基于自适应线性元件的周期误差分离研究

Research on periodic error separate based on adaptive linear element

动态测量误差信号通常为多分量的非平稳信号,可能包含非周期性趋势项误差、周期性误差、随机误差等分量。为了更精确地修正误差,需将各误差分量分离出来。用神经网络自适应线性元件方法分解动态测量误差信号的周期性成分,通过一个动态测试仿真系统进行验证,并与EMD方法分解结果对比,得出神经网络自适应线性元件方法在分离周期性成分(特别是频率较大)时比EMD方法更为准确,说明了该方法在动态误差分解中的实用性和优越性。

Dynamic measurement error signal is usually multi-component non-stationary signal,which contains the aperiodic trend errors,periodic errors and random errors.In order to get a more precise result of correcting the errors,the components of the errors should be separated from each other.The neural network self-adaptive linear element method is used to decompose the periodic components of the dynamic measurement error signals,which is tested by a simulation system of dynamic measurement.Compared with the result from the EMD method,get the conclusion that the result based on neural network self-adaptive linear element method is more precise（in particular when the frequency is high） than the result based on EMD method in isolating the periodic components,and the practicability and advantage of the method are given.