In-situ monitoring of laser metal deposition for additive manufacturing

Laser metal deposition (LMD) is widely used in the aerospace, medical implant, and other industry sectors due to its extraordinary capability of net-shape building of parts with complex geometries, repair, coating, and refurbishment. As a member of the additive manufacturing (AM) family of processes, the LMD fabrication process proceed in a layer-by-layer manner. The feed material (in the form of powder or wire) is caught within the melt pool, which is created by the laser beam and solidified as deposition. However, LMD-fabricated parts still face quality issues, such as geometrical inaccuracy, porosity, cracks, and other types of defects, due to the complex physics involved, which include simultaneous heat transfer, fluid dynamics, and phase changes. In the meanwhile, the production efficiency of metal AM is deficient (despite smaller material wastage) relative to conventional manufacturing methods. If one attempts to improve production efficiency by increasing the LMD system’s power source, more unstable melt pool is created leading to further difficulties in achieving high part quality. To understand the foundations of part quality issues and further quality control, it is essential to integrate multiple sensors into the high-power, high-build-up-rate LMD system to monitor the deposition process. The features derived from raw monitoring data (process signatures) can be used to link the process parameters with part quality. Multiple mathematical models must be established to describe in detail the correlations between process parameters, signatures, features, and part quality. These will then assist in understanding the physics of the LMD process and underpin the design a feedback control system to improve LMD-fabricated part quality and general process reliability. Therefore, this PhD project aims to reveal the mechanisms underlying LMD and to develop an improved, robot-assisted LMD system operating at higher power by integrating different sensors to achieve out-of-the-box fabrication for faster production of larger parts. The important correlations between process parameters, signatures, features, and part quality are established via statistical methods (including data-driven methods). Moreover, the statistical model is deepened with data from a mechanistic simulation model to close the gap in the lack of scientific knowledge within solely statistical modelling methods.