松脱件撞击波形检测算法

Detection Algorithm of Loose Parts Impact Waveform

文中针对核反应设施松动件的检测,以碰撞波形检测为目标,提出了一种基于机器学习的松动件撞击波形检测方法。该方法以核反应设施中的传感器数据为输入基础,利用数据滤波、特征降维、机器学习等组合算法,实现松动件振动波形识别。实验结果表明,对数据进行滤波、降维后,利用GBDT算法,能有效识别不同类型的振动波形。此外,通过t-SNE算法对样本进行可视化后,发现经滤波和特征降维后的数据具有很好的聚类特征,有助于分类算法进一步完成分类识别。

Aiming at the detection of loose parts in nuclear reaction facilities,with collision waveform detection as the goal,this paper proposes a method for detecting the impact waveform of loose parts based on machine learning.The method takes the sensor data in the nuclear reaction facility as the input basis,and uses a combination algorithm such as data filtering,feature dimensionality reduction,and machine learning to realize the vibration waveform recognition of loose parts.The experimental results show that after filtering and dimensionality reduction of the data,the GBDT algorithm can effectively identify different types of vibration waveforms.In addition,after visualizing the samples through the t-SNE algorithm,it is found that the filtered and feature dimensionality reduction data has good clustering characteristics,which helps the classification algorithm to further complete the classification and recognition.