(a) It must be a powerful and user-friendly tool

(b) Accurate models of real industrial must be easily accessible as well as mechanical modelling tools (for example, through a library)

(c) Automated tools placement must be represented to test reach, tool orientation and highlight any changes in cell configuration

(d) Other automated plant components such as conveyor belts, part feeders, CNC machines may also need to be included in the simulation model

(e) A calibration facility must be available to increase model accuracy (32)

(e) A collision detection feature must be included in the software; integration between production scheduling routines and the simulation model is highly desirable

(g) Off-line programming facility for subsequent downloading is a cost-effective feature

(h) Simulation, calibration and off-line programming of industrial automated equipment should be possible, ideally using a standard low-cost desk or laptop PC.

 

 

a) Intentions and objectives as the agent moves towards a pre-defined set of goals given the means available

b) Rationality as an agent will select the best course of action based upon a set of evaluation criteria

c) planning ability in that an agent will co-ordinate its actions with other agents and will plan how to attain its own objectives

d) Adaptability with respect to the environment and the other agents with which one agent interacts; e) intelligence if the agent has all the characteristics listed in a) to d)

Complex intelligent agents have an explicit representation of their environment, may keep track of past performance and events, and are probably small in number. Most systems also contain a large number of reactive agents with limited protocol and communication skills and no explicit representation of the outside world. Not all agents are required to be individually intelligent for a command and control system to be considered intelligent. The three objectives of a multi-agent system can be summarised in three words: communication, control and organisation 35. There are various design options in terms of communication. Agents may be unable to communicate directly with each other but go through a common blackboard. They can exchange asynchronous messages, be spatially distributed with a large degree of autonomy, and rely on a set of pre-defined production rules placed on an inference engine or follow recent trends towards the more abstract holonic or neural communication system.

Command and control architectures can be classified depending upon the forms of communication and control used throughout the system. These may be a meta-object agent, supervisory agents, cell agents, product agents and finally resource agents, the latter being non-intelligent in classic workstations 36. The control architecture may be centralised (with and without pre-planned production schedules), hierarchical, co-ordinated, distributed and finally distributed and supervised. In the case of the co-ordinated and distributed architectures, intra-level communication links exist with or without a global supervisor (37).

 

 

 

Once the best solution for a particular problem is obtained, it will be reused in a similar situation. Evidently, these two philosophies cannot be applied simultaneously to the same type of systems. The second type in particular is reserved for systems when response time is irrelevant (as in the case of long production cycles) but it is imperative to apply the best remedial solution for a particular problem because of costs, for instance. As discussed in detail above, there are three broad choices to automate the command and control function:

¤ A unique application manages all, an ideal choice for small systems because of the risk of an explosive increase in operating time for large systems.

¤ A multi-agent application in which each agent is assigned to a particular task, a much more flexible alternative. There are many architectures of this type, the choice of which depends on the particular configuration of the operational system.

¤ A multi-expert system that is a particular application of the multi-agent system where each agent is an expert system. This particular choice is almost exclusively applied to systems adopting the second philosophy where the quality of the response is more important than response time.

In conclusion, the command and control module plays a central role in the CIM chain. It simplifies the shop-floor manager job by automating some of the managerial functions. It provides a reliable link between scheduling and numerically controlled automated machines by translating and interpreting scheduling requirements. Finally, it relies on a high performance computer system for quick reactivity to endogenous and exogenous shocks to the productive system.

 

_________________________________________
PRODUCTION SCHEDULING (42)

Production scheduling is a short-term activity and provides a detailed specification of the work that the machines are to perform, usually over the period of a few days or weeks. In a scheduling problem, jobs must be processed within the time period fixed at the production planning stage, as well as their arrival times and due dates, whereas the order in which the jobs are to be processed has to be determined.

The most basic model requires a single machine to be scheduled. If there are several machines performing the same function, a parallel machine-scheduling problem results. More complex models require several operations at different stages to be performed on the jobs. In a flow shop, every job passes through the machines in the same order. However, in the more general job shop model, different jobs have different machine routings and a job may revisit a machine (in a re-entrant system).

The objective is sometimes to minimise the maximum completion time or the sum of weighted completion times of the jobs. However, in the presence of due dates, the objective may be to minimise other performance measures such as the maximum lateness, the total weighted tardiness or the weighted number of late jobs.

With the increased use of flexible and automated machinery, scheduling problems differ from the classical ones mentioned above. More precisely, the assumption in classical scheduling is that each type of operation is performed on a dedicated machine or group of machines, whereas in modern production systems, there is often a choice of which machine or tool to use for an operation. Thus, scheduling problems require an allocation of operations to machines as well as the determination of the operations processing order for each machine.

The investment in the provision of an automated production system is often substantial. To reap the benefits of such heavy investment, a high level of utilisation of the machinery should be achieved. For the repetitive manufacture of high volume products, therefore, a common objective is to maximise the throughput of the products.

 

_______________________________________________________
ALTERNATIVE CONFIGURATIONS
FOR THE FMS INDUSTRIAL PLANTS

 

The application of concurrent engineering principles during the first drawings meant that the modules would be given general information about the case study and that the staff would formulate their own set of preferences to select the best design for that FMS.

Given the wide variety of modelling approaches used in process design and the interdisciplinary nature of most planning groups and in this project, the choice of a common and easily understood modelling tool is a major concern. Graphs, 3D Design & Simulation are selected as modelling tools for the FMS design analysis for some reasons:

¤ It is commonly used in disciplines dealing with flow optimisation such as logistics, production planning and plant lay-out and a modern standard in automation and robotics, or command and control;

¤ It allows the joint evaluation of module solutions;

¤ It provides a smooth convergence to a global optimum solution, if one exists.

For one case study, various design features and all feasible combinations have to be assigned, the set of logically consistent options across modules define the best feasible solution, if one exists. Each of the modules is based on experience or modelling results but should take into account preferences from other modules, i.e. optimised solutions.

Internal module preferences may be fully represented in the 3D simulations techniques (such as the A&R - Automation and Robotics module) but be reflected in the total weight placed on a particular design. In some other cases, the complete set of individual module objectives cannot be evaluated in the global context (such as balanced workload, long production runs and location of WIP - Working in Process storage for the P&L - Production & Layout module). In this case, modules may simply assign a global weight to the set of design features available before providing its own contribution (this is the case for the P&L and the M&L - Maintenance & Logistics modules shown bellow in the FMS designs). In some other cases, modules may not assign any intermediate options.

Although this approach provides several feasible solutions, it has the disadvantage that the amount of information increases from top to bottom. Subsequent runs of the same problem in a different order may alter the order of preferences and the ranking of the solutions. The selection of the most desirable module consultation order is one of the simulator's decision variables in its search for optimality.

For example, the C&C Command & Control module may be consulted at the bottom end rather than at the top of the tree because it requires a significant amount of information from other modules. This includes the number and type of machines proposed, the use of cells for production (or not), scheduling, maintenance and logistics requirement.

The M&L module requirements will have an impact on plant layout and equipment choice. And so on. In other words, there is an important role to be played by the 3D simulator. Otherwise it must be decided the order in which it visits the modules, how to obtain global feasibility, how to impose additional constraints on individual modules if necessary and finally how to move towards global optimality.

In a case study, many feasible solutions are identified, a number of which being clearly superior to the others thus providing an important first step to discuss optimisation as well as the best approach for the general case. The drawings below are a summary of the strategic decision variables selected by each of the four modules and the set of solutions assigned to the various options. The complete set of options and options for the case study following the analysis and approval were simulated to propose a Generic FMS design.