电阻点焊熔核尺寸的RBF网络模型

以接头质量在线监控为目的,对电阻点焊电极电压和焊接电流等信号进行了时域特征分析,研究了电压、电流信号的周波幅值、峰值、有效值以及接头等效电阻、加热功率的估计方法.利用周波参数时间序列构造RBF神经网络输入向量,进行了点焊接头熔核尺寸预测.结果表明,采用归一化处理后的电流和电压有效值周波序列联合构造输入向量,经样本训练能有效地预测接头熔核尺寸,熔核直径平均验证误差为5.50%,熔核高度的平均验证误差为3.83%,比单独采用动态电阻、加热功率等参数具有更好的精度.

RBF network model of nugget sizes in resistance spot welding

For on-line quality monitoring of joint, the time-domain characteristics are analyzed for the signals of electrode voltage and welding current in resistance spot welding process. The methods for estimation of cycle amplitude, peak value,and RMS of the welding voltage and current as well as the equivalent resistance of the joint and the heating power are investigated.RBF neural network is used to predict nugget sizes of resistance spot welder, in which the input vectors are constructed by time sequences of cycle parameters. The result indicates that,when the RMS entire cycle sequences of current and voltage having been normalized in resistance spot welding are used as input space, the nugget sizes can be predicted effectively after sample training,the mean proved errors of the nugget diameter and height being 5.50% and (3.83%,) respectively. It has better precision by using this method than those of dynamic resistance and power, solely.