舰用钢板的大功率激光焊接参数优化研究

Research on optimization of high power laser welding parameters of warship-used steel

针对高强低碳合金厚钢板大功率激光焊接过程的非线性、多变量耦合性、不确定性等特点,通过构建基于声信号采集的监测系统,实现舰用钢板焊接过程声音的实时采集,并提出特征向量,构建双权值神经网络(DWNN)模型,充分利用DWNN优秀的非线性拟合能力,实现大功率激光焊接多参数与声信号多特征之间非线性映射的神经网络建模。在拟合精度和迭代次数上,DWNN比径向基函数网络等传统网络更优,为高强度低合金厚钢板的大功率焊接参数的优化和控制提供了良好的基础。

For the characteristic of non-linear, muhivariable coupling and uncertainty of high-strength low-alloy steel in high pow- er laser welding process, a monitoring system based on acoustic signal acquisition was built, real-time acquisition of the sound was achieved, feature vector was extracted, and Double Weights Neural Network （DWNN） model was constructed. Excellent nonlinear fitting ability of DWNN was used to build the nonlinear mapping neural network model between multiple parameters and sound signals. DWNN has higher fitting accuracy and less iteration than traditional radial basis function network （RBF） in the same size, which provides good foundation to parameters optimization and control of high power laser welding.