01挑戰The challenge
前起落架擺振（shimmy）會影響飛機起降安全。過去以傳統可靠度分析安排維修，難以掌握每一架飛機實際的劣化程度，也不容易提早發現潛在故障。Nose landing-gear shimmy affects aircraft take-off and landing safety. Maintenance was planned with traditional reliability analysis, which made it hard to see how far each aircraft had actually degraded or to catch emerging faults early.
02我們的做法Our approach
資料來源為航空器上感測器擷取的訊號，涵蓋兩種行進狀態與多種訊號類型：The data came from sensors on board the aircraft, covering two operating phases and several signal types:
- 行進狀態：起飛、降落Phases: take-off and landing
- 訊號類型：時間、空速、側向振動Signals: time, airspeed and lateral vibration
我們針對這類訊號提出專屬的前處理方法，提取有效的關鍵特徵，分別建立故障診斷模型與剩餘可用壽命（RUL）預測模型，再以超參數最佳化（Hyperparameter Optimization）提升模型表現。We developed a dedicated pre-processing method for these signals, extracted the key features, and built both a fault-diagnosis model and a remaining useful life (RUL) model, then tuned them with hyperparameter optimization.
03成果Results
模型結果與傳統可靠度分析方式比較，並對照實際進廠維修紀錄驗證，準確度達 98% 以上。研究成果已發表於 SCI Q1 期刊 Aerospace 共兩篇。We compared the models with traditional reliability analysis and checked them against actual maintenance records. They reached over 98% accuracy, and the work is published in two papers in the SCI Q1 journal Aerospace.