DSpace community: 資訊工程學系碩士在職專班

應用客製化 MES 與 BI 系統於隱形眼鏡工廠之即時外觀良率優化

Fri, 06 Mar 2026 10:52:41 GMT

title: 應用客製化 MES 與 BI 系統於隱形眼鏡工廠之即時外觀良率優化 abstract: 隱形眼鏡產業正面臨少量多樣需求的挑戰。傳統人工管理無法支援頻繁換線下的數據整合，導致異常修正延遲並負面影響直通率（Yield Rate, YR）。本研究透過整合商業智慧（BI）與強化之製造執行系統（MES）來解決此問題。遵循 ISO 13485 規範，本研究開發了客製化 MES 模組，具備支援合併工單的流程卡設定（Run card）與多人協作檢驗輸入功能。來自八大機台的異質數據經 ETL 整合至 MySQL 資料庫，並由 Power BI 提供即時看板。結果顯示，缺陷報表整理時間由 50 分鐘縮短為即時顯示。即時警報促成參數快速調整，有效減少氣泡等關鍵缺陷。外觀良率從 68.41%（2024 Q2）提升至 70.92%（2024 Q4）。此 2.51% 的提升每月約減少 2.5 萬片報廢，證實本系統能有效將被動檢討轉變為主動的即時製程驗證。;The contact lens industry faces challenges from high-mix low-volume (HMLV) demands. Traditional manual management fails to support data integration during frequent line changeovers, causing delayed anomaly corrections and negatively impacting yield rates. This study integrates Business Intelligence (BI) with an enhanced Manufacturing Execution System (MES) to address these issues. Adhering to ISO 13485, this study developed customized MES modules featuring run cards for merged work orders and collaborative inspection inputs. Heterogeneous data from eight machines were integrated into a MySQL database via ETL, with Power BI providing real-time dashboards. Results show defect reporting time decreased from 50 minutes to real-time. Immediate alerts enabled rapid parameter adjustments, reducing critical defects like bubbles. The cosmetic yield rate improved from 68.41% (Q2 2024) to 70.92% (Q4 2024). This 2.51% increase saves approximately 25,000 lenses monthly, confirming that the system effectively transforms passive reviews into active, real-time process verification.

AI-Qos具反應時間保證與品質可控之AI技術;Latency-Aware Inference Techniques with Controllable Quality and Response-Time Guarantees

Fri, 06 Mar 2026 10:52:26 GMT

title: AI-Qos具反應時間保證與品質可控之AI技術;Latency-Aware Inference Techniques with Controllable Quality and Response-Time Guarantees abstract: 隨著即時 AI 應用與串流推論服務的快速發展,AI 推論系統在實際部署環境中所面臨之反應時間與服務穩定性問題日益顯著。傳統推論架構多以固定模型與離線準確率為主要設計考量,難以因應網路延遲波動、模型計算成本差異與系統負載變化所造成的端到端延遲累積,進而影響即時性與推論結果之可用性。為回應實際應用場域需求,本研究提出一套具反應時間感知與品質可控之AI-QoS 推論架構,透過即時延遲量測與場景特性分析,動態調整推論策略以兼顧即時性與推論品質,並設計自適應模型調度機制,於不同延遲與系統負載條件下選擇最適推論模型。實驗結果顯示,所提出之方法能有效降低端到端延遲對推論品質之影響,並在動態環境中維持推論服務之穩定性與可預測性,證實其於即時串流推論與實際 AI 應用部署中具備可行性與實務價值。;The rapid growth of real-time AI applications and streaming inference services has made responsiveness and service stability critical challenges in deployed AI systems. Conventional inference pipelines rely on fixed models and offline accuracy metrics, which are inadequate for handling end-to-end latency accumulation caused by network variability, heterogeneous model computational costs, and dynamic system workloads, resulting in degraded timeliness and inference usability. This work proposes a latency-aware and quality-controllable AI-QoS inference framework that dynamically adapts inference strategies based on real-time latency measurements and scene characteristics. An adaptive model scheduling mechanism is introduced to select suitable inference models under varying latency and resource conditions, balancing inference quality and responsiveness. Experimental results show that the proposed framework effectively mitigates the impact of end-to-end latency on inference quality while maintaining stable and predictable performance in dynamic environments, demonstrating its practicality for real-time streaming inference and deployed AI applications.

A Transformer-Based Model for Long-Term PM2.5 Forecasting

Fri, 06 Mar 2026 10:52:05 GMT

title: A Transformer-Based Model for Long-Term PM2.5 Forecasting abstract: 準確的 PM2.5 濃度預測對於空氣品質監測與管理決策至關重要，然而，受限於 PM2.5 固有的非平穩性、複雜的空間變異特徵以及時間依賴關係，使得精準的預測任務依然面臨嚴峻挑戰。針對上述難題，本研究提出了一 Transformer 架構模型，透過混合趨勢與殘差的預測機制來有效應對這些問題。具體而言，整合了空間與時間嵌入（spatial–temporal embeddings）以捕捉時空相關性；引入序列分解模組，將複雜的時間序列有效拆解為長期趨勢項與高頻波動項；並採用 De-Stationary Attention 機制，使模型能動態適應隨時間偏移的資料分布變化。為了驗證模型的實效性，我們採用台灣環境部提供的真實 PM2.5 觀測資料進行廣泛的評估實驗。結果顯示，相較於當前先進的 Transformer 變體模型及傳統線性基準方法，本研究所提出的模型在多種不同的預測步長下，均展現出更為穩定且優越的預測精確度。整體而言，本研究提出的方法提供了一個有效框架，可用於提升 PM2.5 之中長期預測能力。;Accurate PM2.5 forecasting is essential for air-quality management, yet remains challenging due to strong non-stationarity, spatial variability, and long-range temporal dependencies. In this work, we propose a Transformer-based modeling method designed to address these challenges through a hybrid trend–residual forecasting mechanism. The method incorporates spatial–temporal embeddings, a decomposition module that separates long-term trends from high-frequency fluctuations, and a De-Stationary Attention mechanism that adapts the model to shifting data distributions. We evaluate the proposed approach on PM2.5 records from Taiwan’s Ministry of Environment. Experimental results show that the model achieves consistently superior accuracy across multiple forecast horizons compared with recent Transformer-based models and linear baselines. Ablation studies further verify the contributions of the hybrid design and non-stationarity handling. These findings demonstrate that our Transformer modeling provides an effective framework for long-term PM2.5 forecasting.

傳統醫療手套業製造執行系統（MES）開發與導入實證－以P公司為例;Development and Implementation of a Manufacturing Execution System (MES) in the Traditional Medical Glove Industry: A Case Study of Company P

Fri, 06 Mar 2026 10:51:50 GMT

title: 傳統醫療手套業製造執行系統（MES）開發與導入實證－以P公司為例;Development and Implementation of a Manufacturing Execution System (MES) in the Traditional Medical Glove Industry: A Case Study of Company P abstract: 醫療手套業屬於典型的連續式製程，其生產線以模具掛載軌道為核心，手套需依序經過清洗、浸漬、烘乾與硫化等多道工序，對製程穩定性與品質管理具有高度要求，然而，傳統手套產線多仰賴人工方式進行品質檢驗，導致品質數據蒐集延遲與缺模等異常狀況無法及時掌握，進而影響產品品質與產線效率，為改善上述製程管理與品質監控之問題，本研究以實際醫療手套產線為研究對象，開發並導入一套輕量化製造執行系統（MES），整合企業資源規劃系統（ERP）工單資訊、監控系統（SCADA）設備參數以及品質檢驗數據，並透過光電計數器與即時儀表板，建立缺模率即時監控與異常警示機制，以數位工具實現即時數據驅動，提升製程透明度與管理效能。本研究採用系統設計與實務驗證並行之研究方法，將所建構之MES實際部署於個案工廠進行驗證。實驗結果顯示，系統有效降低缺陷發生率，提升產線管理效率，研究成果展現即時監控與數據驅動管理的優勢，並為傳統醫療手套產業之數位轉型提供具體且可行的實證依據與參考。;The medical glove industry is characterized by a continuous manufacturing process in which production lines are centered on mold-mounted conveyor tracks. Traditional glove production relies heavily on manual quality inspection, resulting in delayed data collection and limited real-time visibility of quality issues. Among these issues, missing molds are a major common cause of defects such as pinholes, uneven thickness, and surface imperfections. To address these challenges, this study develops a lightweight Manufacturing Execution System (MES) that integrates work order information from an Enterprise Resource Planning (ERP) system, process parameters from a Supervisory Control and Data Acquisition (SCADA) system, and quality inspection data. Photoelectric sensors and real-time dashboards are introduced to enable real-time monitoring and abnormality alerts for mold missing rates, thereby improving process transparency and production management efficiency. This study adopts a combined approach of system design and practical validation. Experimental results indicate that the proposed system effectively reduces defect occurrence rates and enhances production line management efficiency, providing practical empirical evidence for digital transformation in the traditional medical glove industry.