Laser Directional Energy Deposition Area Calculation Method of Full Convolution Neural Network

Zhengxiong Li, et al.

Abstract

The invention provides a method for calculating the laser directional energy deposition area of a full convolution neural network. The method comprises the steps of collecting a plurality of laser directional energy deposition area images, manually marking each molten pool label, and further constructing a training set; building a full convolution neural network, inputting the images in the training set into the neural network, predicting the existing value of a molten pool, further combining with the marked construction loss function, and optimizing to Nash balance to obtain the optimized neural network; collecting an image to be detected, inputting the image into an optimized neural network, and judging whether the image has a molten pool area; if the molten pool exists, graying the image to obtain a gray level image, and further carrying out binarization to obtain a binary image; and obtaining a molten pool contour pixel coordinate set of the image by a topological structure analysis method for the binary image, and further calculating the area of the molten pool. The invention can identify the molten pool in the laser directional energy deposition process and calculate the area of the molten pool, and is beneficial to the regulation and control of the manufacturing process.

More information can be found here