改善滾珠螺桿缺失率之研究－以扭矩變化異常偵測

滾珠螺桿應用在許多的高精密的機台內，如半導體、CNC加工機等，為核心的關鍵零組件。螺桿施加預壓力可消除軸向背隙，使螺帽往覆運動時，較不易發生前後振動而達到高速與高定位精度之性能。在外部無負載的狀態下，連續轉動螺桿或螺帽所需之力矩，即為扭矩。以往的扭力量測為人工為主，但可能因每人操作或看量表問題，造成不良品往下製站程加工。近年來改以半自動量測平台，減少人員干涉，穩定度確有提昇，但仍僅能判斷是否符合公差。本研究以現有不良品資料，利用R語言進行程式開發，並以e1071套件進行建模（svm,機器學習），可做二階段的分析。第一階段利用DTW、相關分析演算法進行量測值波形圖比對，若為相關就可能為不良品。第二階段依量測值再進行機器學習之預測，是否為不良品，其可能發生之原因。減少誤差或測試時間，加快人員處理速度，並降低交接下製程的缺失率。

關鍵字

R語言； DTW ；相關分析；機器學習

並列摘要

Ball screw are used in many high-precision machines, such as semiconductors, CNC machines, etc., as the core components. The pre-pressure applied by the screw can eliminate the axial backlash, so that when the nut is moved over, the front and rear vibrations are less likely to occur, and the performance of high speed and high positioning accuracy is achieved. In the external unloaded state, the torque required to continuously rotate the screw or nut is the torque. In the past, the torsion force measurement was mainly manual, but it may be caused by the operation of each person or the problem of the meter, resulting in defective products going down to the station. In recent years, it has been changed to a semi-automatic measurement platform to reduce personnel interference, and the stability has been improved, but it is still only able to judge whether it meets the tolerance. This study uses the R-language for program development and the e1071 suite for modeling (svm, Machine Learning), which can be used for two-stage analysis. In the first stage, DTW and correlation analysis algorithms are used to compare the waveforms of the measured values. If they are related, they may be defective. In the second stage, based on the measured value, the prediction of machine learning is made, whether it is a defective product, and the reason for its possible occurrence. Reduce errors or test time, speed up personnel processing, and reduce the missing rate of the process under the transfer.

並列關鍵字

R ； DTW ； Correlation ； Machine Learning

國際替代計量

改善滾珠螺桿缺失率之研究－以扭矩變化異常偵測

全文下載

主題瀏覽