基于人工神经网络的电液伺服系统加速度谐波辨识

Electro-hydraulic Servo System Acceleration Harmonic Estimation Based on ANN

对于电液伺服系统,由于系统非线性因素的存在,当正弦信号输入时,系统加速度输出中出现高次谐波,使加速度信号严重谐波失真。提出了基于人工神经网络(ANN)的谐波辨识方法,该方法利用Adaline神经网络在线辨识信号中各次谐波的幅值和相位,用实际加速度输出与辨识得到的加速度信号间的误差,通过LMS算法来调整Adaline神经网络的权值,从而利用权值计算各次谐波的幅值和相位。通过大量仿真试验证实,这种方法能快速有效精确地在线辨识各次谐波信号。

Because nonlinearities occur in the electro-hydraulic servo system, when the command signal is a sinusoidal wave, the acceleration output of the system contains higher harmonics, which causes harmonic distortion of the output acceleration signal. Using artificial neural network （ANN）, a novel method for harmonic estimation was developed here. An Adaline neural network was used to on-line identify the amplitudes and phase of harmonics as well as the fundamental acceleration output. The LMS algorithm was applied to update the weights of the Adaline according to the error between the actual acceleration and the estimated acceleration to calculate the amplitudes and the phases of harmonics with the weights. The simulated results show that the method can identify harmonic signals on-line, fast, effectively and accurately.