Font Size: a A A

Researches On Agent-Based Intelligent Manufacturing Execution System

Posted on:2004-03-08Degree:DoctorType:Dissertation
Country:ChinaCandidate:Z Q BaoFull Text:PDF
GTID:1102360122475567Subject:Mechanical and electrical engineering
Abstract/Summary:PDF Full Text Request
As the bridge connecting the planning & managing layer and the facility control layer, Manufacturing Executive System (MES) is the key to raising enterprises' agility. This dissertation focus on MES in agile enterprises and an intelligent MES is established based on the multi-agent technologies. An extensive study on the system's crucial technologies is carried out as well. The main contents and achievements of the dissertation as follows:1. Combining hierarchical architecture with federation architecture, the hybrid architecture is constructed. The hierarchy corresponds with the stages of factory production decisions and meets the needs of all agents, which are act at the different levels of the system and have different functions and styles.2. An extended contract net protocol is presented in which the task manager may be permit bidding for tasks. As a result, the space of optimization is enlarged for the cooperation planning. The interaction diagram and planning model are described and the validity of the model is proved by genetic algorithm.3. Aimed at the problems of multi-agent scheduling systems, such as lack of global consistence, a novel scheduling approach is put forward which takes advantages of the optimization approaches and dynamic approaches based on multi-agent techniques. In the approach we adopt a launching policities according to the bottlenecks. A new idea is given that is partial global scheduling in the permitted range for resource requests from the scheduler and its principle is stated. The adjustable mechanism of the hybrid scheduling raises the system's adaptability greatly.4. Applying rough sets a knowledge extraction methodology for agents is introduced. The bidding for tasks between the task agent and the resource agents is described in terms of acquaintance set, invitee set, bidder set and bid-winner. The decision table is built with the attributes of tasks and the attributes of cost, quality, load and time for resource agents to accomplish the tasks. The results of the work show that the knowledge deduced from the decision table is consistent highly with knowledge of experts.5. Aimed at the problems in extracting knowledge directly from the experts' experiences, such as description inconsistency, rough fuzzy methods are presented, in which the agents' knowledge is extracted from the data set recording experts' action instances. The new approaches lead to validity and high consistency of decision knowledge and rationalization of decision process.6. A framework for system integration based on CORBA is built and the communication interfaces between agents are discussed.
Keywords/Search Tags:Agile Manufacturing, Manufacturing Execution System, Agent, Multi-agent System, Cooperative Planning, Hybrid Control, Scheduling, CNP, Rough Sets, Genetic Algorithms, CORBA
PDF Full Text Request
Related items