Ph.D. Thesis Defense by Winncy Y. Du
Tuesday, November 30, 1999

(Dr. Steve Dickerson, advisor)

"Real-Time Robust Feedback Control Algorithms for Vibratory Part Feeding"

Parts feeding has been a prominent problem in automatic assembly.  More than 50% of total manufacturing cost is assembly, and up to 30% of assembly cost is parts feeding.  The major problem of existing parts feeders is that they are not flexible.  If the shapes or the sizes of the parts are changed, most of or the entire feeder has to be redesigned and rebuilt.  This not only delays the manufacturing process, but also results in material waste and recycling cost.  A novel idea, simply using a vibrating flat plate to feed various parts, has been conceived at Georgia Tech.  If the feasibility of such a feeder can be verified through theoretical analyses, simulation, and experiment, it would benefit automatic assembly.  A vibratory plate feeder has high flexibility in feeding various parts, completely controllability of partsí motion, simple construction, and low cost.

This dissertation focuses on a detailed study of the modelling and control aspects of the new Vibratory Feeder (VF).  The VF involves the 3D motions: translations in the X and Y, and rotation about Z.  The Coulomb friction is the only force to cause the partís motion. The link between the controlled plate and the moving part is indirect, nonlinear, and highly uncertain because of the friction.  The system was modeled as a 12th order system.  A three-point-contact model for the part was established to simplify the system model and the calculations. Two Modified Pulse Width Modulation (MPWM) signals were developed for the VF control. The duty cycle and the period of the MPWM signals are the only control variables to control both plateís and partís motion so that they can move in a desired way. The corresponding control laws and the 3D control algorithms were developed for the two MPWM control signals.  The Synchronous Period and the Dimension-Independent control strategies are developed to solve the different control period and the angular-linear motion coupling issues in the 3D control. The efficient path of the part and the control parameters were designed to achieve the best feeding rate.  Both the open-loop and the closed-loop simulations have verified the feasibility of the system model, the MPWM signals, and the control algorithms.  The 1D prototype and experiment further examine the validity of the proposed algorithms.