过程(计算)
材料科学
模态(人机交互)
激光功率缩放
卷积神经网络
功率(物理)
激光器
计算机科学
集合(抽象数据类型)
工艺工程
机械工程
人工智能
工程类
光学
物理
操作系统
程序设计语言
量子力学
作者
Noopur Jamnikar,Sen Liu,Craig Brice,Xiaoli Zhang
标识
DOI:10.1016/j.jmapro.2023.05.004
摘要
For wire-feed laser additive manufacturing (WLAM), the molten pool dimensional and thermal profiles are the indicators for achieving the high quality of the build, and they are comprehensive resultant effects of multiple process parameters. For the purpose of in situ quality control, the desired molten pool conditions should be achieved by adjusting the process parameters comprehensively. This paper experimentally analyzes in situ molten pool image modality and molten pool temperature modality under a set of controlled process parameters in a WLAM system. The variations in the steady-state and transient state of the molten pool are presented with respect to the change of multiple process parameters, including laser power, travel speed, wire feed rate, and hot wire power. A multi-modality convolutional neural network (CNN) architecture is developed for predicting the required process parameter to achieve a given desired molten pool condition. The results highlighted that the multi-modality CNN, which receives thermal profile as an external modality on top of the features from the molten pool image modality, has improved prediction performance compared to the image-based uni-modality CNN and standard regression modeling approach.
科研通智能强力驱动
Strongly Powered by AbleSci AI