www.laser-journal.de Optical In-Process Monitoring of Direct Metal Laser Sintering (DMLS) A revolutionary technology meets automated quality inspection Thomas Grünberger and Robert Domröse Additive Manufacturing (AM) is a revolutionary technology enabling the users to design and manufacture products in a completely new way. As the AM technology is proceeding from rapid prototyping to rapid manufacturing, industries ranging from aerospace and medical (Fig. 1) to energy and automotive require detailed documentation of the process. The market demands automated in-process quality inspection. The goal is to get a better understanding of the process and assure repetitive high-end part quality. Plasmo, a worldwide leading engineering partner for quality inspection systems in the field of laser and joining technologies cooperates with EOS, the technology and market leader for the DMLS technology, and with selected industrial partners to develop a diode based in-process monitoring system. Direct Metal Laser Sintering Direct Metal Laser Sintering (DMLS) is an AM process [2] by which digital 3D design data (Fig. 2 a) are used to build up a component in layers by depositing metal material. The system starts by applying a thin layer of the powder material to the building platform (Fig. 2 b). After each layer a laser beam then fuses the powder at exactly the points defined by the computer-generated data, using a laser scanning optic (Fig. 2 c). The platform is then lowered and another layer of powder is applied (Fig. 2 d). Once again the material is fused so as to bond with the layer below at the predefined points resulting in a complex part. Thereby not only the part but also the final material is created in the process and defines the unique characteristics of this technology. Every single welding line creates a new micro segment 40 Laser Technik Journal 2/2014 Fig.1 Additive manufacturing of dental crowns and bridges, material EOS Cobalt Chrome SP2 [1] of the final part and can therefore be monitored. Stacking all monitoring information on top of each other, we can visualize a 3D model of the part quality (Fig. 2 e) [3]. Part quality, quality assurance Besides others [4, 5, 6, 7], the final part quality is heavily influenced by various factors: the powder material, exposure parameters such as laser power, scan speed and exposure-strategies, as well as the inert gas flow and temperatures in the process chamber. Potential negative phenomena in part quality can occur e.g. porosity, lack of fusion and rough surfaces. There are several strategies to enable quality assurance during the process, such as system monitoring, camera based powder bed monitoring and diode-, pyrometer- or camera-based inprocess monitoring. In the following section the technology diode-based inprocess monitoring is described. Diode based in-process monitoring – basic layout The basic layout of the system shows Fig. 3. Light emissions emitted from the process are measured by photo diodes Fig. 2 Description of DMLS technology, [3] © 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Process Monitoring gorithms need less computational effort compared to evaluations in frequency domain enabling a real time implementation. Therefore, the user gets the information during the building process and can change the settings immediately. using the plasmo® fastprocessobserver. Photodiodes can be mounted in lateral “off-axis” configuration with direct view on the process. Photodiodes can also be mounted in “on-axis” or ‘coaxial’ configuration – integrated to the optical path via a beam splitter, which allows view through the scanner, along with the laser beam. The measured diode signals are transmitted to the plasmo® in-process monitoring system where the appropriate visualizations and evaluations are performed. The quality inspection results are transferred to the machine PLC via a machine interface. A suitable wavelength region for reception of the observed light must be choosen. This is done by careful selection of photodiode types (e.g. silicon or germanium diodes have different sensitivity in the visible and near infrared spectrum) and spectral filter types. The building process is highly dynamic. Due to the high scanning speed, a sufficiently high sample rate has to be used. The hardware is capable of sampling rates up to one sample per second. Usually, the sample rate is reduced in order to reduce the data amount. Experiments have been done to determine a sample rate appropriate for the process dynamics. Nevertheless it turned out that a high sample rate should be used. The great advantage of this high sampling rate compared to camera based inprocess monitoring systems is to detect high dynamic effects like keyhole dynamics. Spatial information can be obtained by using neighboring exposure lines and layers. For developing algorithms and proving the reliability of a quality inspection system it is necessary to have NOK samples (defect population) and OK samples (baseline population). In many cases real defects in production are hardly available, so it is necessary to simulate defects. Example: Building of a cube of 7 × 7 mm base area. At the beginning nominal exposure parameters are used, later the laser power is reduced by 10% of its nominal value. The resulting part quality was inspected by computer tomography (Fig. 4 a) obtaining a NOK part containing a pore of size 0.18 mm × 0.18 mm × 0.18 mm [9]. In a first step characteristic values like the mean value of the measurement are plotted over its corresponding position at the building platform layer per layer (Fig. 4 b shows this for a layer with the given pore) in false color representation (dark red equals high mean value, dark blue equals low mean value). The defect can easily be seen by a lowered signal level, which can also be found in Fig. 4 c in the time series plot. Algorithms like short term fluctuations in combination with spatial information enable the user to detect this behavior automatically. Evaluation principles Example “Splashy” process It has been revealed successfully, that changes in process parameters like e.g. laser power, focus position and scan speed have an influence on the resulting measurements, as well as on part properties. To evaluate monitoring data automatically, analysis can be done in time, frequency and time scale domain (wavelet transformation). Algorithms in time domain known from e.g. remote laser welding – like absolute signal limits, signal dynamics and short term fluctuations [8] – are applicable to the DMLS process with slight modifications in parameters caused by higher process speeds. Time domain al- Changes in exposure parameters by the user may yield in an operating point close to stable process boundaries or even an unstable process. It is possible to detect such unstable processes like a splashy process, so the user gets the information during the building process and can change the settings immediately saving metal powder and machine time. Fig. 5 shows at the left side (green marked area) an image of an OK process and the corresponding sensor signal in time domain compared to an NOK process (Fig. 5 right side red marked area). The bottom plot shows the calculated Fourier transform of the measured sig- Example porosity © 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim Fig. 3 System layout diode based in-process monitoring nal over time. The changes in process can be detected by a more noisy signal and by upcoming higher frequencies in frequency domain. Summary and Outlook Benefits of diode based in-process monitoring Diode-based in-process monitoring systems especially using high sampling rates offer great opportunities for the analysis of fast processes like the DMLS process in time and frequency domain to guarantee constant high and repetitive part quality. Calculation and visualization of measurement characteristics lead to a better understanding of the process and enable a faster development of new and optimization of existing processes. Using spatial information from the building process enables a parameterization for different sizes of defects depending on the need of the customer. Fig. 4 Example detection of porosity, Source Fotec [9] Laser Technik Journal 2/2014 41 www.laser-journal.de The hardware of the system is independent from the processed metal powders. Only the parameterization of the OKNOK evaluation needs to be adapted individually. The system enables the user to perform process and quality documentation following international standards. Challenge Fig. 5 “Splash” process, image of process, sensor signal and analysis in frequency domain Next Steps Companies plasmo Industrietechnik GmbH Vienna, Austria plasmo is a global high-tech provider of automated quality inspection systems for the manufacturing industry. Established in 2003, it’s headquartered in Vienna and has a network of international distributors. The expertise ranges from control of welding processes, monitoring of weld seams, geometric shapes and surfaces, laser power measurement, up to tailored solutions in the field of machine vision and analysis software. A wide range of services with comprehensive engineering rounds off the portfolio. A leading position in the market, the company occupies in the real-time quality control for assembly processes. www.plasmo.eu EOS GmbH - Electro Optical Systems Founded in 1989 and headquartered in Germany, EOS is the technology and market leader for design-driven, integrated e-Manufacturing solutions for Additive Manufacturing (AM), an industrial 3D printing process. EOS offers a modular solution portfolio including systems, software, materials and material development as well as services (maintenance, training, specific application consulting and support). As an industrial manufacturing process it allows the fast and flexible production of high-end parts based on 3D CAD data at a repeatable industry level of quality. As a disruptive technology it paves the way for a paradigm shift in product design and manufacturing. It accelerates product development, offers freedom of design, optimizes part structures, and enables lattice structures as well as functional integration. As such, it creates significant competitive advantages for its customers. www.eos.info Laser Technik Journal In a next step, the system will be qualified for additional defects and further automation. Also, the intelligent combining integration of other monitoring technologies such as camera-based powder bed monitoring will be done. The presented results in porosity detection are developed in the project PAM – Powder Additive Manufacturing. It is supported by the Austrian Research Promotion Agency FFG and carried out by FOTEC Forschungs- 2/2014 References [1] EOS GmbH – Electro Optical Systems: press material – images and applications (2012). [2] A. Gebhardt: Generative Fertigungsverfahren, Hanser Verlag, München (2007). [3] EOS GmbH – Electro Optical Systems: EOS system data sheet M 280 (2012). [4] J. Kruth et al.: Part and material properties in selective laser melting of metals. Proc. 16th Int’l Symp. Electromachining (ISEM XVI), Shanghai (2010). [5] A. B. Spierings, M. Schneider, R. Eggenberger: Comparison of density measurement techniques for additive manufactured metallic parts, Rapid Prototyping J. (2011) 380–386. [6] L. E. Murr et al.: Microstructure and mechanical behavior of Ti-6Al-4V produced by rapid-layer manufacturing for biomedical applications, J. mech. behav. biomed. mater. (2009) 20–32. [7] S. Bremen, W. Meiners, A. Diatlov: Selective Laser Melting – A manufacturing technology for the future? Laser Tech. J. 9 (2012) 2, 33–38 [8] plasmo Industrietechnik GmbH, Operating manual processobserver advanced, www.plasmo.eu [9] Fotec Forschungs- und Technologietransfer GmbH, www.fotec.at [10]University of Applied Sciences Wiener Neustadt, www.fhwn.ac.at DOI: 10.1002/latj.201400026 Authors Thomas Grünberger Krailing, Germany 42 One challenge due to high sampling rates of the system is the development and implementation of a data flow model, which can handle big data (approximately 10 MByte/s are generated by the diode based system). Appropriate data compression strategies, which are already available at plasmo can be implemented. und Technologietransfer GmbH [9] together with the University of Applied Sciences Wiener Neustadt [10], TU Vienna and plasmo. Dr. Thomas Grünberger received his PhD at TU Vienna, in 1997. Since then he has been scientist at Vienna University of Technology (Institute for Electrical Control Engineering), continued as scientist a t Forschungszentrum Seibersdorf (Department Quality Assurance for Production). His actual position is CTO of plasmo Industrietechnik GmbH which he cofounded 2003 in Vienna. Dr. Grünberger is lecturer at FH Wiener Neustadt and president of Austrian Association for Thermography. Robert Domröse Robert Domröse is R&D project leader at EOS. Since joining EOS in 1999, he held various positions in the R&D department, with focus on development of hardware and systems. He is a specialist for lasers, optics, metrology and laser processing of metals. In his current projects, he is dealing with process monitoring of metal additive manufacturing processes. Robert Domröse holds a diploma in physics from University of Regensburg, Germany. Dr. Thomas Grünberger, plasmo Industrietechnik GmbH, Dresdner Str. 81-85, 1200 Vienna/Austria, E-mail: thomas.gruenberger@plasmo.eu Robert Domröse, EOS GmbH – Electro Optical Systems, Robert-Stirling-Ring 1, D-82152 Krailling E-mail: robert.domroese@eos.info © 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim