Ph.D. Thesis Summary

Automated Cell Supervisor Control Product Disassembly

Michael Bailey-Van Kuren

September 1996

A global waste storage problem exists due to reduced storage facilities. Shorter product life cycles coupled with "throw away" consumption create greater quantities of consumer good waste. High volumes of used products for material recycling or reuse require automated disassembly systems. A key component of the automated disassembly system is addressed through the development of a new work cell supervisory controller.

A new automated disassembly system description and model are presented. The model was constructed through state spaces describing a automated work cell and a used product. The model incorporated automated equipment constraints and product configuration constraints to define allowable transitions in the system state space.

The controller design is presented. The controller directs automated resources to disassemble a used product in order to retrieve some set of target components while maximizing throughput and minimizing system error. Furthermore, the used product may contain missing, replacement, or jammed components. System states are observed through sensor feedback. Optimality conditions are implemented through a dynamic component search. Search time is reduced by searching in the component state space with topologically defined neighborhoods. Further search logic is presented to handle jammed or replacement components. A feasible sensing technique is presented based on machine vision.

Controller behavior was explored through a disassembly simulation based on a real product and a real work cell. Simulated disassembly of a single use camera and a personal computer pointed out the differences between closed and open architecture used products, respectively. The results show throughput gains for the designed controller for differing products, disassembly goals, and component uncertainties.

The major contributions are: 1) the first system model for automated disassembly, 2) a new controller design for automated disassembly with product uncertainty, 3) a new automated disassembly simulation that uses real product and process parameters, and 4) the automated disassembly system behavior under different product configuration uncertainties (missing, replaced, jammed, or added components) is identified. Minor contributions include: 1) the concept of disassembly group technology reduce the search process when the expected component is not identified on the used product and 2) a fast component identification algorithm utilizing machine vision.