Laser Point Cloud-based Inspection Of The Appearance Of Flat Welds

With the development of science and technology,welding technology is widely used in aerospace,ships,bridges,machinery manufacturing and other industrial fields,and the quality of welding is one of the key factors in determining the reliability and service life of welded components,each weld to do a detailed inspection will cost a lot of time and financial resources,so the initial screening of the weld before the detection of those with obvious surface defects in advance to eliminate the weld for the enterprise to save a lot of time costs.At present,most of this initial screening work by the workers on site through visual inspection to carry out,but this way too dependent on personal experience and did not form a unified standard,in the dark environment will seriously affect the workers’ ability to judge.With the rapid development of the economy,there will be more inspection tasks waiting to be handled,however,the way through the manual initial screening can no longer meet today’s rapid development and increasingly standardized inspection needs.Therefore,a more intelligent and standardized appearance inspection method has become the future development trend.In this paper,two types of flat welds,linear and curved,are studied to improve inspection efficiency by evaluating the quality of welds through visual inspection and thereby screening unqualified welds.Firstly,laser scanning is used to obtain the point cloud information.In order to get the original point cloud data which is consistent with the real weld appearance,the laser sensor is equipped with the threeaxis motion platform.The weld seam is scanned by connecting the computer via PCI bus and with the use of the scanning system developed by the sensor manufacturer.Secondly,according to the imaging characteristics of the outer contour of the weld seam,a combination of nonlinear and curvature algorithm is used to optimize the point cloud data to reduce the interference of noise to the subsequent analysis.According to the forming characteristics of the weld seam,the point cloud of the slice of the weld seam along the trajectory direction is extracted by the combination of PCA and local normal vector algorithm,and the main direction of the slice is projected according to the PCA algorithm.The projected two-dimensional data are fitted with a polynomial of order 9 and a distance threshold K.The difference between the actual contour line and the fitted contour line on the yaxis is compared to determine whether the contour line has defects such as galling,tumor,porosity,etc.,and the defect area is restored by iterative method.Finally,for the convenience of the operator,a visual operating system is designed for planar welds with built-in modules for filtering,alignment,segmentation,tangent,fitting and rating,which can quickly calculate the characteristic parameters of the cross-sectional contour line at any position of the weld,such as weld toe,fusion width,residual height,bus bar angle and other parameters.By comparing the experimental test data with the actual measurement data,it can be found that the maximum error of the calculated contour parameters can be controlled within 0.3mm under normal measurement environment,and the maximum error of the angle deformation on both sides can be controlled within 1°.