Modeling and Control of a Microscale Robotic Deposition Manufacturing System

Microscale Robotic Deposition (μ-RD) is a direct-write- manufacturing technique for solid freeform fabrication of parts with feature sizes on the microscale (0.1m - 100m). The technique places stringent demands on the tracking performance of a robotic positioning system used in the process. To meet these requirements, a dual Cartesian stage positioning robot is presented. The research presented in this thesis focuses on the modeling of and development of control strategies for the first, or coarse, positioning stage. Robotic axes driven by brushless DC motors are subject to deterministic disturbances from motor cogging, amplifier-motor force ripple, and both sliding and pre-sliding friction. These disturbances are nonlinear, difficult to isolate, and can make precision control objectives difficult to achieve, especially at low velocities. Physics based models based on unknown parameters or nonlinear functions of a single variable are developed for each of these phenomenon. A novel technique is presented for obtaining mappings, without the use of external measurement equipment, of these disturbing phenomena which can be used in a feedback linearization approach. The technique is applied to the positioning system and linear models of the augmented feedback linearized system are identified using a swept sine approach. Simple lead-lag controllers are designed for the linearized system. Iterative Learning Control (ILC) methods are identified as being particularly well suited for the μ-RD system and the repetitive nature of the process. Two ILC methods are developed for the positioning system - the fixed Q-filter ILC and the recent Adaptive Q-filter Time-Frequency ILC (TFILC). In both methods, a novel online learning algorithm is presented. Using the fixed Q-filter ILC, prototype lattice parts were fabricated. Contour error plots show that, using the developed controllers, the coarse positioning system is able to achieve very high performance on the order of the feedback resolution for the prototype lattices. However, examination of the actual fabricated lattice parts shows that the part quality is not always consistent with the positioning system performance. These results indicate that ink extrusion and adhesion behavior play a very important role in the part quality. Some preliminary models and compensation techniques for the ink behavior are presented and suggested for future work