應用智慧型咖啡烘焙系統於咖啡烘焙風味口感之研究

在咖啡的製造程序中，透過烘焙製程可降低咖啡豆含水量，增加保存時間，且咖啡豆在烘焙過程中會釋放咖啡的酸、苦、甘等多種風味，改善風味品質的穩定性，焙火程度會影響咖啡的口感，其烘焙技巧需烘豆師大量的經驗。故本研究之主要目的在開發一套可藉由人工智慧自動調整及偵測咖啡豆烘培之裝置，透過IoT設備監控烘焙機內部環境，並使用RNN類神經網路自動控制烘培相關設備來控制烘焙曲線，以達到最適當的烘豆結果。採用人工智慧控制烘培設備，除可達到預期的烘培風味效果外，亦可降低烘豆人力需求及降低成本。使用人工智慧烘豆亦可紀錄烘培過程中的相關參數，不但可透明化整個烘培過程，也可以收集大量與完整烘培數據資料，作為未來更進一步的大數據分析使用。成功結合人工智慧與烘培作業，可將此技術推廣到其他相關烘培產業與設備開發，成為另一波的產業革命，將會是本研究的重大貢獻之一。咖啡豆烘焙後，經咖啡杯測評分表評測後之結果，淺焙咖啡豆平均得分為86.475、中焙咖啡豆平均得分為87.775、深焙咖啡豆平均得分為77.775，在口感風味上得到了最佳化的參數為中度烘焙之咖啡豆，烘焙時的火力大小為150 Kpa mmAq，風門刻度為全程烘焙時刻度5、脫水期刻度9，鍋爐轉速為50 RPM，烘焙時間落在10分鐘，咖啡豆經烘焙後失重率13%。

關鍵字

人工智慧；咖啡豆烘培； RNN類神經網路

並列摘要

In the coffee manufacturing process, through the roasting process, the water content of the coffee beans can be reduced, the storage time can be increased, and the coffee beans will release various flavours such as sour, bitter, and sweet coffee during the roasting process, and improve the stability of the flavour quality. The degree will affect the taste of the coffee, and its roasting skills require a lot of experience from the roaster. Therefore, the main purpose of this research is to develop a device that can automatically adjust and detect the roasting of coffee beans through artificial intelligence, monitor the internal environment of the roaster through IoT equipment, and use the recurrent neural network (RNN) to automatically control roasting-related equipment. Control the roasting profile for the most appropriate roasting results. Using artificial intelligence to control the baking equipment can not only achieve the expected baking flavour effect but also reduce the labour requirement and cost of baking beans. Using artificial intelligence to bake beans can also record relevant parameters in the baking process, which not only makes the entire baking process transparent but also collects a large amount of complete baking data for further big data analysis in the future. The successful combination of artificial intelligence and baking operations can promote this technology to other related baking industries and equipment development, resulting in another wave of the industrial revolution, which will be one of the major contributions of this research. After the coffee beans are roasted, the average score of the lightly roasted coffee beans is 86.475, the average score of the medium roasted coffee beans is 87.775, and the average score of the dark roasted coffee beans is 77.775. The optimized parameters are medium roasted coffee beans, the firepower during roasting is 150 Kpa more, the throttle scale is 5 during the whole roasting period, 9 on the dehydration period scale, the boiler speed is 50 RPM, the roasting time falls within 10 minutes, and the coffee The weight loss rate of beans after roasting is 13%.

並列關鍵字

Artificial Intelligence ； Coffee Bean Roasting ； Recurrent Neural Network (RNN)

參考文獻

Nehlig A. (2016). Effects of coffee/caffeine on brain health and disease: What should I tell my patients?. Practical neurology, 16(2), 89–95. https://doi.org/10.1136/practneurol-2015-001162

Wang, A., Wang, S., Zhu, C. et al. Coffee and cancer risk: A meta-analysis of prospective observational studies. Sci Rep 6, 33711 (2016). https://doi.org/10.1038/srep33711

王士文，2001，當咖啡的香氣瀰漫在巴黎的街頭─論法蘭西咖啡館文化的形成及發展，輔仁大學歷史研究所碩士論文。