Weld joint forming prediction method based on complementary two-channel convolutional neural network

• Main Authors: GUO HENGTONG, LEI ZHENGLONG
• Format: Patent
• Language: Chinese; English
• Published: 24.12.2021
• Subjects: CALCULATING; CLADDING OR PLATING BY SOLDERING OR WELDING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; CUTTING BY APPLYING HEAT LOCALLY, e.g. FLAME CUTTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; MACHINE TOOLS; METAL-WORKING NOT OTHERWISE PROVIDED FOR; PERFORMING OPERATIONS; PHYSICS; SOLDERING OR UNSOLDERING; TRANSPORTING; WELDING; WORKING BY LASER BEAM

The invention discloses a weld joint forming prediction method based on a complementary dual-channel convolutional neural network. Compared with a BP neural network, the convolutional neural network has the biggest characteristic that the extraction of molten pool characteristics is not needed, but the extraction of molten pool characteristic quantity is automatically carried out through a constructed multi-layer convolution kernel; the convolutional neural network takes the whole molten pool image as the input of the model, so that the time consumed for extracting the feature quantity of the molten pool is saved. Meanwhile, the loss of molten pool image information is avoided; compared with a common two-channel convolutional neural network laser welding seam forming prediction method, the method adopts two convolution modules to extract shallow layer features of the molten pool image to extract edge lines of the molten pool, and adopts a two-channel strategy, so that the obtained molten pool image features a