Robust Supervisor Synthesis For Automated Manufacturing Systems Using Petri Nets

A manufacturing system is usually a computer-controlled one with resources such as automated guided vehicles, buffers, and computer numerical machine tools. Different kinds of job types are processed along with preestablished paths competing for limited number of resources. Deadlocks may occur owing to the circular wait caused by the competition among these concurrent processes. As a consequence, research on deadlocks has been a hotspot in automation area. In the past decades, a lot of researchers have devoted to solving the deadlock problem. For simplicity, most of the studies presumptively and arbitrarily suppose that system resources do not fail; nevertheless, this is quite the opposite in reality. Actually resource failures occur frequently due to various causes in an automated manufacturing system. Any subtle error can cause unexpected failures of resources and even unpredictable disruption. In the worst case, the failure of a single resource may lead to the crash of the entire system, the loss of which is immeasurable economically.In this thesis, resource failures will be taken into consideration. A distributed control policy is derived to achieve the tractability in deadlock-freeness, robustness, and system performance matters. The system resources are categorized into unreliable and reliable ones depending on whether or not they are subject to failures. In the paradigm of Petri nets, this thesis proposes a distributed and dynamic searching algorithm. Different from the conventional off-line approaches which require to enumerate all reachable states, the proposed method will only check the resource usage on-line not only avoiding deadlocks but also preventing the conventional state explosion problems. Based on this deadlock avoidance algorithm, a robustness intensification algorithm is developed to guarantee that processes not necessarily using the failed resources can still execute smoothly when resource failures occur. Thus the enterprise production profit can be greatly increased. To improve system performance, we expect the difference of moving steps between any two tokens is minimized thereby improving the concurrency among processes. In this thesis, we propose three algorithms, which focus on deadlock avoidance, robustness intensification, and concurrency improvement issues, respectively. Every time the algorithms are executed, we can obtain a set of fireable transitions among which one and only one transition is going to fire. Whenever the system comes to a new state, these three algorithms will be executed in sequence iteratively so that an appropriate event sequence can be derived that will lead the system to a desirable state.The Petri nets theory is used as a major research tool and analysis method through this paper. This thesis focuses on the research of robustness of automated manufacturing systems with unreliable resources, proposing a robust algorithm on the basis of deadlock avoidance ensuring that the system will not be blocked due to resource failures and further improving system performance.