摘要
采用时变参数模型对航天器某时段非平稳随机振动信号(NSRVS)进行建模分析,利用过程神经元网络(PNN)求解模型的时变参数并以此确定信号的时变自功率谱密度。计算结果表明:由PNN估计的NSRVS时变参数与自相关Levinson法估计的该参数基本一致,但前者建模物理意义明确,和传统的方法相比避免了计算信号的自相关矩阵,减少了存储空间,提高了频率分辨率和计算速度。
A time-varying parameter model is established to analyze the nonstationary random vibration signal(NSRVS) of a spacecraft in a certain time period.Then the process neural network(PNN) is utilized to obtain the time-varying parameters of this model.Moreover,the time-varying parameters of the NSRVS are applied to determine the time-varying auto-spectral density of the signal.Computation results show that the parameters estimated using the PNN are very close to those estimated using the auto-correlation Levinson method.However,the PNN method is visualized and convenient,it can avoid calculating the self-correlation matrix of the signal,reduce the storage space,and also increase the resolution of the freguency spectrum and the computing speed.
出处
《振动与冲击》
EI
CSCD
北大核心
2008年第1期12-15,29,共5页
Journal of Vibration and Shock
关键词
NSRVS
时变参数模型
功率谱
过程神经网络
振动环境
nonstationary random vibration signal(NSRVS),time-varying parameter model,power spectrum,process neural network(PNN),vibration environment
