The broad uses of laser welding in various industrial applications such as shipbuilding, automotive production and battery manufacturing, result from its capabilities of high productivity, flexibility and effectiveness1. However, the complex nature of laser-material interaction requires additional measures in order to reach the high-quality standards of the goods produced. Therefore, continuous process monitoring in laser welding is crucial to achieve reliable mass production and high-quality products at once. Camera-based process monitoring offers great advantages compared to one-dimensional observation techniques. The spatial resolution enables the monitoring of several process characteristics simultaneously, which leads to a more detailed description of the current process state2. In the last few years, we proposed a coaxially integrated camera system with external illumination. Process images taken by this system typically show the keyhole area, the weld pool, but also areas of solidified weld and areas of the blank sheet3. To automate image evaluation with respect to the recognition of aforementioned areas, we propose a convolutional neural network architecture to perform pixel-wise image classification4. In this paper, we investigate the influence of multiple hyper-parameters required for the network architecture in use, but also the amount of data that is necessary for high segmentation accuracies. In a second step, the outcome of the network is used to detect process deviations in laser welding image data using supervised machine learning. With the help of the Random Forest algorithm, assessment of the extracted process characteristics with respect to prediction accuracy takes place. Based on the information of the segmented image data, further investigations are carried out into the possibility of predicting individual process parameters such as laser power, welding speed and focus size simultaneously.
Process observation in 3D printing of metals currently is one of the central challenges. Many companies strive to employ this additive manufacturing process in their production chains in order to gain competitive advantages through added flexibility in product design and embedded features. The new degrees of freedom are accompanied with the challenge to manufacture every detail of the product to the predefined specifications. Products with filigree internal structures for example require a perfect build to deliver the performance that was designed into these structures.
Melting conditions determine properties such as grain structure and density of the finished part before it is sent to post processing steps. Monitoring of such melting conditions is still a challenge where the use of photodiodes, pyrometry and camera systems contribute to an overall picture that might identify errors or deviations during the build process. Additional considerations must be made to decide if these sensors are applied coaxially or from a lateral perspective. Furthermore, setting parameters of focal plane array (FPA) sensors are discussed and events that are seen in the machine vision image are compared against the pyrometry data.
The resume of the experiments suggests the application of multiple sensors to the selective laser melting process (SLM) as they jointly contribute to an identification of events. These events need to be understood in order to establish cause effect relationships in the future.
In additive manufacturing, the quality of products can be traced by observation of process variables track by track and
layer by layer. The stacking of layer wise information can be used to consolidate the entire build up history of a product
thus leading to a truly three dimensional quality histogram. The first step that is necessary to achieve such a quality
histogram is the acquisition of process measurands that are related to product quality.
Successful acquisition of measurements for thermal radiation has been reported in several publications. The authors of
such papers report the detection of changes in boundary conditions of the process by observing the thermal radiation of
the process. It has been reported that for example a change in laser power has an influence on the thermal emission and
that different readings are received for processing a thin powder layer on a solid work piece compared to scanning pure
powder in the situation of an overhang structure. A correlation to the underlying reason for the increase in thermal
radiation however is mostly related to the experimental setup rather than to in process measurements.
This report demonstrates an approach of acquiring and combining synchronous measurements of different physical
properties of the process. The coaxial observation system used in the experiments enables the synchronous acquisition of
measurements of the thermal emission and the acquisition of images that visualize the surface of the powder bed in the
vicinity of the interaction zone. The images are used to monitor the motion of powder particles as they are influenced by
the melting process. This amount of particle motion is then correlated to areas of different powder thicknesses. The
combination of this information with excessive readings in thermal emission classifies the event to be a situation of
noncritical deviation of thermal emission. In fact, this combination of extracted features establishes a first key criterion
for an unequivocal event mapping.
Laser cladding processing has been used in different industries to improve the surface properties or to reconstruct damaged pieces. In order to cover areas considerably larger than the diameter of the laser beam, successive partially overlapping tracks are deposited. With no control over the process variables this conduces to an increase of the temperature, which could decrease mechanical properties of the laser cladded material. Commonly, the process is monitored and controlled by a PC using cameras, but this control suffers from a lack of speed caused by the image processing step. The aim of this work is to design and develop a FPGA-based laser cladding control system. This system is intended to modify the laser beam power according to the melt pool width, which is measured using a CMOS camera. All the control and monitoring tasks are carried out by a FPGA, taking advantage of its abundance of resources and speed of operation. The robustness of the image processing algorithm is assessed, as well as the control system performance. Laser power is decreased as substrate temperature increases, thus maintaining a constant clad width. This FPGA-based control system is integrated in an adaptive laser cladding system, which also includes an adaptive optical system that will control the laser focus distance on the fly. The whole system will constitute an efficient instrument for part repair with complex geometries and coating selective surfaces. This will be a significant step forward into the total industrial implementation of an automated industrial laser cladding process.
This presentation deals with a camera based seam tracking system for laser materials processing. The digital high speed camera records interaction point and illuminated work piece surface. The camera system is coaxially integrated into the laser beam path. The aim is to observe interaction point and joint gap in one image for a closed loop control of the welding process. Especially for the joint gap observation a new image processing method is used. Basic idea is to detect a difference between the textures of the surface of the two work pieces to be welded together instead of looking for a nearly invisible narrow line imaged by the joint gap. The texture based analysis of the work piece surface is more robust and less affected by varying illumination conditions than conventional grey scale image processing. This technique of image processing gives in some cases the opportunity for real zero gap seam tracking. In a condensed view economic benefits are simultaneous laser and seam tracking for self-calibrating laser welding applications without special seam pre preparation for seam tracking.
Production of laser beam welded tailored blanks requires both, high quality processing as well as a quality assurance by reliable monitoring systems for each welded part. Actual quality monitoring systems for tailored blank applications make use of different sensors for seam tracking, seam shape detection and process control. The suitable process diagnostic device is a plasma sensor which detects the optical emission of the weld plasma. Experimental evidence show that the reliability of the seam quality prediction can significantly be improved by using a camera system with a two-dimensional spatial resolution instead of an integrating plasma detector. The improvement is achieved by exploiting the information provided by the spatially distributed intensity of the plasma emission. In particular, by coaxial arrangement of the camera with respect to the laser beam axis, the direction of observation allows to detect significant process characteristics. Based upon these results a coaxial process control system was developed that can be adapted for different laser materials processing applications like welding, cutting and surface treatment. The system consists of a high speed camera mounted directly at the welding head. The optical path of the camera goes coaxial with the laser beam path through the focusing optics. The camera images taken from the process are analyzed using image processing algorithms. The algorithms are chosen depending on the type of application to be monitored. In the case of welding tailored blanks the system can monitor a full penetration of the workpiece, deviations from the desired welding path, seam width, stability of the capillary shape and defects of the seam caused by spatter and ejection of molten material. The camera system offers the ability to perform simultaneously different quality monitoring tasks like determination of seam and capillary shape, seam tracking and process control. Thus the number of sensors required for quality monitoring is reduced to one single system.