微电阻点焊条件下的质量控制问题研究

发布时间：2019-05-07 23:16

【摘要】：微电阻点焊是近十几年发展起来的一项微连接技术,待焊工件更薄、焊接方式更多样化、电极无水冷却、焊接参数精度要求更高,不能仅将常规电阻点焊的焊接规范“缩放”至微尺度。考虑到日益增长的微连接社会需求,研究微电阻点焊的质量控制问题,对于制定合理的焊接工艺规范,实现焊接过程的自动化和智能化有着重要的意义。本文通过使用动态电阻、电极电压信号以及焊接工艺参数,结合试验设计、回归分析、神经网络等技术,对焊点质量进行预测、分类及优化,具体研究内容和取得的成果如下：(1)基于微电阻点焊中的动态电阻曲线,结合神经网络和动态电阻曲线特征量提出了相应质量预测模型。此时动态电阻的获取较常规条件下方便得多,通过分析动态电阻变化与熔核生长之间的关系,提取了若干特征量,使用与焊点质量相关性较强的特征量,结合多元线性回归和神经网络技术,评价了焊点质量,该方法具有较高的质量预测准确率。(2)考虑模型预测的实时性,提出了基于动态电阻和主成分的焊点质量监测模型。通过对离散动态电阻直接进行主成分分析,选取若干方差贡献率较大的主成分作为回归分析和神经网络模型的输入,用于预测焊点质量。主成分思想的引入在很大程度上缩短了数据处理时间,该方法能更容易地实现在线实时质量监测。(3)基于微电阻点焊电极电压曲线变化特征,结合电极电压曲线和神经网络提出了相应质量监测模型。采用方便提取的部分电极电压曲线特征量,建立了可用于预测焊点失效载荷的广义回归神经网络模型,具备较强的在线实时输出能力。并尝试通过离散Hopfield神经网络识别电极电压曲线模式,区分焊点质量。两种方法均能较准确地评价焊点质量级别,可为实现微电阻点焊质量控制提供新的研究思路。(4)针对微电阻点焊质量指标变化范围有限的特点,进行了考虑多质量指标的焊接参数优化。对多个焊点质量指标进行主成分分析,以综合主成分作为焊点综合质量指标,以焊接参数作为自变量,建立了多元线性回归模型,确立了最优焊接参数组合,焊点综合质量得到明显提高。
[Abstract]:Micro-resistive spot welding is a micro-joint technology developed in the last ten years. The workpiece to be welded is thinner, the welding methods are more diversified, the electrode is cooled without water, and the precision of welding parameters is higher. You cannot just "scale" the conventional resistance spot welding specification to a micro-scale. Considering the increasing social demand of micro-resistance spot welding, the research on quality control of micro-resistance spot welding is of great significance for establishing reasonable welding procedure specification and realizing automation and intelligence of welding process. In this paper, the quality of solder joint is predicted, classified and optimized by using dynamic resistance, electrode voltage signal and welding process parameters, combined with experimental design, regression analysis, neural network and so on. The concrete research contents and achievements are as follows: (1) based on the dynamic resistance curve of micro-resistance spot welding, the corresponding quality prediction model is put forward based on neural network and dynamic resistance curve characteristic quantity. At this time, the dynamic resistance is much more convenient than under the conventional conditions. By analyzing the relationship between dynamic resistance change and nugget growth, a number of characteristic quantities are extracted, and the characteristic quantities with strong correlation with solder joint quality are used, and the relationship between dynamic resistance change and nugget growth is analyzed. The quality of solder joint is evaluated by combining multiple linear regression and neural network. (2) considering the real time of model prediction, a quality monitoring model of solder joint based on dynamic resistance and principal component is proposed. By direct principal component analysis (PCA) for discrete dynamic resistance, several principal components with large variance contribution are selected as inputs of regression analysis and neural network model to predict solder joint quality. The introduction of principal component reduces the data processing time to a great extent, and this method can realize on-line real-time quality monitoring more easily. (3) based on the variation characteristics of microresistance spot welding electrode voltage curve, this method can be used to realize on-line real-time quality monitoring more easily. Combined with electrode voltage curve and neural network, the corresponding quality monitoring model is put forward. The generalized regression neural network model, which can be used to predict the failure load of solder joint, is established by using the characteristic quantity of partial electrode voltage curve which is easy to be extracted. The model has strong on-line real-time output capability. The discrete Hopfield neural network is used to identify the electrode voltage curve pattern and distinguish the solder joint quality. Both methods can accurately evaluate the quality level of solder joint, which can provide a new research idea for the quality control of micro resistance spot welding. (4) in view of the limited variation range of micro resistance spot welding quality index, The optimization of welding parameters considering multi-quality index was carried out. The multi-component linear regression model is established and the optimal combination of welding parameters is established by principal component analysis (PCA), taking comprehensive principal component as comprehensive quality index of solder joint and welding parameter as independent variable. The comprehensive quality of solder joint has been improved obviously.
【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TG453.9

【相似文献】