时变AR模型的一种定阶方法

自回归（AR）模型的定阶对其分析结果的准确性、效率和可靠性具有重要影响。对时变AR模型的定阶问题进行了深入研究，提出了一种基于采样频率以及分析频率之比的定阶方法。对一质量随时间变化的悬臂梁进行了有限元建模，运用计算所得到的振动响应建立了梁的时变AR模型，利用所提方法进行了该模型的定阶，采用递推最小二乘法对梁的模态频率进行了估计。对一质量时变的悬臂梁进行了试验，对采集的振动响应信号建立了时变AR模型并识别了其振动模态频率。模拟计算和试验结果都表明，提出的时变AR模型定阶方法是有效且可行的,且识别算法具有一定的抗噪性。

A new method for order selection of a time-varying AR model

The order selection of an AR model has an important influence on accuracy,efficiency and reliability of its analysis result.Here,a deep study on the order selection of a time-varying AR model was conducted and a new method for order selection based on the ratio of sampling frequency to analysis frequency was presented.A model of a cantilever beam with a time-varying mass was obtained with FEM.The calculated vibration response was used to set up a time-varying AR model and its order was selected with the proposed method.The recursive least square method was adopted to estimate the beam’s modal frequencies.An experiment was performed using a cantilever beam with a time-varying mass.With its measured vibration response signals,a time-varying AR model was set up to estimate its modal frequencies.The simulation and the test results indicated that the proposed method for the order selection of a time-varying AR model is effective and feasible,and the method is capable of anti-noise to some extent.