基于熔池运动特征的选区激光熔融过程状态检测方法

Selective laser melting process state monitoring method based on motion feature of melt pool

选区激光熔融是一种极具发展前景的金属增材制造技术,由于加工过程容易产生各种缺陷,过程监控对于产品质量的控制显得尤为重要。熔池变化特征是最有效的监测手段之一,但是传统的熔池几何特征缺乏与不同熔化状态的准确对应关系。针对该问题,根据熔池特点提出一种新的熔池运动特征用于描述运动熔池的变化规律。首先利用激光扫描位置数据确定熔池的移动方向及质心,然后通过设定3个不同尺度的阈值来捕捉飞溅,并结合连通分量分析方法提取熔池区域和飞溅,进一步消除熔池感兴趣区域中的飞溅。接着将熔池质心到熔池边界轮廓的距离按顺时针方向展开,获得一个36维的矢量特征用于描述运动的熔池。最后,利用k均值聚类算法分别对同一工艺参数和不同工艺参数下获取的熔池进行了聚类分析,并与传统的熔池几何特征进行了对比。实验结果表明,提取的36维熔池运动特征可以较好地区分熔池移动方向及不同的熔化状态,为选区激光熔融过程在线监控提供了一种新的思路。

Selective Laser Melting(SLM)is a highly promising metal additive manufacturing technology.Due to various defects existing in SLM process,the process monitoring is particularly important for product quality control.The variation features of melt pool are one of the most effective monitoring methods.However,the traditional geometric features of melt pool lack accurate correspondences with different melting states.Aiming at this problem,a new motion feature of melt pool was proposed to describe the regularity of variation of the moving melt pool.The moving direction and melt pool centroid were determined by laser scanning position data.Then spatters were captured by setting three thresholds of different scales.After that,the melt pool and spatter were extracted by combining the connected component analysis method to further eliminate the spatter in the Region of Interest(ROI)of melt pool.Subsequently,the distance between the melt pool centroid and the boundary of melt pool was expanded clockwise,and a 36-dimensional feature vector was obtained to describe the moving melt pool.The k-means clustering algorithm was used to cluster the melt pool images obtained under the same process parameters and different process parameters,which were compared with the geometric feature of traditional melt pool.The research results showed that the extracted 36-dimensional motion feature of melt pool could distinguish the moving direction and melted states of melt pool well.This research provided a new way for online monitoring of SLM process.