基于受限波尔兹曼机的GMAW管道打底焊的熔透预测方法

摘要

In the background of high development of modern industry,welding is one of the important technology methods in modern manufacturing field.The welding quality directly affects the quality of weldment.RBM (restricted Boltzmann machine)is a machine learning model.It has the prediction ability with high-accuracy when to be set various proper learning parameters and after to be trained by the training sample.By carrying out image processing on the GMAW molten pool image obtained with visual sensing system,extracting detailed parameters of molten pool as the input of RBMtogether with the welding current,the state of welding,welding penetration and burn-through,are predicted. Experimental result demonstrates that it is able to achieve the goal of real-time monitoring of welding quality within the acceptable range.%在现代工业高速发展的背景下，焊接是现代制造领域的重要工艺方法之一，焊接质量的高低直接影响到工件的质量。受限波尔兹曼机RBM（Restricted Boltzmann Machine）作为一种机器学习模型，设置合适的各项学习参数，经训练样本训练后具有高精度的预测能力。通过对视觉传感系统获得的GMAW熔池图像进行图像处理，提取详细的熔池参数，与焊接电流共同作为RBM的输入，预测焊接状态：焊透与焊穿。实验结果表明，在可接受的范围内，能够实现焊接质量实时监控的目标。