1、檸檬乾燥之研究The Study on the Drying of Lemons研究生：陳冠中 Kuan-Chung Chen指導教授：陳和賢、黃卓治 Ho-Hsien Chen、Tzou-Chi Haung 【摘要】乾燥為一種貯存食品的方法，而不同的乾燥方法直接影響脫水食品的品質、色澤和香氣。本研究主要以熱風乾燥機及利用已開發的密閉式循環乾燥機進行檸檬乾燥試驗。密閉式循環乾燥機由 36逐步加熱至 52，利用冷凝除濕原理達到脫水效果，及乾燥末期調降冷凝器風扇風速，則可有效提昇乾燥效率。田口實驗設計法，配合直交表之望小特性及望大特性，針對檸檬乾燥製程中的操作因子進行確認。進而利用田口方法所建立的
2、數據資料作為類神經網路的模式建立，其輸入變數為：階段性加熱時間、乾燥室風速、冷凝器風扇風速；輸出變數為：總乾燥時間、產品含水率、產品白度值。實驗結果顯示，依目前訓練模式所建立的三層式類神經網路，以隱藏層節點在 2 點下進行訓練，其學習可達 91以上的正確率，測試平均正確率在 90以上，並可得到良好的收斂效果。期望利用具有學習歸納能力的類神經網路，應用於同類型密閉式循環乾燥機檸檬乾燥製程之模擬及預測。關鍵字：檸檬、乾燥、田口方法、類神經網路 【Abstract 】Drying in one of the methods used to preserve food. In general, dif
3、ferent drying methods could affect the quality, color and aroma. In this study, a hot- air dryer and a new designed closed-type air-convection dryer were used to dry lemons. The new designed closed-type air-convection dryer was used to dry lemons starting from 36 to final temperature 52. The product
4、s were dehydrated by dehumidifing and condensing water from the flow air. Tuning air velocity of condenser fan at final stage also could improve the drying efficiency significantly. Firstly, L18 orthogonal arrays and signal-to-noise ratios of the Taguchi method were used to find out significant fact
5、ors in lemon piece drying processing. The artificial neural network (ANN) model was set up based on the input data obtained from the significant factors by Taguchi method. Input variables contained heating time, air velocity in drying chamber and air velocity of condenser fan. Output variables conta
6、ined total drying time, moisture content, white index. Its found that 3 layers neural network with 2 nodes of a hidden layer model could have 91 accuracy of test samples. The confirmation test could reach 90 accuracy and convergence. This study showed that neural network with learning and inducting ability was capable for those same type dryers used to simulate and predict lemon drying process.Keywords: lemon、drying、Taguchi method 、artificial neural network (ANN)
