本研究針對車用半導體A公司生產48V Si MOSFET模組時，鋼嘴(Wedge)的保養頻率與更換週期之探討，利用機聯網的技術將機台生產日期、時間、生產參數、3D AOI量測數值傳至雲端資料庫處理中心備查，藉由資料預處理的方式進行初步的整理與歸納，透過迴歸分析判別自變數與因變數之間的相關性，再由主成份分析根據數據中共變異數的特徵性質找出數據中的主成份和特徵值，保留數據集中主要變異數，減少數據的維度，提高數據處理和分析的效率，最後透過類神經網路將數據代入透過類神經網路的輸入層，隱藏層通過非線性激活函數對數據進行處理和變換，輸出層則生成最終的預測結果或決策，藉此預測Wedge保養頻率預更換週期延長的可能性。;This study focuses on the discussion of the maintenance frequency and replacement cycle of steel nozzles (Wedge) used in the production of 48V Si MOSFET modules by automotive semiconductor company A. By utilizing Industrial Internet of Things (IIoT) technology, the production date, time, parameters, and 3D AOI measurement values of the equipment are transmitted to a cloud database processing center for record-keeping. Data preprocessing is performed for preliminary sorting and summarization. Regression analysis is used to determine the correlation between independent and dependent variables. Principal component analysis (PCA) identifies the main components and eigenvalues from the data based on the characteristics of the covariance, retaining the main variances in the dataset, reducing dimensionality, and improving data processing and analysis efficiency. Finally, neural networks are employed where data is input into the input layer, processed and transformed through nonlinear activation functions in the hidden layers, and the output layer generates the final prediction results or decisions. This approach predicts the potential for extending the maintenance frequency and replacement cycle of the Wedge.
