Optimal Coordination of Flexible Manufacturing Systems, with Automatic Generation of Collision- and Dealock-Free Working Schedule
Doctoral thesis, 2008
The ever more rapidly changing markets pose high demands on the modern industry, often making it necessary to have varied and frequently updated product portfolios. As a consequence, modern industrial systems need to be easily adaptable to different kinds of products, which makes the use of flexible manufacturing systems (FMS) increasingly popular. An FMS generally contains a number of moving actors, such as production robots, conveyor belts, etc, that can be configured for different tasks. However, a big challenge with FMS is their high complexity, which makes FMS coordination a time and resource demanding undertaking.
In this thesis, the challenge of FMS coordination is accepted, with the goal of developing methods for automatic and off-line generation of a correct, safe and time optimal working logic for the moving actors of a given FMS. This means that the order of operations in the studied FMS should minimize the total cycle time of the system and respect all specifications, while avoiding collisions and blocking situations between the moving actors.
To represent possible and specified FMS behavior, deterministic finite automata (DFA) models are used. A method to automatically generate such models is presented, whereafter much work is laid at developing optimization methods, applicable to DFAs. While we start out with considering normal FMS behavior, a method for treating FMS that suffer from uncontrollable operations, such as machine breakdown or manually ordered product inspection, is also presented. When relating our results to the real world, we noted that the optimal control logic often induces uneven movement patterns to the FMS actors. This inconvenience is thus studied and amended.
Finally, combining our results with some existing techniques, a framework for automatic generation of control code from 3D simulation models of FMS is presented. In developing this framework, functionality common to most robot simulation environments is used where possible to facilitate the portability of the approach between different simulation tools.
Mixed Integer Linear Programming
Discrete Event Systems
Supervisory Control Theory
Deterministic Finite Automata