模型參考自適應類神經網路之電液伺服控制

Model Reference Adaptive Control based on Neural Network for Electro-Hydraulic Servo System

劉源良(Yuan-Liang Liu)

指導教授 ： 康淵 ； 王世明

中原大學/工學院/機械工程研究所/博士(2003年)

https://doi.org/10.6840/cycu200300666

摘要

本文提出由控制器、受控體模擬器與參數調節器，三個不同功能之類神經網路，構成多重類神經網路適應性控制，應用於電液比例閥控變量泵與伺服閥控液壓系統，在未知受控體數學模型或物理參數的情況下，預先以離線訓練模擬器近似系統動態特性，控制器採用 適應法則，以誤差及誤差微分項分別乘以參數來近似倒傳遞誤差項，並藉由類神經網路參數調整器線上調整參數值，得到優化之參數值，使神經網路收斂過程具有更高的穩定性，讓受控體之輸出可快速追隨參考模型反應，減少參考模型與受控體輸出誤差，改善以人為調整與經驗選擇參數之偏差，使受控體輸出得到理想反應。
本文所提出的多重類神經網路適應性控制，以Lyapunov定理證明其穩定性，並與直接、間接與神經調參類神經網路控制比較，應用於可變排量泵時，以不同步階與偽隨機方波為偏轉角命令，應用於伺服閥控液壓缸時，以週期性方波做為液壓缸定位控制命令，追隨理想之參考模型反應，根據模擬及實驗結果，證明本文提出之方法，具有快速學習能力及較小的追隨誤差，並可達到最佳輸出的反應速率與穩定性。

關鍵字

油壓伺服 ； 類神經網路 ； 可變排量泵 ； 電液比例閥 ； Lyapunov定理 ； 適應性控制

並列摘要

In this study, a novel adaptive control using multiple neural networks is proposed, which consist of neural controller, neural emulator and neural tuner, to control the swashplate angle of a variable displacement axial piston pump (VDAPP) with a three-way electro-hydraulic proportional valve (EHPV) and the position of an actuator with one electro-hydraulic servo valve (EHSV). The neural controller with specialized learning architecture utilizes the linear combination of error and its derivative to approximate the back propagation error (BPE) for weights update. The neural tuner adjusts the parameters of this linear combination. The nonlinear model of VDAPP and EHSV actuator were identified beforehand, and updated on-line by the neural emulator. The triple networks and the plant were integrated to an adaptive control system, which was designed by Lyapunov theorem for the guarantee of global stability. The simulation and experiment results showed that the time responses of control system with square commands of various amplitudes were stable and with fast convergent speed. The mean square error (MSE) of control process was smaller than those of the direct neural, indirect neural and neural tuner parameter controls.

並列關鍵字

Lyapunov theorem ； neural network ； hydraulic servo ； swashplate dynamics ； electro-hydraulic proportional valve ； adaptive control ； variable displacement pump

參考文獻

〔47〕Zhang, Y., Sen, P. and Hearn, G. E., “An On-Line Trained Adaptive Neural Network”, IEEE, Control Systems Magazine, Vol. 15, No. 5, pp. 67-75, 1995.

〔1〕Kallstrom, C. G., str m,, K. J. and Thorell, N. E., “Adaptive Autopilots for a Tanker,” Automatica, Vol. 15, No. 3, pp. 241-254, 1979

〔3〕Clark, D. W. and Gawthrop, P. J., “Self-Tuning Controller," IEE, Vol. 122, pp. 929-934, 1975.

〔4〕AstrOm, K. J. and Wittenmark, B., “On Self-Tuning Regulators," Automatica, Vol. 9, No. 2, pp. 184-199, 1973.

〔6〕Omatu, S. and Yoshioka, M., “Self-Tuning Neuro-PID Control and Applications,” IEEE, International Conference on Systems, Man, and Cybernetics, Computational Cybernetics and Simulation, Vol. 3,