将人工神经网络(ANN)技术引入CO2短路过渡焊领域,利用PC机实时采集焊接过程电流、电压信号,运用统计方法从电弧电压、焊接电流信号数据中提取出6个统计参数作为神经网络的输入量,以质量评估参数W作为网络的输出量,建立了统计参数和质量评估参数之间的关系模型。通过试验证明,在适当的学习训练后,该模型对焊接质量评估参数W的估测误差在10%以内。 Artificial neural network (ANN) technology is introduced into the field of CO2 short-circuit transition welding. The current and voltage signals of welding process are collected in real time by PC, and six statistical parameters are extracted from the arc voltage and welding current signal data as the input of neural network The quality evaluation parameter W is taken as the output of the network, and the relationship model between the statistical parameter and the quality evaluation parameter is established. The test proves that after the appropriate learning training, the error of the model in estimating the welding quality W is less than 10%.