Abstract:  To grasp the internal welding status of welds in real time during welding processes, an on-line acquisition system of arc sound signals was constructed. The correlation analysis between arc sound signal characteristics and sidewall fusion states was carried out under the conditions that the torch swing center were in different positions. Arc acoustic features with strong correlation to side wall fusion state were extracted from time domain and frequency domain respectively. In order to further improve the effectiveness of the fusion state prediction, a support vector regression model for sidewall fusion state recognition was constructed by arc acoustic feature parameters. To reduce the impacts of non-features and improve the prediction accuracy of the model, genetic algorithm was used to optimize the model parameter. After parameter optimization, the recognition rate of the model is  as 93.33%, which realizes the effective recognition of the fusion states of the narrow gap sidewalls.

Key words:  , narrow gap welding, arc sound, sidewall fusion, support vector machine(SVM), pulse gas metal arc welding

摘要： 为在焊接过程中实时了解焊缝内部的焊接状况，构建了电弧声信号实时采集系统。在焊枪摆动中心处于不同位置的情况下，进行了电弧声信号特征与侧壁熔合状态的相关性分析。分别从时域与频域中提取了与侧壁熔合状态相关性较强的电弧声特征。为进一步提高熔合状态预测的有效性，采用电弧声特征参量构建了支持向量回归的侧壁熔合状态识别模型。为减小不良特征对识别模型的影响，显著提高模型的识别精度，采用遗传算法进行了参数寻优。参数寻优后模型的总体识别率达93.33%，实现了窄间隙侧壁熔合状态的有效识别。 

关键词: 窄间隙焊, 电弧声, 侧壁熔合, 支持向量机, 脉冲熔化极气体保护焊