MODELLING OF THE IMPACT OF MELT SURFACE DYNAMICS ON A PHOTODIODE MONITORING SIGNAL IN LASER WELDING Paper 526 Peter Norman, Hans Engström, Alexander F.H. Kaplan Division of Manufacturing Systems Engineering, Luleå University of Technology SE-971 87 Luleå, Sweden Abstract Today merely a few monitoring systems for in- process detection of laser welding defects are commercially available. Despite a trend towards cameras, the industrially most powerful concept is still a photodiode with optical filter, measuring thermal emissions from the melt surface and from the plasma or laser beam reflections. The monitoring rule for each application is identified empirically through temporal correlations between dynamic signal changes and obtained welding defects. The mechanisms behind are widely not understood. Thus the method does not provide a systematic estimation of success for identifying a certain welding defect. The here presented research approach studies the context between welding defects, the physical mechanisms behind, particularly the dynamics of melt pool, plasma and temperature field, and the photodiode signal. Numerical simulation results of the thermal emissions from the weld pool and keyhole dynamics and their non-linear conversion into a voltage signal are presented. An essential outcome is the sensitivity of the sensor signal to certain sub-mechanisms of the motion for judging under which conditions they can be monitored. Various results are discussed for simplified hypothetic cases as well as for observed weld pool dynamics of practical relevance. Introduction Laser welding is a highly complex process governed by the interaction of several optical, thermodynamic and fluidmechanic mechanisms. As the resulting quality of the weld is essential for industrial applications but sensitive to the process parameters, improved process understanding as well as in-process monitoring is desired. Welding defects like lack of fusion, lack of penetration, cold laps, undercuts, holes, pores or cracks have to be avoided. Commercial process monitoring systems either detect thermal emissions or laser reflections from the dynamic welding process by photodiodes or by cameras. Surveys on process monitoring during laser processing are given by several authors [1-3]. The melt or plasma dynamics is often in direct or indirect context with the generation of welding defects and can therefore be suitable for their on-line detection. Photodiode detection has the advantage of delivering a robust signal as a function of time, easy to handle, but the reduction to a single voltage signal corresponds to loss of information due to its integrative nature (in terms of space, wavelength). In contrast, monitoring by a camera provides an image of the process with lot of information, but requires complex signal analysis, difficult to realise in industry in a robust, universal and reliable manner. The present study focuses on analysis of the generation of commercial [4] photodiode signals by the welding process, in particular on its mathematical prediction and analysis by correlating surface motions to the signal dynamics. The main objective of the present research is improved understanding of the context between process dynamics and signal changes in order to judge the probability for correlations. Improved understanding of the process physics can be achieved by experimental observation of the welding process with high speed imaging. An in- house developed X-ray imaging system combined with a high speed camera was developed by Matsunawa and Katayama [5]. This set-up visualizes the keyhole and melt pool motion. With these tools it was explained how the keyhole dynamics, liquid motion in the melt pool and the plasma affect the welding result, e.g. pore formation. For the explanation of spatter or humping, high speed imaging of the melt surface can be more suitable (with spectral narrow illumination for eliminating the plasma radiation). In addition, to understand the physics and mechanisms behind the laser welding process, mathematical modelling and simulation can be applied. Even with today’s computer powers, the laser welding process has been too complex to be