All-position direct energy deposition defect control method and device based on deep learning

Abstract

The invention discloses an all-position direct energy deposition defect control method and device based on deep learning, in the all-position direct energy deposition process, a CCD camera is installed on the side face of a deposition head and collects a molten pool side face image on line, and the molten pool side face image is labeled according to the state of a molten pool; building a deep neural network model, and performing training and hyper-parameter optimization on the deep neural network model by using the data set; in the all-position direct energy deposition process, a molten pool side image collected online by a CCD camera is used as the input of a deep neural network model, the state of a molten pool in the metal all-position direct energy deposition process is recognized and diagnosed, and a closed-loop controller controls the state of the molten pool according to the size ratio of the molten pool or the deviation between a calculated value and a set value of the area of the molten pool. And the material feeding speed or heat input is adjusted, so that the hump defect and the falling defect in the metal all-position direct energy deposition process are inhibited.

Images