In DLD,  a laser beam is used to melt locally deposited material powder and create geometries layer by layer. This complex process involves fast cycles of melting and solidification of metal powder. Our research aims to build a graph-based neurosymbolic modeling framework for the process, and develop a control scheme accordingly. The scheme will manipulate in-process parameters to ultimately improve geometry accuracy and mechanical properties of the parts, and expedite build time.

In hybrid manufacturing, in addition to additively creating a solid part, a subtractive or deformative process is performed in situ on the solid material. These additional processes help to improve microstructural and surface properties of the finished parts but add on further complexity.  Our research here seeks to control the microstructural evolution in an additive + incremental forming hybrid manufacturing setting using graph-based modeling. We also seek to achieve well-controlled interfaces in multi-material hybrid manufacturing.  

Digitalized manufacturing processes, such as AM and incremental forming, open a major path planning question: In what sequence should the manufacturing tools carry out the processes? We seek to use deep generative models to find (sub)optimal tool trajectories that maximize control authority.  We also seek to implement control co-design of fixed process parameters.

In Thermal Energy Storage (TES) systems, energy is stored in form of sensible or latent heat, or by thermochemical reactions. TES devices are employed in a variety of applications ranging from grid-level and household storage for electricity to thermal management systems in aircrafts . Our interest in these systems involve control-oriented graph-based models for state of charge estimation as well as  control co-design.