高子博　郭海欣　李天雷　张莉　周冲　张晨曦

摘 要：电阻点焊技术因其低价高效性在汽车行业广泛应用于车身各部件的连接上，点焊的质量直接影响到整个汽车的品质。本文设计了一个点焊质量评分预测模型，通过传感器采集焊接过程中的动态电流、电压等信号，在与焊接质量评分进行相关性分析后选取可以表征焊接质量的特征量，基于机器学习算法建立点焊焊接质量预测模型并进行误差分析。最后选用随机森林算法模型，得到焊接评分预测值，结果表明预测值与真实值的误差可以控制在2%以内，预测较为准确。

关键词：电阻点焊 数据采集 机器学习 质量预测

Research on Signal Acquisition and Welding Quality Prediction in Spot Welding Process

Gao Zibo Guo Haixin Li Tianlei Zhang Li Zhou Chong Zhang Chenxi

Abstract：Resistance spot welding technology is widely used in the automotive industry for the connection of various parts of the car body because of its low cost and high efficiency， and the quality of spot welding directly affects the quality of the entire car. In this paper， a spot-welding quality score prediction model is designed， which collects dynamic current， voltage and other signals in the welding process through sensors， selects the characteristic quantities that can characterize the welding quality after correlation analysis with the welding quality score， and establishes the spot-welding quality prediction model based on the machine learning algorithm and analyzes the error. Finally， the random forest algorithm model is selected to obtain the predicted value of the welding score， and the results show that the error between the predicted value and the real value can be controlled within 2%， and the prediction is more accurate.

Key words：resistance spot welding， data acquisition， machine learning， quality prediction

1 前言

在汽车的生产制造过程中，车身及各零部件之间的连接通常采用电阻点焊的方式。据相关统计表明，点焊在车身结构连接方式中的占比大于90%，因此车身强度、安全性和使用寿命很大程度上取决于焊点的质量，准确把控汽车车身焊点质量对保证整车质量意义重大。

电阻点焊的基本原理是把装配成搭接接头的焊件压紧在两电极之间，利用电阻热熔化母材金属，形成焊点。但由于其形成过程一直处于封闭状态，无法直接观测，大多数汽车厂商在生产中都是通过抽样来进行破坏性检测或无损检测，破坏性检测主要通过测量焊点的拉剪强度或焊核直径来评价焊点的质量，无损检测大多通过目测或超声波检测来确定焊点质量。这些方法都使得焊点质量检测效率极低，生产成本提高，并且无法全面了解车身所有焊点的质量。……