頁籤選單縮合
題 名 | 由濕穀重量折算為乾穀淨重之轉換比率的估計=Prediction of the Conversion Rate for Estimating the Net Weight of a Raw Rice Grain-Lot Before Dring and Cleaning |
---|---|
作 者 | 林俊隆; 何榮祥; | 書刊名 | 中華農學會報 |
卷 期 | 185 1999.03[民88.03] |
頁 次 | 頁120-138 |
分類號 | 434.119 |
關鍵詞 | 水稻; 溼穀; 乾穀淨重; 折算率; 估計模式; Rice; Raw grain-lot; Net weight of dried and cleaned grains; Conversion rate; Prediction model; |
語 文 | 中文(Chinese) |
中文摘要 | 所謂乾穀折算率即由溼穀之重量折算為經乾燥與風選後之乾穀的浮重(以13.5%含 水率為準)之轉換比率。本研究之目的在開發一個可在一批溼穀被乾燥與風選之前,即能由 其台水率 (Xi, %) 與容積重 (X,g/L) 預估其乾穀折算率 (Y,%) 的模式。 我們自 1993 年第二期稻作起,至 1996 年第一期稻作止, 在臺灣西部各主要水稻產 區內進行調查;總計獲得梗稻數據共 1193 組,秈稻數據共 432 組觀測值。 首先,將雨型 品種之全部數據分別劃分為兩部分:其中一部分約含全部數據的三分之二,被用來估算模式 中的參敷(模式建立數據),另一部分則被保留來測試後所開發之模式的估計能力(模式翰 驗證數據)。 我們以只含 X �竣峖P時含 X �絰P X �砟坐G次多項模式,及其各種退化模式 求配於「模式建立數據」。並由「模式建立數據」中逢機取出 30 套樣品數據(梗型品種每 套含 400 組,秈型品種含 150 組觀測值 ),反複估算各候選模式之參敷,藉之評量各模式 之參數估值的穩定性。而循此選出之參數估值穩定性高並與數據配合良好的模式,則進一步 以「模式驗證數據」測試其估計能力。 結果,我們在兩型品種均獲得一個令人滿意的模式: Y=10734-1.1636X �� +0.0114X �砥@(梗型品種), Y=92.47-1.2809X �� +0.0439X �砥]釉型品種 )。 這兩種模式除了在既有的樣品數據範圍內有準確的估計能力外,在樣品數據範圍之外的外插 能力也令人滿意。 |
英文摘要 | The conversion rate has been defined as the proportion of the net weight of dried and cleaned grains (on a basis of 13.5% moisture content) out of a lot of raw rice grains.This study attempted to develop a model to predict the conversion rate (Y, %) from the initial moisture content (X �� , %) and volume weight (X ��, g/L) of a lot of raw grains, thereby the net weight of that grain-lot could be estimated before the drying and cleaning process. From the second crop season of 1993 to the first crop season of 1996, a survey was conducted over the main rice production areas on the west plain of Taiwan. A total of 1193 observations for Ken varieties and 432 observations for Sen varieties were obtained. The complete data set was randomly split into two portions. One which contained about 2/3 of the total observations (the estimation data) was used to estimate the model parameters, and the other (the validation data) was reserved as an independent data set to check the prediction accuracy of the developed model. Second-degree polynomial models in X ��, and in both X �� and X ��, and their reduced models were fitted to the estimation data. The stability of parameter estimates for each candidate model was measured by fittng the models to 30 data sets (each contained 400 observations for Ken varieties, and 150 ovservations for Sen varieties) randomly resampled from the estimation data. Only those models with stable parameter estimates and good fit to the estimation data were selected for further testing by the validation data. The procedures ended up with a satisfactory prediction model for each type of rice varieties: Y = 10734 -1.1636X �� + 0.0114X �� (for Ken varieties), Y = 92.47-1.2809X �� +0.0439X �� (forSen varieties). The models gave not only accurate prediction within the sample X-space but also good extrapolation outside the region of the data. Key words: rice, raw grain-lot, net weight of dried and cleaned grains, conversion rate, prediction model (1) Professor, Department of Agronomy, National Chung-Hsing University, Taichung. Taiwan. |
本系統中英文摘要資訊取自各篇刊載內容。