本文主要針對台農12號落花生之乾燥製程，利用多項式網路建構落花生含水率預測模式，以監測落花生在乾燥過程中之含水率狀況。在本研究中，利用具有自動合成建模技巧之多項式網路，得到落花生乾燥過程中之含水率狀況。本研究證實，多項式網路可以很正確地建立輸入變數（乾燥溫度、初始含水率，乾燥時間）與輸出變數（含水率）間之關係。經由實驗結果證實，只要輸入乾燥條件，便可以很準確地預測得知乾燥過程中之含水率變化狀況。因此，利用多項式網路所建構之含水率預測模式，可以相當有效地使用於落花生乾燥之含水率監控。

This text mainly aims at the dry manufacturing process of an agriculture No.12 peanut . In order to monitoring the water content of the peanut during the drying process , the polynomial network is used to construct a water content prediction model.
The polynomial network is constituted with several function nodes; these function nodes can be self-organizing into the optimal network structures according to the predicted square error (PSE) criteria.
It is shown that the polynomial network can correctly correlate the input variables (drying temperature, initial water content and drying time) with the output variable (water content). Based on the water content prediction model constructed, the water content of the peanuts can be predicted with reasonable accuracy if the drying conditions are given and it is also consistent with the experimental results very well.