基于深度学习的超窄间隙焊接质量评估方法  

An Evaluation Method of Ultra-narrow Gap Welding Quality Based on Deep Learning

作　　者：张爱华[1,2,3] 白忠领 马晶[1,2,3] 牛万才 朱亮 ZHANG Aihua;BAI Zhongling;MA Jing;NIU Wancai;ZHU Liang(College of Electrical and Information Engineering,Lanzhou University of Technology,Lanzhou 730050,China;Key Laboratory of Gansu Advanced Control for Industrial Processes,Lanzhou 730050,China;National Demonstration Center for Experimental and Control Engineering Education,Lanzhou University of Technology,Lanzhou 730050,China;State Key Laboratory of Advanced Processing and Recycling of Ferrous Metals,Lanzhou University of Technology,Lanzhou 730050,China)

机构地区：[1]兰州理工大学电气工程与信息工程学院,甘肃兰州730050 [2]甘肃省工业过程先进控制重点实验室,甘肃兰州730050 [3]兰州理工大学电气与控制工程国家级实验教学示范中心,甘肃兰州730050 [4]兰州理工大学省部共建有色金属先进加工与再利用国家重点实验室,甘肃兰州730050

出　　处：《热加工工艺》2020年第19期122-126,132,共6页Hot Working Technology

基　　金：国家自然科学基金项目(61866021);甘肃省科技计划项目(18YF1WA068);兰州市人才创新创业项目(2016-RC-72)。

摘　　要：在超窄间隙焊接过程中,焊接质量的无损评估对焊接质量的在线评估及预报有重要的意义。本文结合焊剂片约束电弧超窄间隙焊接方法的特点及焊接工艺,选取了22个影响焊接接头质量的特征参数作为神经网络的输入,其中基于经验选定了9个特征参数,其余13个特征参数通过对焊接过程中焊接电流信号与电弧电压信号进行时域分析提取获得。利用TensorFlow框架建立了基于深度神经网络的超窄间隙焊接接头质量评估模型,对焊接过程中的焊接接头质量进行无损评估。试验结果表明,该模型能较准确地评估出超窄间隙焊接质量,准确率达92%。In the process of ultra-narrow gap welding, the non-destructive evaluation of welding quality is of great significance to the online evaluation and prediction of welding quality. Combined with the characteristics of flux sheet constrained arc ultra-narrow gap welding method and welding process, 22 characteristic parameters affecting the quality of welded joints were selected as the input of neural network. Among them, 9 characteristic parameters were selected based on the experience, and the remaining 13 characteristic parameters were selected by time domain analysis of arc voltage and welding current signals in the welding process. Making use of TensorFlow framework, a quality evaluation model of ultra-narrow gap welded joints based on deep neural network was established to realize non-destructive evaluation for the quality of welded joints during the welding process. The experimental results show that the model can exactly evaluate the ultra-narrow gap welding quality with an accuracy rate of 92%.

关 键 词：超窄间隙焊接 深度神经网络 焊接质量 无损评估 

分 类 号：TG444[金属学及工艺—焊接] TG441.7