Dross attachment estimation in the laser-cutting process via Convolutional Neural Networks (CNN)

Abstract

Laser cutting of metals offers the advantage of high precision and accuracy. Dross attachment, measured as the length of the re-solidified material perpendicular to the surface, has definitely the highest impact on the overall process quality. Dross attachment is commonly judged by skilled technicians that evaluate the cut quality. Process parameters are optimized to maximize the cutting speed while keeping an acceptable level of dross attachment. However, in practice, increased levels of dross may occur due to different processing conditions. In this framework, a real-time dross attachment monitoring system is desired. Within the stream of vision based monitoring systems, in this work we use high frequency images generated by a precision camera, mounted on the laser head, to capture the cutting process light emission. A CNN-based classification system is developed, where captured images are fed into the trained network with the aim of automatically recognize if a predetermined dross attachment level is exceeded. To our best knowledge, this is the first work where a CNN is used for monitoring the quality of laser cutting process via dross attachment classification.

Laser Cut Interruption Detection from Small Images by Using Convolutional Neural Network

“…This error rates were very low, especially keeping in mind the similarity of the cut interruptions in the thin sheets and cuts in the thick sheets. Additionally, compared to burr detection during laser cutting with neural networks, which archived an error rate of 8% [ 10 ], our error rate was about two orders of magnitude lower. This highlights the good results of our cut interruption detection sets.…”

“…A newer approach is to use a complex convolutional neural network to detect the burr formation during fiber laser cutting from images with 210 × 210 pixels [ 10 ], for which a burr detection accuracy of 92% has been reported. Such convolutional neural networks (CNN) are nowadays used very successfully for many image classification tasks, like face recognition and object detection [ 11 , 12 ] in medicine for cancer detection [ 13 ] or electroencephalogram (EEG) evaluations [ 14 ]; in geology for earthquake detection [ 15 ] and in many technical tasks, such as concrete crack detection [ 16 ], road crack detection [ 17 ], detecting wood veneer surface defects [ 18 ] or detecting wafer error determinations [ 19 ].…”

Simultaneous Burr and Cut Interruption Detection during Laser Cutting with Neural Networks

“…Thinner sheet thicknesses require a higher production effort per stack and are more difficult to cut because they are very flexible, and warp under the gas pressure and thermal influence. In these experiments only one sheet thickness is used, but please note that in previous publications with similar systems an adaptation of the results to other sheet thicknesses was possible with only minor additional expenses [ 5 , 13 ].…”

“…By using two cameras in [ 4 ], laser cutting with a CO 2 laser is monitored by observing the spark trajectories underneath the sheet and melt bath geometries and correlate this to the burr formation or overburning defects. A novel approach is used in [ 5 ], employing a convolutional neural network to calculate burr formation from camera images with a high accuracy of 92%. By evaluating the thermal radiation of the process zone with photodiodes [ 6 ], the burr height during fiber laser cutting can be measured from the standard deviation of a filtered photodiode signal.…”

“…In addition, many of the previous studies applied CO 2 lasers, which are often replaced nowadays by modern fiber or disk lasers, for which, in turn, fewer reports are available. To detect both failures with the same system, we chose the evaluation of camera images with neural networks, as they are able to achieve a high accuracy in detecting both cut failures [ 5 , 13 ]. The use of neural networks, especially for convolutional neural networks (CNN), has been demonstrated for various image classification purposes, such as face recognition and object detection [ 15 , 16 ], in medicine for cancer detection [ 17 ] and electroencephalogram (EEG) evaluations [ 18 ] or in geology for earthquake detection [ 19 ].…”