Proceedings of the 2012 International Conference on Industrial Engineering and Operations Management Istanbul, Turkey, July 3 6, 2012

A New Mathematical Model for a Multi-Criteria Group Scheduling Problem in a CMS Solved by a Branch-and-Bound Method
Y. Gholipour-Kanani Department of Industrial Engineering Islamic Azad University - Research and Science Branch Tehran, Iran R. Tavakkoli-Moghaddam Department of Industrial Engineering University of Tehran Tehran, Iran R. Cheraghalizadeh Department of Industrial Engineering Mazandaran University of Science and Technology Babol, Iran M. Mahmoodjanloo Department of Industrial Engineering Amir Kabir University Tehran, Iran Abstract
This paper presents a group scheduling problem for manufacturing cells, in which parts may visit different cells. By addressing intra-cell scheduling, the sequence of parts within manufacturing cells is determined; however, in inter-cell scheduling the sequence of cells is obtained. We design a new mathematical model for a multi-criteria group scheduling problem in a cellular manufacture system (CMS). The primary objective is to minimize the makespan, intracellular moving, tardiness, and sequence-dependent setup costs simultaneously.

Keywords
Group scheduling, CMS, Makespan, Tardiness, Sequencing-dependent setup cost.

1. Introduction
0B

Cellular manufacturing (CM) is a production system, in which parts requiring a similar production process are grouped in distinct manufacturing cells. These similarities reduce setup times where similar parts can be processed with similar jigs and fixtures. The major advantages of CMSs have been reported in the literature, such as reduction in setup time, throughput time, work-in-process inventories, and material handling costs; better quality and production control; enhancement in flexibility; and the like [1]. Cellular production scheduling finds the optimized sequence of the different and related tasks of the machine, and the sequence of the cells. In this paper the objective function includes four criteria which minimize the intra-cell movements of the parts, the total production time, the tardiness, and the setup cost. The CMS scheduling is an NP-Hard problem, which is one of the most difficult and state-of-the-art problems in combinatorial optimization. This paper considers a scheduling problem, in which inter-cellular moves are allowed and parts may visit machines in the other cells. We solve this problem by a branch-and-bound method with LINGO software. This paper is organized as follows: Section 2 includes a review of the literature and discussion of methods developed for group scheduling problems. In Section 3, the cell scheduling problem considered in this paper is described and formulated. In Section 4, the computational results are presented. Finally conclusion is presented in Section 5.

917

2. Literature Review
1B

Despite this, it took a long time to publish the first article of scheduling about operation research. Yang and Liao [2] considered the cellular scheduling problem with two cells and intra-cell movement, and then resolved it by branch-and-bound technique and a heuristic algorithm. Soleimanpur et al. [3] considered cellular scheduling with setup times. In this problem, they considered a two-step policy to resolve their model. In the first step, the job sequencing has been determined, and in the second step bottleneck machine and dual comparison-based cell sequencing is established. At the end, they resolved their model by a two-step SVS algorithm, and compared the result with LN-PT. Finally they concluded that the SVS algorithm is much efficient than LN-PT. Lu and Yoan [4] considered a single machine classifying with the equal family setup times in order to minimize the maximum tardiness. Shih [5] presented a cellular manufacturing scheduling problem with sequence-dependent family parts setup times. It is concluded that the generated optimization by non-permutation scheduling is better than the permutation scheduling in regard with the due date efficiency. Hendizadeh [6] presented a cellular manufacture scheduling problem considering sequence-dependent family part setup times. Tavakkoli-Moghaddam et al. [7] presented a new mathematical model for scheduling a cellular manufacturing system to minimize the makespan and proposed two evolutionary algorithms for their model.

3. Mathematical Model
2B

The scheduling problem considered in this paper consists of two distinct sequencing problems: sequencing of parts within the cells and sequencing of cells [7]. The following assumptions are considered for the scheduling problem: 1. All parts in all part families are available for processing at time zero. 2. Once an operation starts on a machine, it cannot be interrupted before getting completed (nonpreemption). 3. There is no backtracking in the sequence of machines required by a part family. Once a part is completed on a machine, either it continues processing in the same cell or leaves that cell and joins another cell. As this part returns back to its primary cell, it does not go to the machines already visited. We formulate a novel, nonlinear mathematical model based on machine-part in the CMS. This proposed model deals with the minimization of exceptional elements (EE) and number of voids in cells to achieve the higher performance of cell utilization and minimization of the makespan, intracellular moving, tardiness, and sequencedependent setup cost. The integer nonlinear programming model for the CMS design is presented below. The applied notations, parameters and variations in the mathematical model are as follows: Indices P Number of parts M Number of machines C Number of cells K Maximum sequence of parts in their cell KC Maximum sequence of cell Parameters

1 a ij = 0 1 m jc = 0

if part i visits machine j ; otherwise. if machine j belongs to cell c ; otherwise.

t ij
T H V

Processing time of part i on machine j Time of move between cells for each part (this time is constant for move between each cells) Intra-cell movement cost per time unit Factory costs per time unit (including machines and labor costs and variable production cost and the costs dependent to the work time) 918

CTi Tardiness cost of part i.

di
Wnc

Due date for part i Setup cost for cell c when quickly located after cell n.

Decision variables
S ncj Setup time on machine j in cell c when quickly located after cell n.
S ccj

Setup time on machine j in cell c

1 if part i is assigned to cell c ; x ic = 0 otherwise.

1 if cell c is assigned to sequence b ; y cb = 0 otherwise.

ckjcb

The completion time of a part which is assigned to the k-th sequence on machine j in cell c that cell c assigned to sequence b. Total completion time or makespan

Cmax

Mathematical model In this paper, we introduce our new mathematical model as follows.
p P C M Min Z = ( a ij a ij m jc ) x ic T h + C max V + CTi max 0, C i d i + i =1c =1 j =1 i =1 f C C C (1) Wrq + y c1 Wcc G c =1 b = 2 r =1 q =1 brq

S.T.

x =1 cC ic y cb = 1 c y cb = 1 b

;iP
; b Kc

( 2)
(3)

; c C
; i P & c C

( 4)
(5)

z
k

ikc

= x ic

z
i

ikc

; k K & c C

(6)

C11C1 = max (i p : (t i1 + s cc1 ai1 ) z i1c y c1 ) ; c C

(7)

919

C k 1c1

(8) c k 1,1,c ,1 + max(i p : t i1 z i ,k ,C y c1 ) if ck 1,1,c ,1 > 0 = max(i p : (t + s a ) i1 cc1 i1 z i ,k ,C y c1 ) if ck 1,1,c ,1 = 0 k 2 K & c C

C k1cb

C1,1,C ,b

(10) max(n1 C , n1 c & i P : C K ,1, n1 ,b 1 + ( s n1c1 a i1 y n1b 1 + t i1 ) z i ,1,C y C ,b ) if C K ,1, n1 ,b 1 > C K ,1, n1 ,b 1 > 0 max(n1 C , n1 c & i P : C K ,1, n1 ,b 1 + ( s n2 c1 a i1 y n2 b 2 + t i1 ) z i ,1,C y C ,b ) if C K ,1, n1 ,b 1 = C K ,1, n2 ,b 2 > C K ,1, n3 ,b 3 > 0 = max(n1 C , n1 c & i P : C K ,1, n ,b 1 + 1 ( s n f c1 a i1 y n f b f + t i1 ) z i ,1,C y C ,b ) if C K ,1, n1 ,b 1 = C K ,1, n2 ,b 2 = ... = C K ,1, n f ,b f > 0 max(i P : ( s cc1 a i1 + t i1 ) z i ,1,C y Cb ) if C K ,1, n1 ,b 1 = 0 (b 2) Kc & c C

C1 jc1 = C1, j 1,c ,1 + max(i P : (t ij + s ccj a ij ) z i1c y c1 )

; ( j 2) M & c C

(11)

920

C1 jcb = max(max(n C : C K , j , n ,b 1 ), C1, j 1,c ,b ) + max(n C , n c & i P : (t ij + s ncj a ij y n ,b 1 ) z i1c y cb ) ; ( j 2) M & c C & (b 2) Kc (12)

c kjc1

max(c k 1, j ,c ,1 , c k , j 1,c ,1 ) + max(i p : t ij z ikc y c1 ) = max(c k 1, j , c ,1 , c k , j 1, c ,1 ) + max(i p : (t ij + s ccj a ij ) z ikc y c1 )

if c k 1, j ,c ,1 > 0 if c k 1, j ,c ,1 = 0 (13)

; ( j 2) M , (k 2), c C

C kjcb = max(C k , j 1,c ,b , C k 1, j ,c ,b ) + max(n C , n c & i, o P, i o & d K , d k : ( s ncj a ij y n ,b 1 (1 a oj z odc ) + t ij ) z ikc y cb ) ; ( j 2) M , c C , ( k 2) K , (b 2) K c (14)

C (i ) = max( k K ,

j M , c C & b K c : c( k , j , c, b) z ikc y cb )

; i P

(15)

C max = max(C k , j ,c ,b )

; j M , k K , c C & b Kc

(16)

Gbrq = y rb y qb 1

; (b 2) K c & r , q C , r q

(17)

xic , y cb , z ikc : Binary

; i P , k K , c C & b Kc

(18)

C kjcb , C max 0

; j M , k K , c C & b Kc

(19)

In Eq. (1), the objective function includes minimizing four criteria: namely intra-cell movement time, makespan, tardiness time, and setup cost for each cell regarding to the cell sequencing. Eq. (2) assigns each part to one cell. Eq. (3) insures to assign only one cell to each sequence. Eq. (4) insures to assign each cell to one sequence. Eq. (5) assigns each assigned part to each cell, to one sequence in the same cell. Eq. (6) insures each sequence in each cell to be assigned to one part at most. Eq. (7) shows the C1,1,c,1 is equal to the total of processing time of the mentioned part, and setup time of the first machine in the mentioned cell. Eq. (8) shows Ck,1,c,1, if before the mentioned part, a part which is assigned in the previous sequence of the mentioned is equal to Ck-1,1,c,1 and processing time in present sequence; otherwise, it is equal to total processing time of the present sequence and setup time of first machine in the mentioned cell. In Eq. (9), Ck,1,c,b for b2 is equal to Ck-1,1,c,b and the processing time. If no part in the mentioned cell in the previous sequences has been assigned, it is equal to setup time of the first machine in that cell and processing time. In Eq. (10), C1,1,c,b , if the setup has been done for the first machine in that cell, is equal to the total completion time of the last sequence of the previous cell and Sb-1,b,1 and processing time. If in the previous cell, no process has been done on the first machine, the completion time is equal to total setup time of the first machine from the cell whose setup for the first machine of the has been done and processing time of that part. In Eq. (11), C1,j,c,1 for j2 is equal to C1,j-1,c,1 and the setup time of the machine j in the mentioned cell and processing time. In Eq. (12), C1,j,c,b for j2 and b2 is equal to total maximum completion time of this part on 921

the machine (j-1) and completion time of the last sequence of cell (b-1) on machine j and the setup time of the machine j in the current cell and processing time. In Eq. (13), Ck,j,c,1, if setup has been done for machine j on the mentioned cell is equal to total of maximum completion time of sequence (k-1) on machine j and completion time of sequence (k) on machine (j-1) and processing time. In otherwise it is equal to maximum completion time of sequence (k-1) on machine j and completion time of sequence (k) on machine (j-1) and setup time of the machine j in the cell of the first sequence and the processing time. In Eq. (14), Ck,j,c,b for k,j,b2, is equal to total of maximum completion time of sequence (k-1) on the machine (j) and completion time of sequence (k) on machine (j-1) and the processing time. If there would be no part in the previous sequences (d<k) in the cell of the current sequence, it is equal to the setup time of the machine j and processing time. Eq. (15) computes the completion time of each part. In Eq. (16), Cmax is equal to the maximum completion time. Eq. (17) computes the setup cost of cell r if it is located after cell q; otherwise, the setup cost is zero. Eq. (18) introduces 0 and 1 variables, and finally Eq. (19) introduces non-negative variables.

4. Computational Results
3B

In this section, we prepare the solutions obtained using a branch-and-bound (B&B) procedure with LINGO 8.0 software. To validate the model, ten test problems are solved by the LINGO 8.0 software and the results are shown in Table 1. Table 1: Test problems and results
Test problem 1 2 3 4 5 6 7 8 9 10 P 3 4 3 4 3 3 4 5 4 4 M 2 2 4 4 5 7 5 4 3 4 C 2 2 2 2 2 2 2 2 3 3 LINGO optimal solution OFV 158 263 321 475 652 590 750 763 794 840 CPU time 00:00:01 00:00:18 00:15:06 00:12:23 00:27:57 00:35:52 00:23:55 00:47:57 03:23:52 03:32:55

5. Conclusion
4B

This paper deals with a scheduling problem in a CM environment. In many cases, the conventional flowshop group scheduling approach cannot be applied to CM-based systems. The major shortcomings of the group scheduling problem have been discussed. This paper also considers the cell scheduling problem in the presence of bottleneck machines and exceptional elements incurring inter-cell movement costs in the CMS. In this sense, the considered problem is more general than the conventional flow shop group scheduling problem. We have used a branch-and-bound (B&B) method by the LINGO 8.0 software to solve the foregoing problem that minimizes the makespan, intracellular moving, tardiness, and sequence-dependent setup cost. The LINGO 8.0 software has been used to solve 10 problems.

References
1. Shankar, R., P. Vrat, Some design issues in cellular manufacturing using the fuzzy programming approach, International Journal of Production Research, Vol. 37, No. 11, pp. 25452563, 1999. 2. Yang, W.H., C.J. Liao, Group scheduling on two cells with inter-cell movement, Computers and Operation Research, Vol. 23, No. 10, pp. 9971006, 1996. 3. Solimanpur, M., P. Vart, R. Shankar, A heuristic to minimize makespan of cell scheduling problem, International Journal of Production Economics, Vol. 88, pp. 231-241, 2004. 4. Lu, L.F., J.J. Yuan, The single machine batching problem with identical family setup times to minimize maximum lateness is strongly NP-Hard, European Journal of Operational Research, Vol. 177, pp. 1302-1309, 2007.

922

5. Shih-Wei, L., Y. Kuo-Ching, L. Zne-Jung, Meta-heuristics for scheduling a non-permutation flow line manufacturing cell with sequence dependent family setup times, Computers and Operations Research, Vol. 10, pp. 1016/jcor, 2007. 6. Hendizadeh, S.H., H. Faramarzi, S.A. Mansouri, Meta-heuristic for scheduling a flowline manufacturing cell with sequence dependent family setup times, International Journal of Production Economics, Vol. 111, pp. 593-605, 2008. 7. Tavakkoli-Moghaddam, R., Y. Gholipour-Kanani and R. Cheraghalizadeh, A genetic and memetic algorithm approach to sequencing and scheduling of cellular manufacturing systems, International Journal of Management Science and Engineering Management Research, Vol. 3, No. 2, pp. 119130, 2008.

923