基于集成熵KPCA的复杂机电系统状态监测方法

State monitoring of complex electromechanical system based on integrated entropy of KPCA

针对传统KPCA方法的模型参数选取对经验知识依赖程度过高、容易造成漏报和误报的缺点,提出一种基于集成熵核主成分分析的状态监测方法。该方法将传统的KPCA与信息熵结合,在高维空间用信息测度确定模型参数,用Renyi熵贡献提取核主成分,通过构造综合统计量进行状态监测。在TE过程和某企业的压缩机组系统上的仿真研究表明,所提方法较传统KPCA有更好的非线性数据处理能力和更高的故障或异常检测精度。

Aiming at the problem that the conventional Kernel Principle Component Analysis （KPCA） approaches de- pended on experiences to select parameters too much, a state monitoring method based on integrated entropy of KP- CA was proposed, which combined conventional KPCA with the information entropy to determine the parameters by information measuring in the high dimensional space. The kernel principal component was extracted with Renyi en- tropy estimator. Operating state was monitored by forming synthesis statistics variables. The simulation on the TE process and the compressor system of a enterprise were applied to prove the better capability of processing nonlinear data and the more accuracy rate on detecting the fault or abnormal of the proposed approach comparing with conven- tional methods of KPCA.