摘要
为了提取压缩机喘振发作时表现出的非线性特性,引入了一种新的喘振特征提取方法.首先对原始信号进行多元统计分析,构造高维特征空间,然后利用局部切空间排列的流形学习方法提取出一维主流形,进而通过主流形几何结构的变化来反映系统的非线性变化.分析结果表明,与相关积分方法相比,该方法可以提前1 s识别出喘振特征,并且能够降低误报率,因此在喘振监测中具有良好的应用前景.
To extract the nonlinear characteristic of compressor surge, a new monitoring method based on manifold learning is proposed. High dimension space of raw pressure signals is constructed by multivariate statistical analysis, then the local target space alignment （LTSA） algorithm is employed for extracting one dimension principal manifold, to reflect the non-linear dynamic characteristies of compressor system. Experimental results and industrial measurements show that this approach, compared with the correlation integral method, forecasts surge about ls in advance with lower false prediction rate.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2009年第7期44-48,共5页
Journal of Xi'an Jiaotong University
基金
国家"863计划"资助项目(2007AA04Z432)
关键词
喘振
流形学习
局部切空间排列
surge
manifold learning
local tangent space alignment
