In this paper, relevant twenty features were extracted from the dynamic resistance signal of the welding operation and subsequent mathematical models incorporating the qualifying features were developed to investigate the relationship among dynamic resistance, welding quality and welding parameters based on the stepwise regression method. The results of the analysis of variance indicated the peak point location $t_1$, the initial resistance value $r_0$, and the first slope $k_1$ had little correlation with welding process parameters. The welding current is the most significant factor affecting the dynamic resistance signal, followed by welding time and electrode pressure. The welding current has significant effects on all the extracted features, while the electrode pressure only affects seven features. The model utilizing the features of $r_{\mathit{mean}}$, $r_e$, $r_{\alpha }$, $r_{\beta }$ and $r_{\mathit{med}}$ was developed to predict the welding quality. $r_{\mathit{mean}}$ is the most significant factor in the model whose maximum relative error is less than 10%.

本文从焊接操作的动态电阻信号中提取了相关的二十个特征，并基于逐步回归方法，建立了包含合格特征的后续数学模型，以研究动态电阻，焊接质量和焊接参数之间的关系。方差分析的结果表明了峰值点的位置${Ť}_{\mathrm{1个}}$，初始电阻值 ${\mathrm{[R}}_0$，所述第一斜率${ķ}_{\mathrm{1个}}$与焊接工艺参数几乎没有关系。焊接电流是影响动态电阻信号的最重要因素，其次是焊接时间和电极压力。焊接电流对所有提取的特征都有重要影响，而电极压力仅影响七个特征。利用以下特征的模型${\mathrm{[R}}_{\mathit{意思}}$， ${\mathrm{[R}}_{Ë}$， ${\mathrm{[R}}_{\alpha }$， ${\mathrm{[R}}_{\beta }$ 和 ${\mathrm{[R}}_{\mathit{中}}$ 被开发来预测焊接质量。 ${\mathrm{[R}}_{\mathit{意思}}$ 是模型中最重要的因素，其最大相对误差小于10％。