International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 8- August 2013 Minimization of wastage of material for a flat production using different heuristics Bindu Neeharika.V1, Y.L.Vidya2, Kakati.S3 1,2,3 Assistant professor,ANITS,Visakhapatnam,India Abstract— This paper discusses some of the basic formulation issues and solution procedures for solving one- and twodimensional cutting stock problems. Linear programming, sequential heuristic and hybrid solution procedures are described. For two-dimensional cutting stock problems with rectangular shapes, we also propose an approach for solving large problems with limits on the number of times an ordered size may appear in a pattern. order for specified quantities of smaller lengths. This is achieved by generating optimal cutting patterns for the cutting of the stock lengths. Stock lengths have cost associated with them. The greater the quantity of stock lengths used in filling an order, the greater the cost of the order to the customer. This, essentially, means reducing the quantity of stock lengths used. It arises from many applications in industry. .Keywords— Cutting stock, Trim loss, Linear programming, The Cutting Stock Problem is essentially an integer programming (IP) problem. However, integer programming problems are known to be NP-hard and therefore, the Cutting Stock Problem is formulated as a linear program by relaxing the integer constraint. This makes Cutting Stock Problems amenable to linear programming methods of solution. The resultant LP optimum is rounded to get the IP optimum. The columns of the basis matrix represent all the cutting patterns that can be produced from the available stock length. The number of cutting patterns can be very large and this makes explicit enumeration of all feasible cutting patterns impractical. Therefore an initial set of feasible patterns is generated and used as the basis for the simplex method to solve for the dual variables. Two-dimensional knapsack., I.INTRODUCTION Economic resources are scarce and have costs associated. The scarcity and cost associated with these resources generally impose certain constraints in their utilization. The effective management of these constraints aimedat minimizing the overall cost of input resources and the maximization of corresponding profit is the subject of optimization. Practical optimization is the art and science of allocating scarce resources to the best possible effect (Amponsah, 2006). Optimization techniques, a branch of mathematical programming, has enjoyed enormous appeal after World War II, both in the academia and in practice. Subsequently, this interest inspired numerous researches that sought to identify, analyze and substantiate new techniques for improving industrial and business processes. Currently, Optimization techniques have become an indispensable tool for industrial applications including resource allocation, scheduling, decision-making, etc. Optimization techniques have various branches and one such branch is linear programming. The term “programming” in linear programming does not assume programming as used in the field of computer science to denote software development. Instead, it focuses on mathematical modeling and the requirement of a finite number of iterations to solve the model. This project focuses on a special type of linear programming called the Cutting Stock Problem. II.GENERAL OVERVIEW OF CUTTING STOCK PROBLEM The aim of Cutting Stock Problem is to minimize the total cost of stock length of given cost that is cut to fill an ISSN: 2231-5381 III.CLASSIFICATION OF CUTTING STOCK PROBLEM The Cutting Stock problems can be classified by the dimensions of the cutting object. This can be one-, two- or three-dimensional problems. A. One Dimensional Cutting Stock Method The one-dimensional cutting stock problem is to obtain a given set of ordered lengths (patterns) from stock lengths. The objective is typically to minimize the total cost of stock materials used (material input). A cutting pattern describes how many items of each type are cut from a stock material. The one-dimensional cutting stock problem is defined by the following data. Let the one dimensional model is as follows: where is the number of times pattern j is used, is the cost of stock material used for cutting pattern j, is the number of in pattern j and is the quantity of ordered. To be a valid cutting pattern, a pattern must satisfy where is the length of the kth stock material used to cut the pattern. http://www.ijettjournal.org Page 3630 International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 8- August 2013 The huge number of patterns is not available explicitly for practical problems. Usually, necessary patterns are generated during a solution process, hence the term column generation. However, the number of different patterns in a solution cannot be greater than the number of stock lengths and is usually comparable with the number of piece types. Width B. Two Dimensional Cutting Stock Method Two important procedures was utilized here in finding out the Optimal cutting patterns & number of mother coils against each pattern to meet all the required number of slit coils in order to minimize the trim loss. Another important variant of the cutting stock problem is the two-dimensional cutting stock problem. This variant can be divided into regular (rectangular, circular) and irregular shapes (Farley, 1988). Rectangular shapes can be obtained through guillotine or non-guillotine, oriented or non-oriented cutting. An oriented cutting means that the lengths of rectangles are aligned parallel to length of the stock sheet. Slit Coil req. (Ri) 0 5 0 0 64 0 64 5 Finding optimal cutting patterns: The optimal cutting patterns are determined using the algorithm for pattern generation and searching procedure are as follows. Generating feasible cutting patterns: Mother coil width of 1250mm is considered while A two-dimensional cutting stock problem can be defined as determining the cutting patterns. Appling pattern generating follows: A set of rectangular stock sheets of different types is algorithm in generating feasible cutting patterns. The coefficients (no. of cuts) against each individual available. For each type of sheet we know its length and width. From these sheets we have to cut smaller rectangular pieces of slit widths can be determined b using the following equation. Aij = Smallestlnt [W - ∑i=1 (AiWi)/Wi) length and width, in order to satisfy a given demand for pieces of each type. The objective is to minimize the total area of Step 1: Descending the order of requirement widths (Wi) stock sheets required. A sequence of cuts of a sheet into Step 2: Elements of first row (j=1) Finding no. of cuts from the individual pattern rectangular pieces is a cutting pattern. A11 = Smallestlnt[(1250-0/121.5] = 10 A21 = Smallestlnt[(1250-10*121.5)/100.4] = 0 IV.CALCULATIONS A31 = Smallestlnt[(1250-10*121.5*100.4)/89.8] = 0 A41 = Smallestlnt[(1250-10*121.5-0*100.4-0*89.8)/85] A. One Dimensional Cutting Stock Method =0 The amount of steel should be indented should A51 = Smallestlnt[(1250-10*121.5-0*100.4-0*89.8include the possible steel loss during the Slitting operation. 0*85)/80]=0 The Customer orders will be summarized at actual and at the A61=Smallestlnt[(1250-10*121.5-0*100.4-0*89.8item level (against different OD) and based on which the 0*85number of slit coils required will be determined. The number 0*80)/74]=0 of slit coils required for the month of April 11 is as follows: A71=Smallestlnt[(1250-10*121.5-0*100.4-0*89.80*85-0*800*74)/70]=0 A81=Smallestlnt[(1250-10*121.5-0*100.4-0*89.80*85-0*80-0* 740*70/60.1]=0 A91 = Smallestlnt [(1250-10*121.5-0*100.4-0*89.80*85-0*800*74-0*700*60.1)/49.8]=0 Step 3: Determining the cutting loss of pattern 1 using eqn.(6) from section 3.32. TABLE-I Cut loss of pattern 1 using eqn. Tj W - ∑ i=1 (AijWi) T1=1250Outer (10*121.5+0*100.4+0*89.8+0*85+0*74+0*70+0*60.1+0*49. 38.1 31.76 28.6 26.99 25.4 23.42 22.22 19.05 15.9 Dia. 8) (OD) = 35mm 121.5 100.4 89.8 85 80 74 70 60.1 49.8 Step 4: set the level index 1 from the last row, where Aij>0 Slit Coil (row index)= 9.8=1 ISSN: 2231-5381 http://www.ijettjournal.org Page 3631 14 International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 8- August 2013 i.e. level-1, 11>0 and others in that row are equal to zero. Step 5: Since the current of level-1(A11) is greater than zero, a new column j=j+1=2 is introduced with the following points. a) Value of (A11) will be decrement by 1 b) Values above (A11) remain same and c) Values below (A11) will be determined using eqn. 5 Since A11 is the first element in the first pattern A12 = A10-1=9 A22 = Smallestlnt[(1250-9*121.5)/100.4] = 1 A32 = Smallestlnt[(1250-9*121.5-1*100.4)/89.8]=0 A42 = Smallestlnt[(1250-9*121.5-1*100.4-0*89.8)/85]=0 A52 = Smallestlnt[(1250-9*121.5-1*100.4-0*89.80*85)/80]=0 A62 = Smallestlnt[(1250-9*121.5-1*100.4-0*89.8-0*850*80)/74]=0 A72=Smallestlnt[(1250-9*121.5-1*100.4-0*89.8-0*850*800*74)70]=0 A82=Smallestlnt[(1250-9*121.5-1*100.4-0*89.8-0*850*800*740*70)/60.1]=0 A92=Smallestlnt[(1250-9*121.5-1*100.4-0*89.8-0*850*800*74-0*700*60.1)/49.8]=1 Step 6 : Cut loss of pattern 2, 0*700*0.1)/49.8)=0 Step 8: The cut loss for the pattern 3. T3=1250(9*121.5+0*100.4+1*89.8+0*85+0*74+0*70+1*60.1+0*49.8) =606mm Searching of Optimal cutting Patterns: The Haussler heuristic procedure will be applied here in finding out the optimal patterns that keep the trim loss to minimum level. The first descriptor is an estimate of the number of production rolls needed to satisfy the remaining requirements. Ist search: 1st Descriptor (∑I RiQi)W =(0*121.5+5*100.4+0*89.8+0*85+64*80+0*74+64*70+2*64 .1+16*48.9) /1250 = 8.815 Mother coils required to satisfy slit coils requirement The second descriptor is the average number of rolls to be obtained from each production roll. i.e. 2nd Descriptor (∑i Ri/[∑i RjWi)/W)) = (0+5+0+0+64+0+64+2+16)/8.815 = 17.1295 avg. no. of rolls from each Mother coil These descriptors of the remaining requirements are used to set an aspiration level for the next pattern to enter the solution. The characteristics used are T2=1250a) Trim Loss: (9*121.5+1*100.4+0*89.8+0*85+0*74+0*70+0*60.1+1*49.8) Maximum allowable trim loss MAXTL = 6.3mm W-∑i AijWi <=MAXTL (Max. allowable trim loss) Step 7: Similarly with the step 4 This will reduce the number of feasible patterns over Row index = 9-7=2 which search will be carried out. We assumed that MAXTL A22>0 should be less than 25 and its all optimal cut patterns are For a new column j=j+1=3 found out from these patterns. A23=A22-1=0 W-∑i AijWi <=25 A13=A12=9 Minimum allowable Trim loss MINTL (allowing the The remaining elements in the pattern are determined by slitting m/c for coil holding purpose) eqn. (5). W-∑i AijWi > = MINTL (Min. allowable trim loss) A33 =Smallestlnt[(1250-9*121.5-0*100.4)/89.8]=1 This is the constraint from the production of slit coils. A A43 =Smallestlnt[(1250-9*121.5-0*100.4-1*89.8)/85]=0 minimum trim loss should be allowed while doing slitting A53 =Smallestlnt[(1250-9*121.5-0*100.4-1*89.8- operation for holding purposes and also to square an 0*85)/80)=0 unevenness in the edges. With the physical set up in External A63 =Smallestlnt[(1250-9*121.5-0*100.4-1*89.8-0*85- Processing agency responsible for the slitting operation, a 0*80)/74)=0 minimum of 5mm is allowed/taken on either side of the A73=Smallestlnt[(1250-9*121.5-0*100.4-1*89.8-0*85mother coil. 0*80-0* W-∑i AijWi >=10 This will further reduces the cut patterns generated. 74)/70]=0 The total trim loss associated with the problem: A83=Smallestlnt[(1250-9*121.5-0*100.4-1*89.8-0*85= (W*Mother coils assigned-trim loss associated with the 0*80-0*74optimal patterns W*Mother coils assigned 0*70)/60.1)=1 = ((1250*9-(T1+T2+T3+T4+T5)) *100 A93 = Smallestlnt[(1250-9*121.5-0*100.4-1*89.8-0*851250*9 0*80-0*74= 2.05% ISSN: 2231-5381 http://www.ijettjournal.org Page 3632 International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 8- August 2013 The periodic inventory model considered the Lead time period of five weeks and a Review period of four weeks is developed for minimizing the indenting quantity. B. Two Dimensional Cutting Stock Method The amount of steel should be indented should include the possible steel loss during the Slitting operation. The Customer orders will be summarized at actuals and at the item level (against different OD) and based on which the number of slit coils required will be determined. The number of slit coils required for the month of April 11 is as follows: Two important procedures was utilized here in finding out the Optimal cutting patterns & number of mother coils against each pattern to meet all the required number of slit coils in order to minimize the trim loss. Finding optimal cutting patterns: The optimal cutting patterns are determined using the algorithm for pattern generation and searching procedure are as follows. Generating feasible cutting patterns: Mother coil width of 1250mm and thickness 40 mm are considered while determining the cutting patterns. Appling pattern generating algorithm in generating feasible cutting patterns. The coefficients (no. of cuts) against each individual slit widths can be determined by using the following equation. Aij = Smallestlnt [W - ∑i=1 (AiWi)/Wi) Step 1: Descending the order of requirement widths (Wi) Step 2: Elements of first row (j=1) Finding no. of cuts from the individual pattern A11 = Smallestlnt[(1250-0/121.5] = 10 A21 = Smallestlnt[(1250-10*121.5)/100.4] = 0 A31 = Smallestlnt[(1250-10*121.5*100.4)/89.8] = 0 A41 = Smallestlnt[(1250-10*121.5-0*100.4-0*89.8)/85] =0 A51 = Smallestlnt[(1250-10*121.5-0*100.4-0*89.80*85)/80]=0 A61=Smallestlnt[(1250-10*121.5-0*100.4-0*89.80*850*80)/74]=0 A71=Smallestlnt[(1250-10*121.5-0*100.4-0*89.80*85-0*800*74)/70]=0 A81=Smallestlnt[(1250-10*121.5-0*100.4-0*89.80*85-0*80-0* 740*70/60.1]=0 A91 = Smallestlnt [(1250-10*121.5-0*100.4-0*89.80*85-0*800*74-0*700*60.1)/49.8]=0 Step 3: Determining the cutting loss along the width using following ISSN: 2231-5381 Cut loss of pattern 1 using eqn. (Tj W - ∑i=1 (AijWi))*t T1=(1250(10*121.5+0*100.4+0*89.8+0*85+0*74+0*70+0*6 0.1+0*49.8)) *40 = 1400mm2. Step 4: The coefficients (no. of cuts) against each individual slit thickness can be determined by using the following equation. Aij = Smallestlnt [t - ∑i=1 (Aiti)/t i) Step 5: Descending the order of requirement thickness (ti) Step 6: Elements of first row (j=1) Finding no. of cuts from the individual pattern B11 = Smallestlnt[(40-0/3.66] = 10 B21 = Smallestlnt[(40-10*3.66)/2.89] = 1 B31 = Smallestlnt[(40-10*3.66-1*2.89)/2.54] = 0 B41 = Smallestlnt[(40-10*3.66-1*2.89-0*2.54)/2.39] = 0 B51 = Smallestlnt[(40-10*3.66-1*2.89-0*2.540*2.39)/2.24]=0 B61=Smallestlnt[(40-10*3.66-1*2.89-0*2.54-0*2.39 0*2.24/2.05]=0 B71=Smallestlnt[(40-10*3.66-1*2.89-0*2.54-0*2.390*2.240*2.05)/1.93]=0 B81=Smallestlnt[(40-10*3.66-1*2.89-0*2.54-0*2.390*2.24-0* 2.050*1.93/1.58]=0 B91 = Smallestlnt [(40-10*3.66-1*2.89-0*2.54-0*2.390*2.24-0*0*2.05-0*1.930*1.58)/0.9]=0 Step 7 : Determining the cutting loss along the thickness using following Cut loss of pattern 1 using eqn. Cv = A11W1 * [1250 – B11t1] = 10*121.5 *[1250 – 10*3.99] = 1470271 mm2. Step 8: set the level index 1 from the last row, where Aij>0 (row index)= 9.8=1 i.e. level-1, 11>0 and others in that row are equal to zero. Step 8: Since the current of level-1(A11) is greater than zero, a new column j=j+1=2 is introduced with the following points. d) Value of (A11) will be decrement by 1 e) Values above (A11) remain same and f) Values below (A11) will be determined using eqn. Since A11 is the first element in the first pattern Optimum Solution There are 14 feasible cutting patterns available to cut raw material with the dimensions 1250 mm *40 mm into required rectangular shaped items. The mathematical model is developed to design generated cutting patterns so that waste (cut loss) will be minimized and the optimum solution to the model is given in Table II: http://www.ijettjournal.org Page 3633 International Journal of Engineering Trends and Technology (IJETT) – Volume 4 Issue 8- August 2013 TABLE-II Journal of Operational Research Society, Vol.39, pp.41-53, 1988. The total trim loss associated with the problem: [3] Farley, A., A note on bounding a class of linear programming problems, including cutting stock problems, Journal of Operations Research, Vol.38, pp.922-923, 1990a. [4] Farley, A., The cutting stock problem in the canvas industry, European Journal of Operational Research, Vol.44, pp.247-255, 1990b. [5] Gilmore, P. C. and Gomory, R. E., Multistage cutting-stock problems of two and more dimensions, Operation Research, Vol.13, pp.94-120, 1965. [6] Gilmore, P. C. and Gomory, R. E., A linear programming approach to the cutting stock problem, [7] Gilmore, P.C., and Gomory, R.E., "A linear programming approach to the cutting stock problem-Part II", Operations Research 11 (1963) 863-888. = (W*Mother coils assigned-trim loss associated with the optimal patterns) W*Mother coils assigned = ((1250*40*9-(T1+T2+T3)) 1250*40*9 = 1.90% *100 [8] Haessler, R.W., "Single-machine roll trim problems and solution procedures", Tappi 59 (1976) 145-149. [9] Haessler, R.W., "A note on some computational modifications to the Gilmore-Gomory cutting stock algorithm", Operations Research 28 (1980) 1001-1005. [10] Pierce, J.F., Some Large Scale Production Scheduling Problems in the Paper Industry, Prentice-Hall, Englewood Cliffs, N J, 1964. [11] Sculli, D., "A stochastic cutting stock procedure: Cutting rolls of insulating tape", Management Science 27(8) (1981)946-952. V.CONCLUSION In this work an algorithm to solve an important variant of the one-dimensional cutting-stock problem and twodimensional cutting-stock problem encountered in paper and steel industries, was proposed. This algorithm first identifies all feasible non dominated combinations through a pattern generation procedure, to construct the constraints matrix. Then a relaxation of the problem is performed in order to obtain a linear formulation, and finally a solution to our original problem is generated from the solution of the relaxed problem. The propose algorithm has been tested on a set of randomly generated problems. The experimental work shows the following performances of our heuristic: a good quality of the solution (the difference between the solution and the lower bound to the optimal solution, is less than 1%) and this heuristic is capable to solve problems with medium size (not only small size) because the search tree does not consider dominated feasible patterns. ACKNOWLEDGMENT The authors thank the referees for their constructive comments, which have contributed to the improvement of this paper. VI.REFERENCES [1] Benati, S., An algorithm for a cutting stock problem on a strip, Journal of the Operational Research Society, Vol.48, pp.288-294, 1997. [2] Farley, A., Mathematical programming models for cutting stock problems in the clothing industry. ISSN: 2231-5381 http://www.ijettjournal.org Page 3634