| In manufacturing industry, design and implementation of flexible manufacturing systems (FMS) are a very complicated task. It is necessary to model and analyze FMS before they are put in practice and run. In this way, many problems existing in the systems can be solved at the design stage. This dissertation focuses on deadlock prevention issues arising from the needs to optimize a production plan according to product information and manufacturing capabilities, and overall ensure smooth material flow and high productivity. A number of novel deadlock prevention policies by using Petri net theory are developed.Usually, the capability of every manufacturing resource is fixed in a system. For instance, a robot can pick and hold a work piece every time. However, the rapid development of manufacturing technology keeps updating the capability of the manufacturing resources, thus poses new challenges in system design. When a robot arm in a manufacturing system can handle two or more work pieces at same time, the initial markings in its Petri net plant model are changed accordingly. Then, it is necessary to construct a new supervisor again for the new net model that has the same structure but different initial markings. According to the existing methods in the literature, even if the structure is not changed, the changes of initial markings mean redesigns of new supervisors again and again. The methods proposed in this dissertation aim to solve this problem. They can be used to design a new supervisor by simply computing a set of constraints to adjust the initial markings of the control places in the former supervisor without changing its control structure.This dissertation ensures deadlock-freeness of a system by proper model and control methodology, and proposes several control methods for the increasing productivity of every section in FMS. Furthermore, in order to cope with the problem of less reachability states, this dissertation analyzes the reachability graph first, and designs an initial monitor based on the theory of regions, which can receive a maximally permissive supervisor. Then all the marking constrains are computed for every strict minimal siphon in the process of quadratic control, such that deadlock-free initial markings are obtained. Finally, a controller is designed by combining the initial controller with the deadlock-free initial markings. It is worth noting that all strict minimal siphons are divided into elementary siphons and dependent ones to compute the marking constrains by their linear relationships, which reduces the amount of the computation. During computing the marking constrains, control places are added to elementary siphons only to reduce the number of control places. This dissertation uses INA that is a popular analysis tool of Petri net to check whether the supervisor is deadlock-free or live, and the results show that the supervisor system performs excellent and is deadlock-free or live, which mean the control algorithms are reasonable and feasible.With the expansion of the system scale, the computational cost increases. For the purpose of coping with the rapid growth, this dissertation proposes a new control policy for relatively complicated net systems. Due to the large numbers of siphons in these systems, an idea on sort operations is presented. For the siphons in original plant model, marking constrains are obtained by invariant control method. For the new siphons in the initial supervisor, marking constrains are obtained by reallocating the markings in idle places and resource places. A relatively complicated Petri net system is illustrated to show its rationality and feasibility.All the proposed algorithms in this dissertation are generalized, and can be used not only in simple net systems but relatively complicated net systems. To optimize the steps of supervisor design and reduce the computational cost in practice, one can choose a proper algorithm for a given system according to its complexity.
|
PDF Full Text Request