Embedded Files
1.工業機器人與智慧自動化 (Industrial robotics & intelligent automation)
研究機械手臂如何更穩定地運動、辨識工件，並與製程設備協同作業。從演算法到實際產線，連結路徑規劃、機器視覺與系統整合。
We investigate stable robot motion, workpiece perception and coordination with manufacturing equipment, connecting path planning, machine vision and system integration with industrial practice.
學生參與：機器人整合、產線專題與自動化競賽。
Student involvement: robot integration, production-system projects and automation competitions.
A.高速運動與振動抑制
High-speed motion & vibration control
以 B-Spline 軌跡與 L-BFGS-B 最佳化調整運動參數，兼顧軌跡精度與作業效率。
B-spline trajectories and L-BFGS-B optimization tune motion parameters to balance trajectory accuracy and operating efficiency.
B.製程需求導向的噴塗路徑
Process-aware spray paths
依鍛造模具幾何與潤滑需求，結合蟻群演算法與貝氏最佳化，規劃噴塗覆蓋與材料分配。
Ant colony optimization and Bayesian optimization plan spray coverage and material allocation according to forging-die geometry and lubrication needs.
C.視覺取放與彈性產線
Vision-guided handling & flexible production
整合合成資料、點雲補全與機械手臂，處理亮面工件反光及遮擋；串接 PLC、CNC 與機器人進行系統驗證，並以 LLM／MCP 整合生產資料與智慧工廠決策。
Synthetic data, point-cloud completion and robotic manipulation address reflections and occlusions, while PLC, CNC and robot integration enables system-level validation. LLM/MCP integration connects production data with smart-factory decisions.
代表成果｜鍛造噴塗產線驗證
Selected result | Forging-line validation
研究中的特定產線，鍛件不良率由約 725 PPM 降至約 35 PPM，展現製程需求導向路徑規劃的應用價值。
In the production line studied, the forging defect rate decreased from approximately 725 to 35 PPM, demonstrating the value of process-aware path planning.
代表論文
Selected publications
2025 · Robotic manipulator vibration control ↗
02
2.智慧加工與製程最佳化(Smart machining & process optimization)
2.智慧加工與製程最佳化(Smart machining & process optimization)
以虛擬加工模型與實機資料支援加工決策，探討品質、效率與能源使用之間的取捨，並將工業 AI 應用於加工狀態辨識。
Virtual machining models and machine data support decisions about quality, productivity and energy use, with industrial AI applied to machining-state recognition.
學生參與：CNC 程式實作、設備資料擷取與智慧製造系統建置。
Student involvement: CNC programming, equipment data acquisition and smart manufacturing systems.
A.數位雙生與加工決策
Digital twins & machining decisions
整合虛擬加工時間與切削力模擬，透過實體加工驗證參數選擇與品質預測。
Virtual machining-time and cutting-force simulations are checked through physical machining to support parameter selection and quality prediction.
B.進給率與負載最佳化
Feed-rate & load optimization
利用 CNC 內建主軸功率資訊，結合高斯過程回歸與 Pareto 多目標分析，兼顧加工時間、能耗與表面品質。
Built-in CNC spindle-power data, Gaussian process regression and Pareto analysis support trade-offs among machining time, energy use and surface quality.
C.合成資料與切屑辨識
Synthetic data & chip detection
以 Omniverse 建立帶標註的合成影像，搭配真實影像訓練 YOLOv8，辨識切屑堆積與纏屑異常。
Annotated synthetic images generated with Omniverse are combined with real images to train YOLOv8 for machining-chip and entanglement detection.
代表成果｜加工能耗與視覺辨識
Selected results | Energy use & visual detection
切削條件研究在論文比較條件下，總能耗降低 18.19%；另一項切屑辨識研究於獨立測試集達 mAP@0.5＝0.91、F1＝0.90。
The cutting-condition study reported an 18.19% reduction in total energy under its comparison conditions. A separate chip-detection study achieved mAP@0.5 of 0.91 and F1 of 0.90 on an independent test set.
代表論文
Selected publications
2026 · Digital-twin milling decision support ↗2025 · 進給率最佳化與負載均衡
2025 · Feed-rate optimization & load equalization ↗2025 · 切削條件與加工能耗
2025 · Cutting conditions & energy use ↗2025 · 合成資料驅動的切屑辨識
2025 · Synthetic-data-driven chip detection ↗
03
3.精密傳動元件與動態分析(Precision drives & dynamic analysis)
3.精密傳動元件與動態分析(Precision drives & dynamic analysis)
以傳動機構設計、結構分析與非線性動態建模為基礎，研究精密機械的穩定性，並探索機器人關節可靠度的後續應用。
Transmission design, structural analysis and nonlinear dynamic modeling underpin our study of precision-machine stability and future work on robot-joint reliability.
研究方法：機構建模、有限元素分析、動態模擬與資料模型比較。
Research methods: mechanism modeling, finite element analysis, dynamic simulation and model comparison.
A.雙臂機械手應力數位雙生
Stress digital twin for dual-arm manipulators
整合 CAE 降階模型與機械手運動模擬，預測不同姿態與路徑下的應力響應，支援路徑規劃與結構評估。
CAE reduced-order models are integrated with manipulator motion simulation to predict stress responses across poses and paths, supporting path planning and structural assessment.
B.諧波減速機與機構設計
Harmonic drives & mechanism design
累積波形產生器、減速機構與諧波齒輪相關研發，結合齒形設計及有限元素分析評估嚙合與結構響應。
Work on wave generators, reduction mechanisms and harmonic gearing combines tooth-profile design with finite element analysis of meshing and structural response.
C.氣浮軸承非線性動態
Nonlinear aerostatic-bearing dynamics
分析不同轉速與轉子質量下的動態軌跡、分岔與混沌行為，以最大李雅普諾夫指數等方法辨識系統穩定性。
Trajectories, bifurcations and chaotic behavior under varying rotational speeds and rotor masses are analyzed using measures including the maximum Lyapunov exponent.
D.機器學習與穩定性預測
Machine learning for stability prediction
以隨機森林與 XGBoost 混合模型預測氣浮軸承動態狀態，建立從物理分析到資料預測的研究流程。
A hybrid random-forest and XGBoost model predicts aerostatic-bearing dynamics, connecting physical analysis with data-driven prediction.
代表成果｜氣浮軸承狀態預測
Selected result | Aerostatic-bearing prediction
2024 年氣浮軸承研究的混合模型達 RMSE＝0.36、R²＝0.77。此結果對應氣浮軸承模型；機器人關節狀態辨識與可靠度為後續延伸方向。
The hybrid model in the 2024 aerostatic-bearing study achieved RMSE of 0.36 and R² of 0.77. These results concern the bearing model; robot-joint condition recognition and reliability are future extensions.
代表論文
Selected publications
2024 · Nonlinear bearing dynamics & forecasting ↗2024 · 投票模型與動態行為預測
2024 · Voting-model prediction of bearing dynamics ↗
NEXT / RESEARCH DIRECTIONS
未來研究規劃
未來研究規劃
Future research directions
Future research directions
延伸既有成果，規劃投入亮面與透明工件的智慧取放及精密組裝、微型諧波傳動與視觸覺靈巧操作、高階五軸與薄壁件加工，以及 LLM AI Agent 輔助的製造決策。
Building on existing work, planned directions include intelligent handling and precision assembly of reflective and transparent parts, miniature harmonic drives and visual–tactile manipulation, advanced five-axis and thin-wall machining, and LLM-agent-assisted manufacturing decisions.
Page updated
Google Sites
Report abuse