檢測
LPG車輛
類神經網路
出版社: 機械工程學系
摘要: 以較低的成本及時間篩選出高污染懸疑車輛,是遙測的主要功能。行政院環保署於民國八十五年引進此項技術,至九十年止共取得超過七百萬筆數據,本研究即是利用這些數據來探討國內汽車污染排放的特性。 本文首先針對LPG車輛進行分析,發現LPG車輛的污染排放較一般車輛或是營業車輛為嚴重。且常使用LPG當燃料並不會降低污染排放。 在重覆檢測方面,不論重覆檢測多少次,其污染分佈皆和全體的污染分佈雷同,且前幾次的檢測結果和下次檢測的不合格率是具有相關性的,而要降低遙測的不確定性,增加遙測的次數是可行的方法之一。本文發現如果有輛車前二次被檢測皆為不合格,則再次被檢測到時,應有50%的可能性被判定為不合格。 在高污染車輛預測模式方面,本文以類神經網路來建立高污染車輛預測模式,發現單以遙測數據來判定高污染車輛是不夠的,必須配合廠牌、車用途、出廠日期和排氣量等車籍資料。而單就本模式來看,遙測CO值、遙測HC對數值(負相關)和車齡是影響模式CO濃度的主要原因。而在HC方面,則為廠牌、排氣量和車齡。車齡不論對CO還是HC,都有相當大的影響。在誤判率方面,CO為9-11.2%,HC為4.5-5.6%,皆有不錯的效果,且高污染模式在HC上的運用較佳。而拿另外的數據來驗証模式,可以發現誤判率介於11~13%。由此可見,本高污染模式具有相當的普遍性。
Screening highly emitted vehicles and reducing the testing cost and time are the main functions of remote sensing devices. The ROCEPA started the remote sensing program in Taiwan since 1996, and more than 7 millions data have been collected up to 2001. The objective of this study was to analyze these data to find out the characteristics of vehicle emissions in Taiwan. It was found from the remote sensing data that LPG vehicles emitted more pollutants than conventional gasoline vehicles. The frequency of refueling does not affect the emission levels of these LPG cars. Analysis of the reproduced data showed that the emission distributions were the same for vehicles tested more than once. It was found in this paper that the reproduced data of the same vehicle are related. Increasing the reproduction times is one of the ways to reduce the uncertainty of remote sensing data. If a car had been sensed twice and failed each time, then the probability that it would be fail again is over 50% in the third time. A high emitter profile model (HEP) has been built in this paper with the artificial neural work model. It was found that the RSD CO ,the log of RSD HC and the model year are the most important parameters for the CO model. As for the HC model, the automobile manufacturer, the engine displacement, and the model year are the most important parameters. The false ratio of CO is 9-11.2% and 4.5-5.6% in HC for the training data and testing data. For validating data, the false ratio were between 11-13%.
URI: http://hdl.handle.net/11455/1489
Appears in Collections:機械工程學系所

文件中的檔案:

取得全文請前往華藝線上圖書館

Show full item record
 
TAIR Related Article
 
Citations:


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.