- 學位論文
機器學習於無人搬運車與自動倉儲設置間之搬運時間預測
Machine Learning for Transportation Time Prediction between AGV and Automatic Stocker
摘要
隨著全球疫情持續延燒，大大改變了人們原本的生活方式，也侷限了人們的社交行為，遠距學習與線上會議已是社會的常態，同時也加劇了企業、校園以及家庭對於3C產品的需求。面板和半導體產業因為疫情的關係，生產業務反而逆向急速成長。雖然國內面板廠和半導體廠的高科技生產設備以及技術都已日漸成熟，邁向高度自動化、電腦整合化，然因典型的生產流程大多將物料搬運設備視為服務或支援系統，普遍採用需求發生時才派遣的即時派工策略，當生產作業驟增時會發生搬運系統塞車、延誤、指令大量累積等，以致搬運效率明顯不佳。當工廠的生產作業爆量增加，在產能全開且吃緊下常發現依循既定法則執行的自動物料搬運系統，無法進行有效物料搬運，成為訂單驟增下的生產瓶頸。具有智慧能迎合需求變動敏捷派遣的物料搬運系統，已為高科技業生產系統邁入智慧製造不可避免的挑戰。 本研究案將著重在給定生產排程下TFT-LCD面板搬運的議題，主要研究範圍為自動倉儲設置與無人搬運車連結的搬運系統，藉由機器學習方法來對總運輸時間做出預測，從已知的歷史資料加以分析，利用整理各個需搬運物件在各站點的等候時間，以及等候搬運之物件數量，以及各物件在各個站點之間所需要的傳送時間，同時也考慮各個物件搬運的優先權重等，並且根據工廠實際狀況模擬可能發生之問題，得出資料中之規律，並利用此規律來對未來的資料進行預測。
並列摘要
As the global epidemic continues to spread, people's original lifestyles have been greatly changed. Distance learning and online meetings are normal in the world, and they have also increased the demand for 3C products from companies, campuses and families. The panel and semiconductor industries grow rapidly due to the epidemic. Although the high-tech production equipment and technology of domestic panel factories and semiconductor factories have matured and been moving towards a high degree of automation and computer integration, because most of the typical production processes regard material handling equipment as a service or supporting system. The real-time dispatching strategy of dispatching only when demand occurs is generally adopted, when the production operation increases sharply, the handling system will be found to be exhausted, traffic jams, delays, a large number of instructions...etc., so that the handling efficiency will be obviously not good. This research will focus on the issue of TFT-LCD panel handling. The main research is the handling system of automatic stocker and automation guided vehicle. The total transportation time is predicted by machine learning methods. Analyzing the known historical data, using the waiting time of each object to be transported between each station, and also considering each object's priority of job to find the actual conditions of the factory and simulate possible problems and predict the transportation time in the future.