Deadlock avoidance in automated manufacturing systems with unreliable resource and flexible process sequencing

Widodo Sulistyono, Purdue University

Abstract

In face of increasing pressure on global competition, industry needs to adopt flexible manufacturing in order to expedite manufacturing of customized products and to respond faster to highly unpredictable marketplace. The work was motivated by observation that large scale flexible automation would be the key enabling technology in this emerging manufacturing strategy. Automated flexibility is becoming more important as rapid technological innovation and intense competition shorten product life cycles. Exploitation of AMS advantages relies heavily on a robust and sophisticated control system that monitors the progress of jobs in process and coordinates the activities of the workstations and transport mechanism. One crucial requirement in supervisory controller design is recognizing and addressing the many subtle operational issues associated with full factory automation, taking into account the complexity of the system. As a mission-critical component of the AMS, the supervisory controller must manage resource allocations in real time and utilize the flexibility offered by such systems to achieve the desired performance while making sure that there is no single point of failure that can bring production to a standstill. There are two competing types of control in AMS, performance control and structural control. Performance control focuses on prioritizing parts so that system resources are allocated to achieve certain performance measures. However, the dynamic nature of AMS environment greatly increases the number of decisions that need to be made at any event. In previous studies, it has been suggested that there are two major mechanisms that can cause manufacturing system to stall. First is the careless resource allocation by the controller leading to manufacturing system deadlock. Second is the resource failure which could be devastating to the operation and control of the system, by exacerbating parts blocking and starvation problems which can render the system inoperable. In a complex environment, the performance control can easily drive the system into a situation where a set of parts is deadlocked due to competition for finite resources in the system or where the whole system becomes stalled by the failure of one machine. On the other hand, the foremost concern of structural control is to guarantee deadlock-free resource allocation to concurrent competing processes in AMS. Deadlock avoidance is a difficult problem for many systems that allocate resources in real-time. AMS structural control in the presence of flexibility and failure-prone resource is still in its developmental stage. Most of existing efforts in the area address theoretical scheduling models with respect to long-term average performance measures. In this research, we characterize deadlock and parts blocking structures in AMS and analyze the computational complexities of deadlock detection and avoidance. In the process, we develop a resource allocation model that can represent the system under study in a succinct yet consistent manner. Based on results of the analysis, we develop provably correct and scalable real-time deadlock and parts blocking avoidance algorithms. Then, by exploiting further the characteristics of the AMS structures under study, we identify several special resource allocation structures under partial order that render optimal deadlock avoidance easy and develop method to dynamically allocate central buffer space to increase operational flexibility of robust supervisory controller.

Degree

Ph.D.

Advisors

Lawley, Purdue University.

Subject Area

Industrial engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS