Division of Information Technology, Engineering and the Environment
School of Advanced Manufacturing and Mechanical Engineering
School of Advanced Manufacturing and Mechanical Engineering
Thesis Abstract
Rapid market changes, a growth in demand for product quantity, concerns for product quality and requirements to increase dramatically product mix have forced manufacturing industries to enhance their flexibility. Flexible manufacturing systems (FMS) are designed to cope with these conditions by using high technologies and automation in transfer lines which enable factories to reconfigure rapidly to produce a variety of products by using same resources. Due to the high investment required, higher system utilization must be achieved. Appropriate planning and scheduling have important role in implementation of FMS and may lead to manufacturing industries optimizing their productivity.
There are several issues in production planning stage such as part type selection, machine grouping, production ratio, resource allocation and loading problem. The next stage after production planning is scheduling. All methods presented in literature review consider production planning problem and scheduling problem as separated problems or do not cover all issues in short term planning and scheduling problem. These kinds of approaches only lead to achieving local optimum solutions. However, the complexity of integrated production planning and dynamic scheduling require robust methods which minimize effects of a various unexpected disturbances and also adapt quickly with change in dynamic shop floor.
Multi-agent based systems (MAS) provide an architecture which several different tasks can be decentralized in several levels. Each level contains several autonomous agents which represent physical or functional manufacturing entities such as resources, operators, parts and jobs. This hierarchical multi-layer architecture limits the computational complexity and reduces software engineering efforts by using modularity approach. These modular components can be run on several parallel computers to perform a distributed computing which reduce computational time. However, negotiation mechanism among agents in MAS concerns more on flexibility and quick response on environment changes rather than reaching optimality. Integration with AI-based meta-heuristic search methods improves agents’ capability. The power of genetic algorithms (GAs) to explore a wide search space within a required time constraint fills this need. The combination of MAS and GAs may provide a robust and adaptable framework for these complex problems and also give a better result in a reasonable amount of time.
The system performance is analyzed by using cost and computational time. The cost is calculated from several weighted parameters such as completion time (makespan), mean flow time, total tardiness, number of tardy jobs and waiting time. The computational time is also used to analyze system efficiency. A system with lower cost and computational time is considered as a better system.
