工學院: 環境工程學研究所指導教授: 駱尚廉楊博清Yang, Po-chingPo-chingYang2017-03-062018-06-282017-03-062018-06-282016http://ntur.lib.ntu.edu.tw//handle/246246/277179隨著半導體產業製程技術的演進，間接帶動著臺灣整個相關產業鏈蓬勃發展，投資的金額龐大，但產品的生命週期卻不長。因此當有建廠需求時，從營建工程進場到產品量產的時間約在18個月左右。本研究以MATLAB為工具，建構倒傳遞類神經網路預測模式，預測半導體廠房中水處理系統的工程造價預估，資料來源及範圍為第三方專業水處理系統工程公司於2005年至2015年間，參與國內外半導體廠的標案及專案資料共20筆。輸入參數為原水水質、系統要求產水水質及系統類別，輸出值為工程造價預估金額。由案例驗證可知，倒傳遞類神經網路模式可得到快速、精確的預估成果，可有效地預估專案成本。其預估準確率約為93.72%~99.65%。透過本研究所建置的類神經網路預估模型估算專案成本，可作為後續精算的參考依據。The evolvement of semiconductor manufacturing technology indirectly promotes the development of relative industry chain. The investment amount is large; however the product life cycle shows relatively short and it takes nearly 18 months from factory construction to mass production. Based on MATLAB, this thesis constructs Back-propagation neural network(BPNN) to predict the construction cost of water treatment system of semiconductor factories. The reference collected from a third party construction company specialized in water treatment system from 2005 to 2015 were taken for 20 cases including domestic and foreign semiconductor factory bids and cases. The quality of raw water is inputted and system demands the quality of outflow and system category and cost prediction of construction is outputted. Results reveal that a quick, accurate and effective cost prediction could be achieved by Back-propagation neural network(BPNN) and its accuracy is about 93.72% to 99.65% which means the Back-propagation neural network(BPNN) prediction model through this thesis could be a useful reference for following calculations.論文使用權限: 不同意授權工程造價倒傳遞類神經網路水處理系統工程Project costBack-propagation neural networkWater Treatment Systems Engineeringthesis10.6342/NTU201600745