Queueing network modelling of flexible manufacturing system using

Queueing network modelling of flexible manufacturing system using mean value analysis Presented By: YAMIL PEREZ Authors: M. Jain, Sandhya Maheshwari, K. P. S. Baghel Accepted: February 16, 2007

Function of Paper Develop a queuing model to predict the performance of FMS using multiple discrete material handling devices (MHD). Importance Provide an insight on how manufacturing systems can be upgraded today by improving the throughput and reduction in expected waiting time.

References • • • • [1] D. Dubois, A mathematical model of a flexible manufacturing system with limited in-process inventory, Int. J. Prod. Res. 14 (3) (1983) 66– 78. [2] Y. D. Kim, A study on surrogate objectives for loading a certain type of flexible manufacturing systems, Int. J. Prod. Res. 31 (1993) 381– 392. [3] H. T. Papadopoulos, C. Heavy, Invited review: queueing theory in manufacturing systems analysis and design. A classification of models for production and transfer lines, Eur. J. Oper. Res. 92 (1996) 1– 27. [4] J. A. Buzacott, The structure of manufacturing systems: insights on the impact of variability, Int. J. Flexible Manuf. Syst. 11 (1999) 127– 146. [5] H. C. Co, R. A. Wysk, Robustness of CAN-Q in modeling automated manufacturing systems, Int. J. Prod. Res. 24 (6) (1986) 1485– 1503. [6] R. Suri, R. R. Hildebrant, Modeling flexible manufacturing systems using mean value analysis, J. Manuf. Syst. 3 (1) (1984) 21– 38. [7] Y. Dallery, R. David, Some new results on operational analysis, in: E. Galenbe (Ed. ), Performance, vol. 84, Elsevier Science Publishers, B. V. , Amsterdam, 1984. [8] M. M. Srinivasan, Y. A. Bozer, M. Chao, Trip-based material-handling systems. Throughput capacity analysis, IIE Trans. 26 (1) (1994) 70– 89. [9] D. D. Yao, J. A. Buzacott, Modeling a class of state-dependent routing in flexible manufacturing system, Ann. Oper. Res. 3 (1985) 153– 167. [10] H. F. Lee, M. M. Srinivasan, C. A. Yano, Algorithm for the minimum cost configuration problem in flexible manufacturing systems, Int. J. Flexible Manuf. Syst. 3 (4) (1991) 213– 230. [11] D. H. Lee, Y. D. Kim, Loading algorithms for flexible manufacturing systems with partially grouped machines, IIE Trans. 32 (2000) 33– 47. [12] S. H. Choi, J. S. L. Lee, Theory and methodology: computational algorithms for modeling unreliable manufacturing systems based on Markovian property, Eur. J. Oper. Res. 133 (2001) 667– 684. [13] Y. D. Kim, D. H. Lee, C. M. Yoon, Theory and methodology: two stage heuristic algorithms for part input sequencing in flexible manufacturing systems, Eur. J. Oper. Res. 133 (2001) 624– 634.

• • • • • [14] J. A. Buzacott, J. G. Shanthikumar, Design of manufacturing systems using queueing models, Queueing Syst. 12 (1992) 135– 213. [15] P. R. Philipoom, T. D. Rry, Capacity based order review/release strategies to improve manufacturing performance, Int. J. Prod. Res. 35 (1992) 3303– 3322. [16] K. E. Stecke, N. Raman, Production planning decisions in flexible manufacturing systems with random material flows, IIE Trans. 26 (1994) 2– 17. [17] T. M. Smith, K. E. Stecke, On the robustness of using balanced part mix ratios to determine cyclic part input sequence into flexible flow systems, Int. J. Prod. Res. 34 (1996) 2925– 2941. [18] Y. D. Kim, C. A. Yano, Impact of throughput based objectives and machine grouping decisions on the short-term performance of flexible manufacturing systems, Int. J. Prod. Res. 35 (1997) 3303– 3322. [19] I. H. Nam, Dynamic scheduling for a flexible processing network, Oper. Res. 49 (2) (2001) 305– 315. [20] M. Jain, G. C. Sharma, K. P. S. Baghel, Performance prediction of a flexible manufacturing system, Int. J. Eng. 15 (4) (2002) 1– 10. [21] G. C. Sharma, M. Jain, S. Maheshwari, V. Shinde, Transient analysis for maintenance planning of material handling system (MHS), Int. J. Manage. Syst. 20 (2) (2004) 167– 176. [22] F. Kianfar, A numerical method to approximate optimal production and maintenance plan in a flexible manufacturing system, Appl. Math. Comput. 170 (2) (2005) 924– 940. [23] H. Takagi, Fusion technology of fuzzy theory and neural networks-survey and future directions, in: Proceedings of the International Conference on Fuzzy Logic and Neural Networks, Japan, 1990, pp. 13– 26. [24] S. Haykin, Neural Network a Comprehensive Foundation, IEEE Computer Society, Macmillan Press, 1994. [25] J. S. Jang, C. T. Sun, Neuro fuzzy modelling and control, Proc. IEEE 83 (3) (1995) 378– 406. [26] A. Tettamanzi, M. Tomassini, Soft Computing-Integrating Evolutionary, Neural and Fuzzy Systems, Springer, New York, 2001. [27] S. Kartalopoulos, Understanding Neural Network and Fuzzy Logic, third ed. , IEEE Press, 2003. [28] I. Z. M. Darus, M. O. Tokhi, Soft computing-based active vibration control of a flexible structure, Eng. Appl. Artif. Intell. 18 (1) (2005) 93– 114. [29] D. Gross, C. M. Harris, Fundamental of Queueing Theory, second ed. , John Wiley and Sons, New York, 1985.

Relation to the Course

Parameters

Design Description (p. 702) Assumption: Stations cannot wait for the MHD

Design Principle • Algorithm used for calculating the throughput (X) of the material-handling device (MHD) based on Mean Value Analysis. (p. 703)

Design Principle Cont. • • • Queueing Network Model Pallets come out from a blackbox and wait for service in a queue where the MHD serves each pallet in FIFO fashion. Request for service are made by the pallets which are in the blackbox. Assumption: rate at which pallet arrives from the blackbox is a function of (N-m): N=num. of pallets / m=pallets waiting in the queue for MHD M/M/C/N queueing model. Reference [29] • Using this model, the service rate (λ) for the pallet can be found, and hence the average waiting time in the queue or MHD.

Design Principle Cont. • Algorithm to calculate the average waiting time (Wr) of MHD.

Results - Adaptive Neuro-Fuzzy System (ANFS) network Fuzzy toolbox in MATLAB used for approximating Tr, X and Wr. The performance measures are found by varying different system parameters: C, δ, N and S. Using these parameters as inputs, an ANFS network was built. ANFS measure values serve as a comparison for the analytical values calculated using MVA. Fix set of Homogeneous and Heterogeneous processing times for fuzzy and MVA measurements:

Results Cont. Parameters: N=24, S=15, Q=5 As the number of MHD (C) increases, the throughput (X) and average time (Wr) decreases.

Results Cont. Parameters: N=24, δ =15, Q=5 - Effect of S on X and Wr for C=1 and C=5 - X and Wr increase with increasing S. - However, for heterogeneous processing time these take lower values in comparison to the homogeneous one. - Both X and Wr are larger for C=1 compared to C=5

Results Cont. Parameters: S=15, δ =0. 25, Q=5 - As N increases, both X and Wr decrease - For homogeneous values of processing time these take the higher value in comparison to the heterogeneous one

Results Cont. - Result of mean service time (Tr) and (Wr) by varying the move time multiplier (δ). - As δ increases, Tr increases linearly - The waiting time (Wr) increases exponentially as δ also increases.

Correlation of Results with Model The results obtained by adaptive neurofuzzy inference systems are in good agreement with the numerical results obtained using MVA algorithm.

Practical Use • Improvement of the throughput time and reduction in expected waiting time help in upgrading existing manufacturing systems • Minimize the total cost of FMS in the long run

Technical Advancement • The use of neuro-fuzzy techniques for developing approximations for complex problems. • The use of queueing network modeling for FMS quantitative analysis

Industries Most Impacted • Any industry that implements FMS