Selective Laser Melting (SLM) is an additive manufacturing technique that uses a moving laser source to melt and solidify an area of a powder bed by scanning over the area with a laser beam, thus, fabricating a solid part layer by layer. Although SLM can print complex geometries that are difficult to achieve by machining, quality and repeatability of a printed part are still two challenges that must be addressed. These challenges arise due to the complex physical transformation of the metal powder and the lack of mature process control schemes.
Researchers often use optical sensors for the feedback control of SLM processes. Since fast motion of laser beams may not allow real time feedback control given the large quantity of data to be processed, some researchers have applied layer-to-layer control, i.e., collecting data during the fabrication of an entire layer and then updating process parameter profiles for the next layer. In this paper, by specifying the input and output as laser power and peak temperature, a parameter adaption model is used to estimate the unknown input-output model, which is difficult to evaluate analytically due to complex process physics. In addition, laser paths in SLM usually vary to create more isotropic parts than can be achieved with constant laser paths. The variation of laser paths results in varying local thermal histories. To handle the common situation where laser paths are varying from layer to layer, a switched model is designed and trained by a Parameter Adaption Algorithm (PAA). In a simulation of an overhang part with a constant cross-section and varying laser paths, the switched adaptive model shows the ability to achieve a desired output profile and also better performance if more switched gains are utilized.