基于改进PSO-SVM的薄壁件铆接质量检测

Quality Inspection of Thin-Walled Parts Riveting Based on Improved PSO-SVM

针对传统铆接几何公差质量检测极易造成错检、漏检等问题,提出了基于改进PSO-SVM的铆接质量检测方法。采用惯性权重自适应调整的策略,并选择合适的学习因子,有效提高了检测准确性;针对小样本提出最小二乘SVM算法,提高计算速度获得最优解;利用改进PSO优化最小二乘SVM的惩罚因子参数值和核函数参数值。并以制孔和铆接后的6061铝合金板模拟飞机薄壁件铆接样本,使用搭配远心镜头的CCD相机采集图像并建立数据集,验证了方法的有效性。

Regarding the issues of missed and false detections in traditional riveting geometric tolerance quality inspection,a riveting quality inspection method based on an improved PSO-SVM is proposed.Utilizing a strategy of self-adaptive adjustment of inertia weight and selecting appropriate learning factors can effectively improve detection accuracy.For small sample sizes,the least squares SVM algorithm is proposed to enhance computational speed and obtain the optimum solution.The improved PSO algorithm is employed to optimize the penalty factor parameter values and kernel function parameter values of the least squares SVM.Using 6061 aluminum alloy plates,which simulate aircraft thin-walled riveted samples after punching and riveting,images are obtained with a CCD camera equipped with a centrifugal lens and a dataset is established,thus verifying the effectiveness of the method.