基于聚类分析的航天继电器多余物检测方法研究

以往的航天继电器多余物检测算法不能有效的去除可动组件固有振动信号的干扰,难以保证判别结果的准确率。依据固有振动信号的周期性和微粒信号的随机性,提出了一种基于聚类分析的多余物检测方法。运用三门限端点检测算法,识别出信号脉冲,从而选取脉冲中心位置、脉冲起止时刻和脉冲个数3个特征参数。采用基于改进的k-平均聚类分析法对脉冲序列进行分类,计算其标准差作为是否存在多余物的主判别依据,并对信号脉冲序列数据集相位和次数表示于二维平面内,以其密度分布作为辅助判别依据。试验验证表明,该方法可以有效判别航天继电器是否存在多余物微粒,并对其它密封电子元器件和电子装置的多余物检测具有一定的借鉴意义。

Remainder detection method for aerospace relays based on cluster analysis

Previous algorithm of remainder detection cannot effectively remove the vibration signal interference caused by the internal movable components for aerospace relay, so it is difficult to ensure the accuracy of identification. In this paper, a new method of remainder recognition based on cluster analysis is proposed, according to the periodicity of component signal and randomness of particle signal. The signal pulses were extracted through the start-end point detection method with three thresholds. The center, the start-end time and the number of pulse were regarded as characteristic parameters. Pulse sequence was classified base on improved k-means cluster analysis, and the standard deviation of different classification of pulse was taken as the main criteria of remainder particles inside. The phase and number of pulse sequence data set was presented in a two-dimensional plane, and the phase density as the auxiliary criteria. The test result shows that the presented algorithm is effective to recognize remainder for aerospace relay, and the conclusion is of great value for research on remainder detection for seal electronic components and equipments.