This section provides a complete FMS example. Zhou and DiCesare (1993) introduced this FMS, as developed at Rensselaer Polytechnic Institute, USA. The following demonstrates that the method proposed here is feasible for such an FMS.
The system takes two types of raw stock, machines them into desired shapes, and then assembles these two ® nished parts into a product. Assume that two product types are to be manufactured. Each has two di erent parts (block and peg). The type of parts are numbered one to four. Figure 10 illustrates the FMS layout. The major components of the system are one CNC mill and drill machine, one CNC lathe, a Microbot robot to load and unload the materials between the lathe and conveyor 3 (C3), and between the mill and C2, and AS/RS with 19 usable pallet-storage bins for bu ering the raw materials and intermediate parts, four two-way conveyors with sensors (sensor1 and sensor2) at each end, a Gantry robot for transferring the materials between the four conveyors, and a Scorbot robot to assemble the parts.
Therefore, the main routes for the block and peg are
B® C4
--- ---
® C1(® AS /RS® C1)
®® C2® M®C2
--- ---
® C1(® AS /RS® C1)
®® C4® A® Fand
Figure 10. The FMS layout.
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014
P® C4
--- ----
® C1(® AS /RS® C1)
®® C3® L®C3
--- ---
® C1(® AS /RS® C1)
®® C4® A® Fwhere B is the block storage, P is the peg storage, Cn is the conveyor, M is the CNC mill, L is the CNC lathe, A is the assembly station, and F is the ® nal product carousel.
The material ¯ ow in the FMS is stated as in the following (Zhou and DiCesare 1993).
(1) The Scorbot robot moves an empty pallet from the gravity fed storage to C4.
(2) The Scorbot robot takes a raw block from the block storage and places it in the empty pallet on C4.
(3) C4 moves the loaded pallet to the Gantry robot.
(4) the Gantry robot moves the pallet from C4 to C2 (C3). (5) C2 (C3)moves the pallet to the CNC machine.
(6) The Microbot robot takes the raw block from the pallet and loads it into the CNC machine.
(7) The CNC machine ® xes the raw block and machines the part.
(8) The Microbot robot unloads the ® nished part from the CNC machine to the pallet on C2 (C3).
(9) C2 (C3)moves the pallet to the Gantry robot.
(10) The Gantry robot moves the pallet from C2 (C3)to C4.
(11) C4 moves the pallet with the ® nished part to the Scorbot robot.
(12) The Scorbot robot takes the ® nished part and places it in the assembly cell.
If two relative parts are present, it assembles the ® nal product.
(13) The Scorbot robot moves the ® nished product to the output carousel.
(14) The Scorbot robot moves the empty pallet from C4 to the pallet storage.
The system can process concurrent works simultaneously.
The functional speci® cation for the FMS is next constructed. Figure 11 (a) illus-trates the MI diagram for the functional speci® cation of the FMS. The material
¯ ow’s ¯ exibility is easily inspected. When the CNC machines are busy, the raw parts can be stored in the AS/RS. While the assembly station is busy, the machined parts can also be stored in the AS/RS for bu ering. When the AS/RS is full, the C1 can hold an additional one part for bu ering. Figure 11 (b) presents the FMI dia-gram for the FMS. The control information is created by the transformation rules discussed in the previous section. Figure 11 (c)to (g)highlights the details of A1 to A5 pages. The deadlock avoidance policies are described as in the external expres-sions of each activity box. Table 2 only illustratively lists the transformed rules for the A1 page of the FMS controller.
To test the controller, total test cases are 422/419 =64, since the AS_RS is a simple bu er. An optimal (shortest) test schedule can be available as indicated in the following,
214131224244431144143211133323243442223411212412334331421342313221, where each number denotes the part type of each product. Only 66 cases are required to test the controller. Figure 12 illustrates the inventories of input, output and AS/
RS, and the machine utilization for the Scorbot, Gantry and Microbot robot of this
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014
Figure 11. The MI diagram for the FMS, (a)the functional M1 diagram, (b)the augmented MI diagrams, (c), (d), (e), (f), and (g)are the details of the A1 to A5 pages.
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014
Figure 11 (continued).
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014
(g
)
Figure 11 (continued).
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014
Activity box Rules
A11 if the sensor of input_storage is busy and the sensor1 of c4 is ready and the sensor2 of c4 is ready and the sensor1 of c1 is ready and the sensor2 of c1 is ready and the sensor1 of c2 is ready and the sensor1 of c3 is ready and ((the sensor of assembly_station is ready and ((the part of c1 is machined and the type of c1/= 1 and the type of c1/= 3) or the status of c1 is ready))or (the sensor of assembly_station is busy and ((the part c1 is machined and the type of c1/= 2 and the type of c1/= 4)or the status of c1 is ready)))and the request of s_robot= 0 and the status of s_robot is ready then conclude that the request of s_robot= 1 and conclude that the type of s_robot= 5 and start start_s_robot(1)and conclude that the status of s_robot is busy
A12 if the sensor1 of c4 is busy and the pgm of s_robot is ® nished and the request of s_robot= 1 and the status of s_robot is busy and the status of c4 is ready then conclude that the request of s_robot= 0 and conclude that the type of c4= the type of s_robot and conclude that the status of s_robot is ready and conclude that the status of c4 is loaded
A13 if the type of c4= 5 and the sensor of input_storage is busy and the sensor1 of c1 is ready and the sensor2 of c1 is ready and the request of s_robot= 0 and the status of s_robot is ready then concude that the request of s_robot= 2 and conclude that the type of s_robot= the type of input_storage and conclude that the part of s_robot is raw and start start_s_robot(2)and conclude that the status of s_robot is busy.
A14 if the sensor1 of c4 is busy and the pgm of s_robot is ® nished and the request of s_robot= 2 and the status of s_robot is busy then conclude that the request of s_robot= 0 and conclude that the type of c4= the type of s_robot and conclude that the part of c4= the part of s_robot and conclude that the status of s_robot is ready
A15 if the type of c4/= 5 and the type of c4/= 0 an the part of c4 is raw and the sensor1 of c4 is busy and the sensor2 of c4 is ready and the request of c4= 0 and the status of c4 is loaded then conclude that the request of c4= 1 and conclude that the status of c4 is busy and start start_conveyor(c4,1)
A16 if (((the type of c4= 1 or the type of c4= 3)and (the sensor1 of c2 is ready and the sensor2 of c2 is ready)and the sensor of mill is ready)or((the type of c4= 2 or the type of c4= 4)and (the sensor1 of c3 is ready and the sensor2 of c3 is ready)and the sensor of lathe is ready)or ((the sensor1 of c1 is ready and the sensor2 of c1 is ready)and not ((((the type of c4= 1 or the type of c4= 3)and (the sensor1 of c2 is ready and the sensor2 of c2 is ready)and the sensor of mill is ready)or ((the type of c4= 2 or the type of c4= 4)and (the sensor1 of c3 is ready and the sensor2 of c3 is ready)and the sensor of lathe is ready))))) and the sensor2 of c4 is busy and the request of c4= 1 and the status of c4 is busy and the request of g_robot= 0 and the status of g_robot is ready then conclude that the type of g_robot= the type of c4 and conclude that the request of g_robot= 1 and conclude that the request of c4= 0 and conclude that the part of g_robot= the part of c4 and conclude that the status of c4 is ready and conclude that the status of g_robot is busy and start start_g_robot(1).
Table 2. The transformed rules for the A1 page of the FMS controller.
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014
simulation testing. The steady state for the FMS results in an inventory of AS/RS as 14 parts. The Gantry robot gains a maximal machine utilization. Those statistics reveal that the FMS controller is very stable.
7. Conclusion
In summary, this work has presented a systematical method of designing a rule-based FMS controller. The functional speci® cation of the FMS controller is ® rstly synthesized by the individual MI diagram primitives. The control ¯ ows can be created by transforming the resource utilization cycles. The manufacturing policies are also speci® ed and attached. The FMI diagrams can be created by the transfor-mation and attachment. The FMI diagrams can be transformed to the production rules which can be executed on an expert system. The rule-based FMS controller is therefore constructed.
To verify this controller, two approaches are proposed. For a small system, the C
Â
PN model is suggested. This model can be created by transforming the FMI diagrams. Properties of the CÂ
PN model, e.g., the safeness and liveness, can then be inspected by the P-invariants method, or the reachability tree. For a large system, however, simulation testing is recommended instead since the analytical CÂ
PN modelmay not be feasible. Some general guidelines for the simulation process are pro-posed. Two examples are employed to illustrate the method provided here. The ® rst example is the common FMC with one robot, two machines and one bu er. The other is the famous FMS example introduced by Zhou and DiCesare (1993), where
Figure 12. The inventory and machine utilization chart.
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014
one AS/RS, two machines, three robots, four conveyors and an assembly cell are employed. Both examples demonstrate the usefulness and rapid prototyping cap-ability of the proposed method.
The fault diagnosis problem for the approach may be an area of future research.
Results in this study can hopefully contribute toward the design of an intelligent FMS controller.
References
Bau er, A., Bow den, R., Brow ne, J., Du gga n, J., and Lyons, G., 1991, Shop Floor Control Systems ± From Design to Implementation (London: Chapman & Hall).
Cecil, J. A., Sriha ri, K., and Emerson, C. R., 1992, A review of Petri-net applications in manufacturing. International Journal of Advanced Manufacturing Technology, 7, 168±
177.
Coffma n, Jr, E. G., Elphic k, M., and Shoshani, A., 1971, System deadlocks. Computer Surveys,3 (2), 67± 78.
Dav id, R., and Alla, H., 1992, Petri Nets & Grafcet ± Tools for Modeling Discrete Event Systems (Englewood Cli s, NJ: Prentice Hall).
Dupond-Gatelmand, C., 1982, A survey of ¯ exible manufacturing systems. Journal of Manufacturing Systems,1(1), 1± 16.
Franks, I., Loftus, M., and Wood, N. T. A., 1990, Discrete cell control. International Journal of Production Research,28(9), 1623± 1633.
Gensym, 1994, G2 Reference Manual, Version 4.0 (USA: Gensym Corporation).
Goodenou gh, J. B., and Gerhart, S. L., 1975, Towards a theory of test data selection.
IEEE Transactions on Software Engineering,SE-1(2), 156± 173.
Hollowa y, L. E., and Krough, B. H., 1990 Synthesis of feedback control logic for a class of controlled Petri nets. IEEE Transactions on Automatic Control,35(5), 514± 523.
Hong, H. M., 1993, IDEF/CPN/G2 approach to the implementation of real-time shop ¯ oor control system. MS thesis, National Chiao Tung University, Taiwan.
Jafa ri, M. A., and Bou cher, T. O., 1994, A rule-based system for generating a ladder logic control program from a high-level systems model. Journal of Intelligent Manufacturing, 5, 103± 120.
Jones, A. T., and McLean, C. R., 1986, A proposed hierarchical control model for auto-mated manufacturing systems. Journal of Manufacturing Systems,5(1), 15± 25.
Kusia k, A., 1990, Intelligent Manufacturing Systems (Englewood Cli s, NJ: Prentice Hall). Lia ng, G. R., and Hong, H. M., 1994, Hierarchy transformation method to manufacturing
system speci® cation, design, veri® cation, and implementation. Computer-Integrated Manufacturing Systems,7(3), 191± 205.
Meyer, W., 1990, Expert Systems In Factory Management ± Knowledge-Based CIM (London:
Ellis Horwood).
Mu rata, T., Komoda, N., Ma tsumoto, K., and Haruna, K., 1986, A Petri net-based controller for ¯ exible and maintainable sequence control and its applications in factory automation. IEEE Transactions on Industrial Electronics,IE-33(1), 1± 8.
Sa uv e, B., and Collinot, A., 1987, An expert system for scheduling in a ¯ exible manufac-turing system. Robotics and Computer-Integrated Manufacmanufac-turing,3 (2), 229± 233.
Teng, S., and Blac k, J. T., 1989, An expert system for manufacturing cell control. Computers in Industrial Engineering,17(1± 4), 18± 23.
Va lette, R., 1987, Nets in production systems. In Advances in Petri Nets: Petri Nets Applications and Relationships to Other Models of Concurrency, L ecture Notes in Computer Science,255(New York: Springer), pp. 191± 217.
Wu, S. Y. D., and Wysk, R. A., 1988, Multi-pass expert control system ± a control/scheduling structure for ¯ exible manufacturing cells. Journal of Manufacturing Systems,7(2), 107±
Xia ng120., D., and O’Brien, C., 1995, Cell control research ± current status and development trends. International Journal of Production Research,33(8), 2325± 2352.
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014
Zhou, M., and Dicesare, F., 1990, Modeling bu ers in automated manufacturing systems using Petri nets. Proceedings of Rensselaer’s Second International Conference on Com-puter Integrated Manufacturing, pp. 265± 272.
Zhou, M., DiCesare, F., and Desrochers, A. A., 1992, A hybrid methodology for synthesis of Petri net models for manufacturing systems. IEEE Transactions on Robotics and Automation,8(3), 350± 361.
Zhou, M., and DiCesare, F., 1993, Petri Net Synthesis for Discrete Event Control of Manufacturing Systems (Boston, MA: Kluwer Academic Publishers).
Downloaded by [National Chiao Tung University ] at 05:38 28 April 2014