本研究之目的是使用類神經網來補償壓電式力量感測器 (piezoelectric) 低頻時量測不準的現象。相較於改變感測器本身結構的方法來，本研究所提出之估測方法無須改變感測器的結構，使用商用的壓電式力量感測器，將其輸出給予力量估測系統；力量估測系統藉由類神經網路估出系統之干擾，並將受控體的受力量估測出來。
實驗平台是由實驗室成員共同設計出的，含一維線性伺服馬達系統，並採用美國德州儀器(Texas Instruments, TI)生產之TMS320C6713 DSP (Digital Signal Processor)開發板，搭配實驗室成員自行研發具備FPGA (Field-Programmable Gate Array) 等IC之擴充子板，作為控制核心。於FPGA方面，以VHSIC硬體描述語言(VHDL)實現編碼器、ADC與DAC等周邊界面訊號處理介面；在控制法則實現上，透過TI提供的Code Composer Studio (CCS)發展環境軟體，以C/C++撰寫控制器程式，並下載至DSP執行。藉由實驗室成員自行設計、組裝之一維線性馬達平台進行實驗，並感測負載受力與其他物理量。實驗結果顯示，本文提出之方法能有效地改善壓電式感測器的低頻量測失準現象;與先前的研究相比，使用類神經網路能在較劇烈變化的路徑上有更好的干擾估測效果，使系統能達到良好的位置與阻抗控制結果。

This research presents RBF neural networks for estimating an external force acting on a linear motion stage. The lower-frequency part of estimated force is then used to compensate force measurement from a piezoelectric force sensor, which is unable to measure low-frequency components of an external force. Compared to previous schemes that change a sensor's mechanical structure in order to solve this problem, this research uses an RBF neural networks for estimating an unknown disturbance, and estimate the real contact force is estimated without modifying the sensor's mechanical structure. In other words, a commercially available piezoelectric force sensor provides force data to an adaptive algorithm that can estimate force precisely, including low-frequency range.
The experimental system consists of a linear servomotor system, and a TMS320C6713 DSP (Digital Signal Processor) from Texas Instruments is used with a self-developed FPGA (Field-Programmable Gate Array) daughter board, as the control kernel. ADC, DAC and other interface are realized on the FPGA by employing VHSIC (Very High-Speed Integrated Circuit) hardware description language (VHDL), and control algorithm is realized on the DSP by employing the C/C++ language under CCS (Code Composer Studio) development environment. Force and other data are acquired from a one-dimensional linear platform to a personal computer for data analysis. The experimental results show that the proposed scheme improves the accuracy of the quartz force sensor in terms of lower-frequency contact force, and also compensates for the perturbation to the system.

[1] Ming Li et al., “A high performance piezoelectric sensor for dynamic force monitoring of landslide,” Sensors, pp.17(2):394, 2017.
[2] 姚家翔，結合徑向基底函數網路與粒子群演算法於小型風機之最大功率追蹤，中原大學電機工程學系研究所，碩士論文，2012。
[3] 鄭哲偉，線性參數化伺服系統之低頻接觸力估測，臺灣師範大學機電科技學系研究所，碩士論文，2018。
[4] 李柏辰，加速度觀測器於線性平台之追蹤控制應用，臺灣師範大學機電科技學系研究所，碩士論文，2013。
[5] 鄭百恩，雙軸機械手臂之適應性神經網路滑動模式控制，國立臺灣師範大學機電工程學系，碩士論文，2011。
[6] J. Song et al., “Experimental study on cascaded attitude angle control of a multi-rotor unmanned aerial vehicle with the simple internal model control method,” Journal of Mechanical Science and Technology, pp.5167－5182, 2016.
[7] J. Liu et al., “Development of a commercially viable piezoelectric force sensor system for static force measurement,” Measurement Science and Technology, vol. 28, no. 9, pp.1－7, 2017.
[8] J. Chen et al., “Low-frequency compensation of piezoelectric micro-vibrations test platform,” Technical Gazette, pp.1251－1258, 2016.