| 摘要: | 時至今日，端銑在工業上已是一項最常被使用的加工程序，而現代工業要求產品同時具備生產速度與高品質。對切削程序來說，振動是一項影響產品品質的重要因素，影響範圍包括了加工精度以及產品的表面特徵。而為了抑制、甚至做到消除振動的影響，我們需要一個有效的方法來偵測或是重建振動信號，因此如何使用信號分析一直以來都是一個重要的議題，因為這是唯一一個幫助我們接觸真實、類比世界的手段。然而現實上，受限於技術與分析的難易度，我們經常被迫簡化或放棄一些資料中的重要信息。
希爾伯特黃由黃鍔院士在1998年提出，其具有的適應性令其在處理非線性與非穩態信號上有良好的表現。為此，本方法已經廣泛應用於聲音學、影像處理、醫學、氣象學與海洋學等各種領域之中。其包含兩個主要步驟，分別稱為經驗模態分解與希爾伯特轉換。由於相較其他常用分析法，此方法相對新穎並在實用證明例子上還不夠充分，我們需要透過實作來嘗試證明在加工方面應用的有效性。
在本研究裡，我們使用應變規與加速度來採得在端銑過程中產生的振動訊號，再透過希爾伯特黃轉換來分析、觀察其結果是否能真實呈現加工過程中所產生之各種振動現象。
;Until today, milling process has been became one of the most commonly used processes in industry. Present industry requires not only productivity but quality of products. To the cutting process, an important factor that infects qualities, such as accuracy and surface characteristic, is vibration. To ease or even cancel the effect of vibration, we need an effective way to detect or process the vibration signal. Therefore, there is always an important issue that how we make a signal analyzing. Because this is the only way to help us to connect the real analog world with the digital world. However, most of these algorithms are forced to premise their targets are linear or stationary that may lose the information of what exactly happened in the real world.
Hilbert-Huang Transform (HHT) was proposed by Huang et al. in 1998, which is well at solving non-linear and non-stationary data due to its adaptation. Therefore, it has been used in widely used in acoustics, image processing, medical, meteorology and oceanography now. It mainly consists of two steps, empirical mode decomposition (EMD) and Hilbert transform (HT). Since this is a relatively new method and lack of theoretical proof, we need as possible as we can to prove its efficiency.
In this research, we used the HHT to analyze data which collected by strain gauge and accelerometer in end-milling process to investigate the applicability of HHT on machining process and, due to the experiment results, we had some conclusions as below:
1. The Margin spectrum which built by HHT had almost the same distribution as FFT spectrum in frequency domain, which means it could be used to get frequency domain data instead of using FFT.
2. Different with wavelet transform, HHT spectrum is able to present the change of vibration generated in cutting process in time-frequency domain and keeps the same resolution.
3. In a extra experiment, the combination of using strain gauge and HHT could also be useful to detect the tool wear. However, accelerometer data had a worse presentation than strain gauge was. The reason should be further investigated. |
