基于独立分量分析的切削声发射源信号分离

Source Separation of Cutting AE Signal Based on ICA

针对切削声发射(Acoustic Emission,AE)信号的多目标状态源并行分离问题和同频干扰源分离问题,引入独立分量分析(Independent Component Analysis,ICA)技术作为研究工具,用刀具破损、切屑折断和环境噪声三个AE源的线性混合模拟切削AE信号,尝试用Fast ICA算法分离目标状态源。结果表明:实现了各目标状态源的并行分离,相对误差小于10%;目标状态源的同频干扰不影响基于独立性的ICA分离。最后,针对分离结果的鉴别排序问题进行了初步探讨。

The Independent Component Analysis （ICA） technique was introduced as a tool for simultaneously separating from cutting Acoustic Emission （AE） signal multi condition sources with common frequency band. Cutting AE signals were simulated by linearly combining AE sources of tool breakage, chip fracture and environmental noise. It was attempted to separate condition sources with Fast ICA algorithm. The results show that ICA is capable of estimating multi condition sources with errors less than 10%, and that the problem of common frequency band does not affect ICA solvability. Finally, further investigation was made with regard to identification of separated sources.