The International Journal of Advanced Manufacturing Technology https://doi.org/10.1007/s00170-019-04279-9 ORIGINAL ARTICLE Model predictive control of laser metal deposition Yangbo Liu 1 · Liuping Wang 2 · Milan Brandt 2 Received: 24 April 2019 / Accepted: 6 August 2019 © Springer-Verlag London Ltd., part of Springer Nature 2019 Abstract Laser metal deposition (LMD) is one of the efficient processes in laser additive manufacturing (LAM) systems that uses metallic materials to produce metallic components additively. One of the remaining challenges for this type of systems is its product quality control. This paper proposes automatic control of melt pool size during the process of laser metal deposition. Using a low-cost near-infrared monochrome (NIRM) camera, the melt pool size is measured from grey melt pool image and is modelled by a first-order transfer function with time delay where the laser power is chosen to be the manipulated variable. Experimental data are used to identify the first-order plus delay model, which is then converted to a non-minimal state space realisation that accurately captures the time delay in discrete time with all state variables being measurable. A model predictive controller (MPC) is designed and implemented for controlling the melt pool size with the state space model. Experimental results are conducted on the platform TruLaser Cell 7020 from Trumpf, showing that the melt pool size has been successfully controlled to their desired specifications and the product quality measured by the variations of the brick height has been significantly improved. Keywords Laser metal deposition · Product quality control · Model predictive control · Melt pool size Abbreviations LMD Laser metal deposition LAM Laser additive manufacturing NIRM Near-infrared monochrome MPC Model predictive controller RLS Recursive least squares PID Proportional integral and derivative GPC Generalised predictive control CMOS Complementary metal oxide semiconductor FoV Field of view FPS Frames per second LP-MPS The relationship between the laser power and the melt pool size CNC Computer numerical control LPM Low power model HPM High power model FAM Frequency average model ISA Industry standard architecture  Yangbo Liu yangbo liu@hust.edu.cn 1 Wuhan National Laboratory for Optoelectronics, 1037 Luoyu Road, Wuhan, 430074, People’s Republic of China 2 School of Engineering, RMIT, Melbourne, VIC, 3001, Australia 1 Introduction LMD is a type of laser additive manufacturing (LAM) that uses metallic materials to produce metallic components additively and it possesses most of the advantages of LAM, such as fast prototyping and design validation, a short production period, low dilution rate, low physical distortion, better surface properties and flexible building capabilities [1–3] (building lattice structure or fully dense components or other special applications). LMD also gains a number of benefits from using metal materials, including that LMD can repair metallic parts directly, and that its products can gain special properties by combining different types of metal powders. As a result, LMD is widely adapted in metallic part repair, refurbishing, prototype fabrication and surface coating [4–6]. Similar to other LAM technologies, the procedures of LMD start from tool path planning. The first step is to generate a tool path that can be either defined by a program or generated based on a sliced CAD model. The second step is to set the process parameters for a specific LMD task. After placing a substrate on the base of the LMD machine, powder and laser are delivered simultaneously as the laser head moves along the tool path. In the meantime, a shielding gas, for example, Argon, is conveyed to the processing area to isolate it from oxygen. Finally, few machining procedures