Intelligent scheduling and control of automated guided vehicle considering machine loading in a flexible manufacturing system: using hopfield networks and simulation.

摘要

Flexible manufacturing systems (FMS) have received increasing attention from researchers and practitioners due to their potential advantages: quicker response to market changes, reduction in work-in-process (WIP), high inventory turnover and high levels of productivity. Two groups of problems in an FMS are of importance: (1) design problems and (2) operational problems. Operational problems can be effectively separated into 4 sub-problems: planning, grouping, machine loading problem (MLP) and scheduling. Problems from machine loading to scheduling and control of an FMS can be handled with neural networks approaches and simulation.The machine loading problem as a combinatorial optimization problem is actually a classic problem in operations research and is known to be NP-hard. MLP formulated as 0-1 integer programming problems has been solved by the methods of linearizing the nonlinear terms, branch and bound algorithm, and heuristic methods which have also been popularly applied.Hopfield Networks as a class of artificial neural networks have been adapted as an efficient method to solve the MLP, as these are able to find the solutions quickly through massive and parallel computation. Unfortunately, the quality of the solutions can occasionally be poor owing to the values of the weighting parameters in the energy function of the Hopfield Networks. One alternative approach used is to imbed mean field annealing into Hopfield Networks. The hybrid method of Hopfield Networks and mean field annealing can find near-optimal solutions as well as overcome the difficulties with decisions about the weighting of parameters in the energy function. The AGV scheduling problem can be regarded as the problem of selecting appropriate dispatch rules. Many dispatch rules have been introduced by a number of researchers. Even though vqarious formulations of the FMS scheduling problem can be presented, simulation methods are popular and often used.A solution methodology for MLP and AGV scheduling problems is proposed and specific models based on the literature are subjected to experimented through simulation. The proposed methodology can be also applied without difficulty to of breakdowns of machines and AGV. Results from simulation experiment s show that superior performance and capability of the proposed to existing methods are demonstrated by applying them to the test problems represented by simulation..