基于最小二乘支持向量机的船舶水下焊接质量在线监测

On-line Monitoring of Submerged Weld Quality of Marines Based on Least Squares Support Vector Machines

以径向基函数(radial basis function,RBF)为核函数,将最小二乘支持向量机(least squares support vector machines,LSSVM)预测模型应用于船舶水下焊接质量在线监测。提出了一种自适应优化方法确定该模型中的可调超参数和核宽度参数,并建立了实时显示和报警系统。实验结果表明,该方法预测误差较小,建模耗时少,适合于船舶水下焊接质量在线监测。

Taking the radial basis function （RBF） as kernel function, the least squares support vector machines （LSSVM） prediction model is used for on-line monitor of the marine submerged welding quality. A method is proposed to adaptively optimize the LSSVM regularization parameter and the kernel width parameter. In addition, display and alarm systems are established. With higher predictive accuracy and less modeling time, the experimental results show that this method ig suitable for on-line monitoring of submerged welding quality of marines