Quality detection and control during laser cutting progress with coaxial visual monitoring

Abstract

A system of coaxial visual detection and closed-loop control is developed in laser cutting to automatically optimize cutting speed to obtain best cutting quality. The sparks behavior is observed under different cutting speeds by side visual monitoring. It is proved that the maximum number of pixels of various radiation brightness ranges of the coaxially detected sparks corresponds to the optimal cutting speed, represented by the lowest bottom roughness on cut face. By coaxial visual monitoring of cutting front and sparks, the standard deviations of shape parameters of cutting front and the average length of certain brightness range of sparks are, respectively, used as the signals for cutting defects and the bottom roughness. The results show that the optimal cutting speed for free defect and low bottom roughness could be determined and kept on-line by the developed system.

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

“…The usual monitoring methods are cameras or photodiodes to measure the optical, primary or secondary radiation from the cut kerf. In addition, some elaborate approaches like using a fiber Fabry-Pérot cavity microphone have been demonstrated [ 3 ]. To use such sensors as an industrial product, the sensor must be able to detect cut interruptions independent from the cutting direction, which is not feasible by below-bed sensors, as shown in [ 4 , 5 ].…”

“…Additionally, by using photodiodes [ 8 ], the burr height during laser cutting is calculated from the standard deviation of the photodiode current. From camera images of flame cutting with a CO 2 laser, it is already possible to calculate the roughness, striation angle and the burr formation [ 3 , 9 ]. In detail, a NIR camera with a 40-Hz sampling rate was used, and the quality was calculated by the size of the hot process zone and the size of its circumscribed rectangle.…”

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

“…Related works employ photodiodes [7], [8], to transduce the cut image from an optical signal to an electrical one. More recently, camera based monitoring systems are used, which guarantee a much more high quantity of information of the cut, [9], [10], [11], Fig. 2.…”

Analysis of Photodiode Monitoring in Laser Cutting

“…Duflou et al [20] proposed as control variables the monitored NIR images, the laser power, cutting velocity, assist gas pressure, and the duty cycle of the laser beam. Wen et al [21] included sparking behavior as a variable to monitor. They proved that the optimal cutting speed is ensured when a maximum number of pixels of various radiation brightness ranges were obtained.…”