论文部分内容阅读
提出了一种新的时变系统健康监控的损伤分类方法。将函数级数时变自回归平滑时序模型应用于时变系统的振动信号,以估计TAR/TMA参数和革新方差。这些参数是时间的函数,将它们在以特定的基函数构成的某种函数子空间上展开得到相应的投影系数组。所估计得到的TAR/TMA参数和革新方差可进一步用来计算潜在成分(LCs),将LCs用于健康评估比原来的参数更具信息。并将LCs联合并归化为数值得到特征集,输入概率神经网络(PNN)进行损伤分类。为了评价该方法,对一个时变系统进行了仿真,以各种不同的质量和刚度减少来模拟不同的损伤类别。算例表明:该方法能够在时变系统的背景下对损伤进行归类。
A new method of damage classification based on time-varying system health monitoring is proposed. Applying the function-level time-varying autoregressive smoothing time series model to the vibration signal of time-varying system to estimate TAR / TMA parameters and innovation variance. These parameters are functions of time, and they are expanded to a certain function subspace formed by a specific basis function to obtain a corresponding projection coefficient set. The estimated TAR / TMA parameters and variances can be further used to calculate potential components (LCs), and the use of LCs for health assessment is more informative than the original parameters. The LCs are combined and naturalized into numerical feature sets, and the probabilistic neural network (PNN) is used for damage classification. To evaluate this method, a time-varying system was simulated to simulate different damage categories with a variety of different mass and stiffness reductions. The example shows that this method can classify the damage in the context of time-varying system.