【摘要】 目的针对液压振动台加速度正弦波波形复现、加速度扫频两类试验项目,利用反馈误差学习(FEL)控制策略解决传统三状态控制策略不能解决的跟踪精度低、鲁棒性差的问题.方法利用ADALINE神经网络、归一化LMS算法、结合正弦函数的特点,设计FEL控制策略中的前馈神经网络.结果对于正弦波和扫频信号参考输入,即使在阀控缸系统参数发生变化的情况下,所设计控制器仍然在幅值和相位上达到很高的跟踪精度.结论基于FEL思想设计的幅相控制器跟踪精度高,结构简单,运算量小,能满足对实时性要求很高的液压振动台控制系统的要求,也有利于基于DSPs的嵌入式实现.更多还原

【Abstract】 For the two test items of hydraulic vibration generator system,acceleration sinusoidal signal waveform replication,acceleration sweep,Feedback-Error-Learning(FEL) control strategy is utilized to solve the weakness of tracking accuracy and robustness which traditional three variable control strategy can’t do.Feedforward neural networks of FEL control strategy was designed,using ADALINE neural networks,normalized LMS,plus characteristics of sinusoidal function.For sinusoidal and sweep reference input,even in cases that valve-controlled cylinder system parameters changed,the designed controller still achieved high tracking accuracy in amplitude and in phase.Furthermore,it has the advantage of simple construct and smaller amount of computation,which can satisfy the demand of control system that requires higher real-time capability and DSPs solution.更多还原