Robust supervision and condition based control for single-unit resource allocation systems

Shengyong Wang, Purdue University

Abstract

A flexible manufacturing system is an integrated computerized system of resources and material handling devices that is capable of processing simultaneously a variety of part types. The underlying workflow logic, the resource allocation structure, and the reliability of the system resources constitute various aspects of research for the system. We first look into deadlock and blocking problems in resource allocation systems with unreliable system resources. Almost all of the work to date focuses on systems without unreliable resources. In this work, several supervisory control policies are developed to ensure failures or repairs of any unreliable resources do not cause blocking to propagate through a system and stall unfailed portions of the system. Thus continuous production of part types not requiring failed resources is realized. This work differs from other research in that we allow distribution of failure dependent parts along their respective routes to gain better production rate on failure dependent parts, instead of using an "absorbing policy" to push all failure dependent parts into a "neighborhood". We then conduct a simulation study on the performance of different supervisory controllers on systems with multiple unreliable resources. We identify four system ratios which can be used to describe more elaborate system structures. Based on these ratios and several factors, we determine which of the compared three controllers, RO, NHC+BA and RO4 are most suitably employed for the production of FD, NFD and total parts. We also verify these implementation guidelines by conducting a simulation study of other systems. The verification results substantiate the proposed guidelines. We finally incorporate degradation signals of unreliable resources into the supervisory control of the flexible manufacturing systems. A Markov model capturing the degradation status of an unreliable resource is developed. Based on this model, we establish a control policy for systems with a single unreliable resource. Generally, the condition based policy is more permissible than our previous more restrictive policies for systems with unreliable resources since more unrestricted states are allowed when the unreliable resource is unlikely to fail.

Degree

Ph.D.

Advisors

Lawley, Purdue University.

Subject Area

Industrial engineering

COinS