从数据信息中对焊接的再认识

Re-understanding of Welding Process From Data Information

借助于“弧焊信息分析平台”的功能,介绍了当前在弧焊过程信息分析方法的研究与应用中取得的新进展。提出了如何用数据找问题以及在数据中发现和获得新的认知;如何根据数据对问题定位、发现数据信息中的分析“盲点”。多个焊接生产中的检测实例提供了通过分析数据实现对弧焊过程关键信息的认识途径。即:以数字化焊接生产车间的机器人生产线为例,分析了不同的焊机输出特性对焊缝产生可直接观察的质量差异以及数据监管标准实施中存在的“短板”;以铝合金板件多层焊的工艺设计为例,提出了脉冲电流焊接中对熔滴过渡时刻的检测,以及对熔滴过渡形态实现调控等工艺细节的讨论;从焊丝与导电嘴之间接触电阻的波动信息及其对电弧压降的影响,揭示了铝合金焊丝表面氧化而产生的相应电弧特征,为铝合金焊丝表面状态及其焊接过程电弧稳定性提供了一个新的分析途径。对如何从图像到数据的“可视化”信息的映射作了专门介绍,为数据信息的深度学习与图像智能识别奠定了基础。

By means of "Arc Welding Information Analysis Platform", the current progress in the research and application of arc welding process information analysis methods is introduced. It proposes how to use data to find problems and obtain new cognition in data;how to locate problems and find analytical "blind spots" according to data;and provides a way to realize the understanding of key information in the arc welding process through analysis data with multiple testing examples in welding production. That is, taking the robot production line of the digital welding production workshop as an example, the output characteristics of different welding machines can intuitively produce quality differences in the weld seam and the "short boards" in the implementation of data supervision standards;taking the process design of multilayer welding of aluminum alloy plates as an example, the detection of the transition time of the droplets in pulsed current welding, as well as the regulation of the transition amount of the droplets are discussed;the fluctuation information of the contact resistance between the welding wire and the conductive nozzle and its influence on the arc voltage drop are revealed, which reveals the corresponding arc characteristics generated by the oxidation of the surface of the aluminum alloy welding wire, which also provides a new analysis method for the surface state of the aluminum alloy welding wire and the arc stability of the welding process. The article specially introduces the mapping of "visual" information from image to data, laying the foundation for deep learning of data information and intelligent recognition of images.