【作者】 赵明；

【导师】 高晶波；

【作者基本信息】 哈尔滨工业大学 ， 一般力学与力学基础， 2010， 硕士

【摘要】 基于物体出水的模态参数辨识,主要考虑两个方面的问题:一是由于附加水质量逐渐减少,系统频率是随时间变化的,是时变系统辨识问题;二是采集到的信号时间短,是短时信号辨识问题。因此本文在总结了线性时变结构模态参数识别方法的基础上,主要研究了适合短时信号的时变参数辨识方法,主要的研究内容如下:首先,研究了经典的Prony辨识方法对时不变系统结构参数的辨识过程,并在不同噪声强度和采样时间的情况下,通过此方法对输出响应信号的模态频率与阻尼比识别精度的数值仿真,研究了采样时间和噪声对模态参数识别精度的影响。其次,基于Prony辨识系统参数所需样本小的特点,将其推广为能够辨识线性时变结构参数的分段Prony方法,同时对该方法进行了仿真,并与分段AR方法作了比较,辨识结果显示分段Prony方法的识别精度要优于分段AR方法,但也发现其对快变参数的辨识精度不理想。最后,将模型系数通过一组基函数加权和拟合,建立了时变Prony方法,该方法能快速的跟踪时变参数,不仅能很好的辨识出随时间缓慢变化的参数,对快速变化的参数同样有很好的识别精度,但是其对噪声比较敏感,需要大幅度提高模型阶数来降低噪声的影响。接下来通过分析六种常用基函数对时变Prony法识别精度的影响,发现基于DCT基函数的该方法,不仅计算量相对较少,需要的数据长度也小,而且辨识精度也高于基于其它基函数的时变Prony法,通过对一个线性时变系统结构辨识的仿真,验证了基于DCT基的时变Prony方法的有效性。同时对一个悬臂梁装置的实验数据,使用时变Prony法进行了分析,结果表明该方法基于短时信号的辨识是有效的。更多还原

【Abstract】 Modal parameters identification of objects exiting water need to solve two problems: the first is the time-varying feature, this is because the added mass of water decreasing, the system frequency is time-varying, and the second is the sampling signal is very short. Therefore this paper main studies the varying parameter identification method based on short-time signals. The main research work and conclusions are as follows:Firstly, the process of the classical Prony method base on the line time-invariant system has been discussed. It has been studied about relation of the sampling time, noise and modal parameter identification accuracy by some numerical simulations at different noise level and sampling time.Secondly, since the Prony identification method just need a small quantity of sample, So that segmented Prony method is establish to identify the linear time-varying structural parameters, the method was also simulated, and compared with the segmented AR method, results show that the identification accuracy of segmented Prony method is superior to segmented AR method, But also found that method is not satisfactory for identifying the fast varying parameters.Finally, The model coefficients are fitted by a set of basic functions, and a time-varying Prony method is establish, This method can fast track the time-varying parameters, Can not only identify good for parameters change slowly, also has good identification accuracy on rapidly changing parameters, but its more sensitive to noise, need to substantially improve the model order to reduce the impact of noise. Then through analysis the time-varying Prony method identification accuracy with six commonly used basic function, found that the method with DCT basic function, not only need little calculation, require small data length, but also have a higher identification accuracy than the time-varying Prony method with other basic functions. The time-varying Prony method with DCT basic function is proved to be effective for time-varying parameters identification by simulate a linear time-varying system, and convincible for short-time signal’s identification by analyze the test data of the experiment facility of cantilever beam.更多还原