基于响应的运载火箭时变结构模态参数辨识方法研究

The performances and features of launch vehicles predominate the success and quality of space launch missions. The structural dynamic characteristics of launch vehicles, which are kinds of important performances, directly influence the design, control and other aspects of dynamics. Estimating the modal parameters based on the responses is an effective way to acquire the real characteristics of structural dynamics for launch vehicles in operation. Based on the practical needs of launch vehicle engineering and considering the theoretical advances of the identification of time-varying dynamic systems, the methods of response-based modal parameter estimation of time-varying structures will be deeply studied in this project. Focusing on the key scientific issues for launch vehicles, this project will carry out researches on: 1) by introducing the advanced mathematical tool, to develop the new method of parameter estimation based the support vector regression machine for the sake of improving the estimation quality in the case of the short response data and the complex noise environment; 2) to eliminate the time-shift errors by using the more complete model of time series; 3) through expanding the classic frequency-domain methods to the time-frequency domain, to develop the enhanced methods in time-frequency domain for the estimation of highly damped structures and the estimation of operational mode shapes; 4) considering the poly-reference and the multiple-output, to improve the estimation of weak modes. The research products of this project will supply the necessary theoretical basis for the practicality of the response-based modal parameter estimation of launch vehicles, and further theoretically support the improvement of the design and control of dynamics for launch vehicles.

运载火箭的性能品质决定了航天发射任务的成败与质量。结构动力学特性作为重要的性能品质，直接影响运载火箭动力学设计、控制等方面。基于响应辨识模态参数，可更为真实地获取运载火箭工作状态下的结构动力学特性。本项目面向运载火箭的工程实际需求，结合时变系统动力学辨识的理论前沿，深入开展基于响应的时变结构模态参数辨识方法研究。项目围绕运载火箭应用方面的关键科学问题展开研究：1）引入先进的数学工具，提出基于支持向量回归机的参数估计方法，提高方法在短数据和复杂噪声下的辨识质量；2）构建更为完备的前后向时间序列模型，消除低采样率造成的时移误差；3）将经典频域法拓展到时频域，发展针对高阻尼结构辨识和振型辨识的参数化时频域方法；4）在模型和方法中考虑多参考和多输出，提高弱模态辨识能力。研究成果将为基于响应的运载火箭模态参数辨识的实用化提供必要的理论基础，进而为提高动力学设计与控制的质量提供理论支持。

基于响应辨识模态参数，可更为真实地获取运载火箭工作状态下的结构动力学特性。本项目面向运载火箭的工程实际需求，结合时变系统动力学辨识的理论前沿，深入开展了基于响应的时变结构模态参数辨识方法研究，取得了较为丰富的研究成果：（1）引入支持向量机、脊回归、最小二乘支持向量机等先进的数学工具，提出若干性能好、精度高的时变结构模态参数辨识方法，较现有的最小二乘、最大似然方法，在短数据和复杂噪声下的辨识能力更佳；（2）将经典频域法拓展到时频域，发展针对高阻尼结构辨识的多参考右矩阵分式多项式最小二乘和最大似然辨识方法，方法在高阻尼结构和密集模态方面具有优势；（3）在模态频率和阻尼比已知的情况下，提出了时频域时变结构模态振型最小二乘辨识；（4）开发了变质量充液实验系统，为模拟飞行状态运载火箭结构动力学系统提供了有效途径。研究成果将为基于响应的运载火箭模态参数辨识的实用化提供必要的理论基础，进而为提高动力学设计与控制的质量提供理论支持。