應用於高動態研磨工具之研磨力控制系統開發

Abstract

本研究使用實驗室開發的自動化研磨系統進行研磨實驗，優化研磨工具性能以達到更好的研磨品質。首先針對舊版的研磨工具，又稱mini-robot，在研磨過程中無法穩定維持轉速以及六軸力規受干擾的問題進行了硬體改版，並取得了顯著的進步。改版後的mini-robot在研磨表現方面有了明顯的改善。因應硬體改變進行系統識別實驗，以確保系統的穩定性和韌性。 在系統識別的過程中，為了避免雜訊和力感測器的延遲特性對結果產生影響，採用了單一頻率個別掃頻的方式進行頻域系統識別。根據識別結果調整各控制器的參數，包括比例積分控制器（PI controller）、重複式控制器（repetitive controller）和峰值濾波器（peak filter）。其中，比例積分控制器（PI controller）的目標是穩定控制系統，而重複式控制器（repetitive controller）和峰值濾波器（peak filter）則用於抑制研磨輪旋轉引起的周期性震動。兩種控制器具有不同的特性，重複式控制器（repetitive controller）可抑制基頻與倍頻的雜訊，因為對高頻的雜訊也有響應，較容易造成系統不穩定。峰值濾波器（peak filter）抑制單頻率的雜訊，在使用時較不會產生不穩定的情形。 在評估不同控制器組合的系統時，考慮到砂布輪的磨耗問題進行了砂布輪的磨合實驗，找出適合的研磨正向力數值以及數據較穩定的研磨次序區段。此外，為了避免砂布輪個體差異對結果的誤判，我們依據多組實驗結果的統計數據，對控制器的效果進行了評估。最終，使用比例積分控制器（PI controller）、重複式控制器（repetitive controller）組成的控制系統使得mini-robot在研磨過程中能夠將力震盪控制在正負1N以內。也將mini-robot作為被動式吸震的研磨工具使用，有許多研究是以此種設計進行研磨。雖然被動式吸震的研磨工具因為不需要補償用的致動器製作成本較低，但無法補償機械手臂的軌跡誤差，本研究所使用的主動式的力補償研磨工具因為使用音圈馬達，可有效補償位置誤差。 研磨結果以表面粗糙度Ra值為基準，相較於之前的結果（Ra值平均值為2.198um），改進後的mini-robot表現更好，平均Ra值達1.85um，下降了約0.3um。然而在抑制方面仍有進一步的改進空間，通過改善硬體結構的剛性並配合控制器的調整有望取得更好的研磨效果。綜合而言，本研究有效提高mini-robot的研磨性能，並使其能夠在研磨過程中保持穩定的力震盪，從而獲得更好的研磨表面品質。

This study focuses on optimizing the performance of a robotic grinding tool developed in the laboratory to achieve better grinding quality. The initial version of the grinding tool, referred to as the mini-robot, faced challenges in maintaining stable rotational speed and dealing with disturbances in the six-axis force sensor during the grinding process. To address these issues, hardware modifications were made to the mini-robot, resulting in significant improvements in its grinding performance. System identification experiments were conducted to ensure stability and robustness. Frequency domain system identification techniques using single-frequency sweeping were employed. Based on the identified system characteristics, various control parameters were adjusted, including PI controllers, repetitive controllers, and peak filters. The PI controller aimed to stabilize the control system, while the repetitive controller and peak filter were implemented to suppress the periodic oscillations caused by the rotation of the grinding wheel. These controllers exhibited distinct characteristics, with the repetitive controller effectively suppressing noise at fundamental and harmonic frequencies, although it had a higher risk of instability due to its response to high-frequency noise. The peak filter targeted the suppression of specific frequency components and demonstrated greater stability during operation. In evaluating the performance of different control combinations, wear experiments were conducted to address the wear-related issues of the grinding wheel, determining suitable grinding force values and stable grinding sequences. To mitigate the impact of individual variations in the grinding wheels on the results, statistical data from multiple experiments were used to assess the effectiveness of the controllers. Ultimately, the combination of PI and repetitive controllers enabled the mini-robot to maintain force oscillations within ±1N during the grinding process. The mini-robot was also utilized as a passively dampened grinding tool, which has been commonly adopted in various studies. However, passive dampening tools cannot compensate for trajectory errors caused by the robotic arm, while the actively compensated grinding tool used in this study, employing voice coil motors, effectively compensated for position errors. The grinding results were evaluated based on surface roughness (Ra value). Compared to previous results (average Ra value of 2.198 μm), the improved mini-robot demonstrated better performance with an average Ra value of 1.85 μm, representing a reduction of approximately 0.3 μm. However, further improvements are still possible. By improving the rigidity of the hardware structure and making adjustments to the controller, it is expected to achieve better grinding results. In summary, this study successfully enhanced the grinding performance of the mini-robot, enabling stable force oscillations during the grinding process and yielding improved surface quality.