非平稳信号盲源分离理论及高速列车声场分解方法研究

The acoustic environment within high speed train is under the influence of several exterior noise source signals. Although the commonly methods,such as Time-Frequency Analysis, Transmission Function, Acoustic Holography, etc.，could have a good analysis about the magnitude and the spectrum characteristics of the compound noise, they couldn't separate the noise source signals to know their properties,transmission ways and the mode of joint action. There are amount of countermeasures are proposed by world-wide experts to control the high speed train noise, but being not very effective. It is necessary to propose a new acoustic analysis method. This project would decomposit the compound noise based on the blind source separation theory, which is a newly branch of signal analysis science. But the internal noise of the high speed train is nonstationary random signal, and the number of noise sources as well as the length of signals are both time-varying, the existing methods couldn't deal with the difficult problem. This project will study and propose a new blind source separation method of time-varying nonstationary signal as well as conduct scientific experiments and simulation analysis to separate the noise source signals and reveal their action mechanism of the high speed train. The achievements would be contributed to the development of the theory study in signal analysis, and enhance the design and optimization capacity in sound quality and noise control of the high speed train.

高速铁路车内声场主要由来自车厢外部的若干噪声源信号混叠而成。目前通过时频分析、传递函数、声全息等方法虽能对车内合成声场的总体强度分布、时频域结构加以分析，但不能分离出与各噪声源对应的车内噪声信号分量，从而明晰噪声源的性质、传播特性、对车内声场的作用机制。由于缺乏更有效的研究方法，尽管各国学者在高速铁路噪声控制领域开展了大量研究，实际降噪效果仍然差强人意。本项目拟基于信号处理领域新兴的盲源分离理论对车内声场进行分解，但因为高速铁路车内噪声是非平稳随机信号，并且具有时变信源数目与时变信号长度,增大了信号分析的难度，既有的盲源分离方法不适用。本项目拟研究提出非平稳时变信源的盲源分离新方法，并在理论研究、实验研究、数值模拟的基础上，应用新方法进行车内声场分解，深入揭示各噪声源对车内声场的作用机制。研究成果可以促进测试信号分析基础理论发展，也能为改善高速铁路车内声品质、开展低噪声设计与优化提供依据。

高速铁路噪声源识别与低噪声设计受到全球发达经济体重视，盲源分离技术由于具有在激振源与传输通道未知的情况下分离观测信号提取源信号的优点引起关注。本项目开展了盲源分离理论与算法研究，针对非平稳源信号的稳健高效分离问题提出了基于自适应滤波理论的全局最优信噪比盲源分离方法、基于拉氏逼近贝叶斯选择原理的时变信源数目辨识方法。开展了大量不同工况下的高速列车振动与噪声测试试验，分析了高速铁路车内声场分布、噪声来源与传播机制。建立了试验相符的车体有限元结构模型、声学空腔模型与声固耦合模型，基于声学传递向量特性开展了噪声贡献量分析，提出了高速列车降噪优化设计建议。研究成果促进了信号处理基础理论发展，为高速铁路系统减振降噪优化设计提供了研究依据。