Following the higher degree of automation and adaption to Industry 4.0 standards, monitoring and controlling the manufacturing process has attracted significant attention.
Ever growing demands on increased productivity with a focus on product quality imply the need to both trace and control the manufacturing process. Specifically in high-power laser welding, where increased process speed comes in combination with requirements and demands on process quality, process traceability and process reproducibility.
The monitoring process for high-power laser welding can typically be divided into three categories – pre-process, in-process, and post-process. The in-process category mainly addresses the key-hole’s shape stability in the welding zone, where optical and acoustic detectors, as well as x-ray radiography and camera solutions have been implemented. There are various ways to integrate sensors for process monitoring. Typically, sensors are integrated into the process optics using dichroic mirrors or optical elements to extract signal light. However, adding such optical elements results in increased size and weight of the process head. Alternatively, sensors can be integrated into the laser cabinet or into a laser beam switch that are located further away from the process. This would not require any extra weight on the process head, but the process signals are attenuated by all optical elements in the optical chain, which ultimately reduces the signal quality. Another approach is using non-coaxial setups located close to the process, but these systems strongly depend on the angle of incidence of the process light and are thus inherently difficult to install and repeatability is low. Optical coherence tomography (OCT) is increasingly being used to monitor and control laser-based welding processes. However, these systems are expensive and include complex measurement setups.
Considering these alternatives, the SmartQD fibre optic cable by Coherent offers a simplified yet capable in-process welding monitoring solution. With photodiode-based process monitoring sensors and functions integrated inside the fibre connector, the SmartQD design provides a compact solution that avoids the need to add additional optical components or sensors to the beam guiding system. Furthermore, installation is simple since the sensors are pre-aligned for monitoring parallel to the optical axis, and the setup is compatible with all process head configurations with the European Automotive Standard Fiber Interface (QD).
Fibre connector with process monitoring capabilities: SmartQD
A common system setup for laser material processing consists of a laser source and a fibre to guide the laser light from the laser to a robot equipped with process optics to focus the laser light onto a workpiece.
The optics, optimised to guide the forward propagating laser light from fibre to the workpiece, also guides reflected laser light and emitted process light from the workpiece back to the fibre connector.
In such systems, the fibre connector is one of the most exposed parts to back-reflected laser light from the process and must be able to handle back-reflected power levels close to the power level of the laser source.
Therefore, integrated sensors in previous generation QD fibre connectors have mainly focused on monitoring the amount and effect of back-reflected laser light to prevent system damage by shutting down the laser in time. However, being on the optical axis, with optical elements aligned to the workpiece, this provides an excellent location for photodiode-based process monitoring. The use of photodiodes inside the connector for process monitoring showed potential in [1] and was further explored for cutting and welding applications for various materials in [2].
The back-reflected and emitted light consists of light of different wavelengths, originating from different physical processes as the laser interacts with the workpiece, see Figure 1. The laser back-reflection light comes from direct laser reflections in the processed material and will vary in intensity depending on processing conditions and processed material. As the laser power is absorbed by the workpiece its temperature rises and black-body radiation increases. If the workpiece reaches vaporisation temperatures, the immediately surrounding atmosphere will be ionised and emit luminous radiation.
The SmartQD photodiode-based sensors capture process light at three different wavelength ranges: back-reflected laser light (900-1,100 nm), near-infrared light from temperature radiation (1,200-1,600nm) and visible/UV light (300-700nm). The SmartQD comes equipped with an analogue or digital data interface with sampling frequency per sensor channel up to 2MHz for the digital version.
Process monitoring experimentation
The process monitoring capabilities of the SmartQD were tested at Coherent’s Applications Lab in Hamburg with a Coherent ARM laser. Primarily it was evaluated for anomaly detection in 0.8mm galvanised mild steel sheet welding. A total of nine reference welds were first performed to optimise laser power, welding speed and focal position, before defects were introduced.
In one experiment, a thin layer of oil was applied in between the sheets and, using the same process parameters as for the reference welds, three welding trials were executed. Results showed that the defect was clearly detected by both visible and NIR sensors showing strong deviations and increased levels, see Figure 2, where the NIR sensor is partly saturated. Looking at the weld seams, spatter and blow holes are clearly visible.
Figure 2: Process monitoring signals and corresponding weld seams with oil present in between sheets
In a further experiment, three 2mm holes were drilled on the top sheet before welding. Using the reference weld settings, the three holes could easily be detected in the signals for back-reflected light and temperature radiation, see Figure 3. The diameter of the holes corresponds to the duration of the signal deviation.
Figure 3: Process monitoring signals and corresponding weld seams with three 2 mm holes applied to the top sheet prior to welding
By optimising the dynamic range of the sensors, the results are promising for further tests on other welding geometries, material properties and types of material.
As a tool for anomaly detection the SmartQD clearly is capable as deviations from normal operation can be resolved. Furthermore, sensor signals show promise for classification usage, that is to not only detect that an anomaly has occurred but also to classify what type. A process control system should be connected to the SmartQD for data analysis of the sensor signals. That control system could be an advanced machine learning system such as SmartSense+ or a more basic system based on for example reference envelopes. Basic signal processing is possible to implement on the already existing hardware inside the SmartQD connector.
Fredrik Johansson is a Research and Development Engineer and Mats Blomqvist a Research and Development Manager at Coherent-Mölndal, Sweden
--
References
[1] Blomster O.; Blomqvist M.; Bergstrand H.; Pålsson M.: High-power fibre optic cable with integrated active sensors for live process monitoring, Proc. SPIE 8239, 82390U (2012).
[2] Belke, S.; Bode, M.; Kallage, P.; Rath W.; Aleryd S.; Johansson F.: Laser integrated process monitoring, Lasers in Manufacturing Conference (2019).