Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 126 (2024) 390–395

17th CIRP Conference on Intelligent Computation in Manufacturing Engineering (CIRP ICME ‘23)

Development of a Linear Integer Programming Model for Solving

Cutting Stock Problem in the Manufacturing Industry

Ilesanmi Daniyana, Sesan Peter Ayodejib, Felix Alec, Adefemi Adeodud, Khumbulani Mpofue

a
Department of Mechanical & Mechatronics Engineering, Achievers University, Owo, Nigeria.
b
Department of Industrial & Production Engineering, Federal University of Technology, Akure, Nigeria.
c
Department of Engineering & Space Systems, National Space Research and Development Agency, Abuja, Nigeria.
b
Department of Mechanical Engineering, University of South Africa, Florida, South Africa.
d
Department of Industrial Engineering, Tshwane University of Technology, Pretoria 0001, South Africa

* Corresponding author. Tel.:+2349031906016; E-mail address: [email protected]

Abstract

Cutting stock problem is a problem generally encountered in many manufacturing industries such as the furniture, clothing, glass production,
leather, paper, textile, metals industries amongst others most especially during mass production. The problem usually arises during roll slitting
whereby a large role is to be cut into smaller pieces. The challenge may be further compounded due to constraints and different variants that
arises from product customization, process and machinery as well as customer requirements and quality issues. To address this problem, this
study develops a linear integer programming technique and provides a practical guided approach. First mathematical formulation (having
objective function and constraints) comprising of a list of q orders, that requires 𝑝𝑝𝑗𝑗 pieces was formulated and a list of all possible cuts
combination of configuration and patterns were derived. The linear integer program was solved in the MATLAB 2022 environment. The results
indicated that the large rolls can be cut into the desired length, number, patterns and configuration in a time effective manner using the
developed linear integer programming technique. The results further show that for the total wastes generated was 24 for the optimal solutions.
This number of wastes can be considered to be minimal considering the volume of demand and patterns generated from the stock material.
Furthermore, 6 out of the 7 pattern optimal solutions generated zero waste. This study provides theoretical and empirical findings that can assist
the manufacturing industries in minimizing time and cost variables while addressing stock cutting problems.
© 2024 The Authors. Published by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the 17th CIRP Conference on Intelligent Computation in Manufacturing Engineering
(CIRP ICME‘23)

Keywords: Cutting stock problem; Linear integer programming model; Manufacturing industries; Mathematical formulation

1. Introduction The cutting operation is one of common manufacturing

processes, for product development in the furniture, clothing,
The manufacturing industries are faced with the challenges glass production, leather, paper, textile, metals industries
of time and cost effectiveness of the manufacturing process. amongst others. The cutting operation involves the cutting of
To meet the bottom line of profitability and sustainability, stock materials into predetermined sizes for product
they must embrace decision-making tools for effective development. However, the determination of the number,
resource allocation and usage. This is important in dimensions, configurations or patterns of cut from a stock
manufacturing lead-time and achieving product quality [1-3]. material is critical decision, which must be carefully

2212-8271 © 2024 The Authors. Published by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the 17th CIRP Conference on Intelligent Computation in Manufacturing Engineering
(CIRP ICME‘23)
10.1016/j.procir.2024.08.382
 Ilesanmi Daniyan et al. / Procedia CIRP 126 (2024) 390–395 391

considered to avoid waste. The cutting-stock problem is a heuristic based model [20] as well as decomposition
problem of meeting an order at minimum cost for a specified approaches [21].
number of material dimensions to be cut from a given stock The genetic algorithm performs the best among them,
and cost [4]. If the material is cut manually, significant wastes demonstrating the value of striking a balance between local
may be generated in addition to quality issue. This may make convergence and global search for enhancing outcomes. In
the cutting process less sustainable. For mass production, it is order to calculate the 1-D CSP of a single specification and
more time and cost effective to employ decision-making tools many specifications, Zhu [22] not only created a genetic
such as the linear programming approach to determine the algorithm with fixed-length real coding, but also established
optimum dimensions of cut from a sheet metal. penalties for impractical solutions with individual fitness
Generally, cutting stock problem is a problem that is often larger than 1. This approach produced superior outcomes. Cui
encountered in many manufacturing industries most et al. [23] and Ma et al. [24] suggested a heuristic approach
especially during mass production. The problem usually arises based on column generation to handle the issue of available
during roll slitting whereby a large roll is to be cut into surplus materials in the 1-D BPP for the 1DCSP in a random
smaller pieces. It Is mostly one dimensional and a environment. Sequential value correction (SVG) was used to
combinatorial optimization problem with NP-hard study the 1-D CSP by Belov et al. [25] and Cao et al. [26].
characteristics. Traditional approaches such as heuristic and Recent studies on the 1-D BPP [27-29] focused on improving
genetic algorithms have limitations in solving the cutting meta-heuristics and heuristic algorithms based on the use of
stock problem (CSP) efficiently for small-scale problems. useful remaining raw materials, which entails creating a
Fang et al. [5], solved a one dimensional cutting stock cutting pattern by sorting the items that are a pair or an odd
problem by developing a mathematical model is to minimize length in descending order. However, there are typically
raw material consumption and maximize remaining material millions of cutting methods for the large-scale 1-D CSP. As a
length. Additionally, a novel algorithm based on deep result, it is inefficient to tackle the problem using traditional
reinforcement learning (DRL) was proposed. The DRL algorithms. Furthermore, rules that are created for certain
algorithm comprises two modules: a pointer network with cutting stock situations are necessary for heuristic algorithms
encoder and decoder structure serves as the policy network, and improved-heuristic algorithms. Such standards require the
leveraging the shared underlying mode of 1-D CSP, while a highest level of professional design expertise. The heuristic
model-free reinforcement learning algorithm trains network method can find a variety of answers; however, it is not
parameters and optimizes the cutting sequence. Experimental always able to find an optimal solution. Additionally, there
results demonstrate that the DRL-based one-dimensional are not many straightforward, useful, and adaptable solutions
cutting stock algorithm (DRL-CSP) achieves approximate to the 1-D bin packing problem (BPP) of large-scale and
satisfactory solutions for 82 instances across three datasets in repeated batches in actual production, which leaves the 1-D
a short time. Moreover, the algorithm exhibits strong CSP mostly in the condition of experience cutting [30]. The
generalization performance and practical application recent development of machine learning has seen extensive
potential. The challenge may be further compounded due to adoption across numerous industries. Amazing advancements
constraints and different variants that arises from product in path optimization and scheduling [31-32] and artificial
customization, process and machinery as well as customer intelligence and operations research [33–36] are revealed. The
requirements and quality issues. Generally, the objectives of majority of machine learning approaches to combinatorial
solving a cutting stock problem are to minimise waste and optimization issues concentrate on teaching the creation of
costs as well as to maximize the number of items used to heuristic algorithms. A certain set of input nodes can produce
make profit [4]. To address this problem, this study develops solutions. Sequences are frequently used to represent these
a linear integer programming technique and provides a strategies [37]. High-quality solutions are generated by the
practical guided approach. attention mechanism [38] and graph neural network [39].
According to Dyckhoff [6], cutting stock problems can be However, the construction approach needs other algorithms,
categorised into two groups: pattern-oriented and item- such as cluster search, classical enhanced heuristic algorithm,
oriented and can be further classified into dimensionality, and random sampling, to get advanced performance. Some
kind of assignment, assortment of small and large objects. cutting stock problems based on reinforcement learning can
Some of the approaches employed for solving cutting stock be solved using heuristic algorithms and learning rules. In
problems include integer programming [7], hybrid approach addition to proposing a discounted infinite-horizon Markov
such as the combination of the pattern-oriented linear decision process formulation of the stochastic cutting stock
programming model and item-oriented sequential heuristic problem, Pitombeira-Neto et al. [40] also put out a heuristic
approach [8], have now been developed by academics for the approach based on reinforcement learning. According to the
1-D CSP [11–16]. findings, the typical cost might be up to 80% less than the cost
To solve the one dimensional cutting stock problem, existing attained by a myopic policy. The earlier study by Fang et al.
works have also reported on the use many optimiszation [5] also demonstrates that the cutting stock problem can be
strategies including linear programming, the simplex method, solved using a mix of reinforcement learning search and
dynamic programming, heuristic algorithms, genetic heuristic algorithm, although there are some restrictions, such
algorithms, simulated annealing algorithms, and evolutionary as the randomness of the search. DRL, the widely used
algorithms [9-14], specifically the cutting plane algorithm machine learning technique, has been effectively used in a
[15], mixed integer linear model [16], modified branch-and variety of applications, including image analysis [41], video
bound [17], metaheuristics approach which integrates an categorization [42], intelligent translation [43], and others.
adaptive pattern generation technique [18], branch-and-cut- The algorithm's findings are evaluated and contrasted, and it
and-price algorithm [19], dynamic programming-based is discovered that DRL has the potential for strong scalability,
392 Ilesanmi Daniyan et al. / Procedia CIRP 126 (2024) 390–395

generalization, and adaptability. The traveling salesman patterns having a positive integer variable 𝑥𝑥𝑖𝑖 . 𝑥𝑥𝑖𝑖 represent the
problem (TSP) and knapsack problem were both solved using number of times pattern 𝑖𝑖 is to be used (In other words, is the
the pointer network model that Vinyals et al. [44] introduced number of roll cuts using pattern i).
as the first contemporary depth model for combinatorial Therefore, 𝑖𝑖 = 1, … … . . 𝐺𝐺
optimization issues. Bello et al. [45] employed reinforcement The objective function 𝑍𝑍 is to minimise the cost of producing
learning training to solve the sequence problem based on the pattern 𝑖𝑖. The cost of producing pattern 𝑖𝑖 is denoted as 𝑔𝑔𝑖𝑖 and
pointer network, which not only reduced computational costs may be in the form of waste or scrap materials or the cost
but also produced the ideal answer for the large-scale TSP incurred due to tardiness. This is expressed as Equation 1.
problem. Raffensperger [46] proposed two one dimensional 𝐺𝐺
cutting stock models namely the generalized assortment and
best cutting stock length problems. The novel models show 𝑍𝑍 = 𝑚𝑚𝑚𝑚𝑚𝑚 ∑ 𝑔𝑔𝑖𝑖 𝑥𝑥𝑖𝑖 (1)
capability for waste reduction and selection of a small fixed 𝑖𝑖=1

set of stock lengths. Attempts have also been made to

combine the column generation and branch-and-bound This is subject to the number of order 𝑗𝑗 which appears in
algorithm to obtain exact solution to cutting stock problems pattern 𝑖𝑖. This constraint is expressed as Equation 2.
[47-51]. However, when the cutting stock problem is
expressed as an integer programming problem, large number 𝐺𝐺
of variables may make the computation infeasible most ∑ 𝑏𝑏𝑖𝑖𝑖𝑖 𝑥𝑥𝑖𝑖 ≥ 𝑝𝑝𝑗𝑗 , ∀𝑗𝑗 = 1, … … . 𝑞𝑞 (2)
especially for mass production. The same challenge applies to 𝑖𝑖=1
an approximate solution by linear programming. Although
efforts have been made to develop robust algorithms and 𝑥𝑥𝑖𝑖 integer
technique for solving this kind of challenge but the challenge
still remain a major issue in cutting stock problem. The Sets
objective of this study is to determine how sheet metal rolls Q is the set of order ( 𝑗𝑗 = 1, … . . 𝑞𝑞 )
can be cut into smaller sizes or patterns to meet the demand G is the set of patterns (𝑖𝑖 = 1, … … . . 𝐺𝐺)
and minimize the amount and cost of scrap.
To address this problem, this study develops a linear integer Parameters
programming technique and provides a practical guided 𝑏𝑏𝑖𝑖𝑖𝑖 is the number of rolls order q having pattern g
approach for solving a cutting stock problem of large stock of 𝑝𝑝𝑗𝑗 demand for order the cut pieces
cuts and patterns. The novel algorithm developed for solving
Decision Variables
this problem was implemented in the MATLAB 2022
𝑥𝑥𝑖𝑖 is the number of roll cuts using pattern i
environment and the results obtained indicated that it is The objective function is to minimize the number of rolls cut
feasible for solving large cutting stock problem. This study to meet the demand and minimize the amount and cost of
provides theoretical and empirical findings that can assist the scrap.
manufacturing industries in minimising time and cost In this example, the length of the sheet metal is 100 mm, the
variables while addressing stock cutting problems. The guillotine cuts the sheet metals into fixed lengths for further
novelty of this study lies in the fact that the algorithm
processing. The problem is how to make the cuts so that the
employed for solving the cutting stock problem is unique
sheet metal satisfies a set of orders with the fewest sheet metal
providing exact solutions with high computational time.
The production specification is presented in Table 1.
The first section of the paper provides the motivation for the
use of the linear programming approach to solve the cutting
Table 1. Production specification for stock cutting.
stock problem. The second section presents the mathematical S/N Length of sheet metal (mm) Quantity demanded
formulation for the linear programming approach while the 1 10 29
third section presents the results and discussion. Finally, the 2 12 90
conclusive remarks and future areas of research are provided 3 16 110
in the conclusion. 4 18 54
5 20 30
2. Methodology 6 24 35
7 30 46
Total 130 394
This section explains the mathematical formulation and the
procedural steps for solving the cutting stock problem using a
From Table 1, there are different possible combinations of
linear integer programming approach.
lengths that the sheet metal roll can be cut into that will sum
up into 100 mm or less without any scrap. Each such possible
2.1 Cutting stock problem and mathematical formulation
combination is called a pattern. The determination of the
number of each pattern to cut to meet the demand while
First mathematical formulation (having objective function and minimizing the amount and cost of scrap cannot be easily
constraints) comprising of a list of q orders, that requires 𝑝𝑝𝑗𝑗 achieved manually with precision, hence the need for
pieces (where 𝑗𝑗 = 1, … . . 𝑞𝑞) was formulated and a list of all optimization.
possible cuts combination of configuration and patterns were Without optimisation, the direct cutting of the length and the
derived. The linear integer program was solved in the quantity demanded from the sheet metal, may result in time
MATLAB 2022 environment. and materials wastage. This will make the cutting process less
Let G be the number of the required patterns with each
 Ilesanmi Daniyan et al. / Procedia CIRP 126 (2024) 390–395 393

sustainable. It will be more efficient to firsts generate the 6. If the optimal objective value 𝐹𝐹 𝐹 𝐹, then the values
cutting patterns as the solution of a subproblem and solve the of integer variable 𝑦𝑦𝑖𝑖 indicates the number of rolls
linear programming problem with the objective function to of the new pattern.
minimise the number of sheet metals used subject to the 7. Add the new pattern and solve the linear
constraint that the cuts satisfy the demand, using the existing programming problem. Otherwise, the linear
patterns. After solving the problem, a new pattern will be programming relaxation solution of the cutting stock
generated by solving an integer linear programming problem is considered to be optimal.
subproblem. The essence of the subproblem is to find the best 8. If the linear programming relaxation solution is
new pattern, comprising of number of cuts from each length integer, then stop. Otherwise, an advanced algorithm
in the length list whose sum does not exceed the total length can be employed to obtain an integer solution.
of the sheet metal. The optimised quantity is the reduced cost
of the new pattern, and this is a minimisation problem. 3. Results and Discussion

The optimization presents 89 optimal solutions as presented in

2.2 Optimisation procedural steps Table 2 while Table 3 presents the quantity demanded vs
quantity produced. Table 2 indicates that for the total wastes
1. Select the initial set of patterns. generated was 24 for the optimal solutions. This number of
2. Solve the linear programming relaxation of the wastes can be considered minimal considering the volume of
cutting stock problem demand and patterns generated from the stock material.
3. Relax the initial 𝑥𝑥𝑖𝑖 integer as 𝑥𝑥𝑖𝑖 ≥ 0 in the Furthermore, 6 out of the 7 pattern optimal solutions
mathematical formulation and solve the resulting generated zero waste. This implies that the proposed solution
linear program. is feasible in solving large cutting stock problem in a time and
4. Generate a knapsack problem to generate new cost effective manner. When the item material’s length is
patterns. For instance, Let 𝑧𝑧𝑖𝑖 be the dual variable relatively long with high quantity demand, the waste
corresponding to 𝑔𝑔𝑖𝑖 in the linear programming generated may be high [46]. However, this waste can be
relaxation, let 𝑦𝑦𝑖𝑖𝑖𝑖 be the number of rolls order having further reduced by implementing the generalized assortment
pattern new pattern, and 𝑎𝑎𝑖𝑖𝑖𝑖 be the demand for order model can help reduce or eliminate waste. However, all the
the cut pieces. solutions generated were optimal and feasible. Table 3 also
indicated that the demand is either met or exceeded. This
5. Let L be the uniform length of the rolls.
further implies that the developed model is suitable for
Therefore, the problem of generating new patterns is meeting the demand requirement.
formulated as equations 3 and 4. Now that there is a solution of the linear programming
problem, to complete the solution and further eliminate waste,
𝑍𝑍
there might be a need to solve the problem again with the
𝐹𝐹 𝐹 𝐹𝐹𝐹𝐹𝐹𝐹 ∑ 𝑧𝑧𝑖𝑖 𝑦𝑦𝑖𝑖 (3) final patterns with all variables being integers. The amount of
𝑖𝑖𝑖𝑖 unused sheet metals and scraps represent for each pattern and
for the problem as a whole.
This is subject to Equation 4. The results obtained indicated that from a stock of 100 mm,
the total fixed length is 130 and there are 394 total demanded
𝐺𝐺 quantities. There are 71 possible patterns for this example.
The optimisation gave 89 solutions with a 0.40% reduction in
∑ 𝑦𝑦𝑖𝑖𝑖𝑖 𝑥𝑥𝑖𝑖 ≤ 𝐿𝐿𝐿 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿
waste. It can be shown computationally that the quantity
𝑖𝑖𝑖𝑖
produced exceeded the quantity demanded by 3.04%. These
results lend credence to the fact that the implementation of the
𝑦𝑦𝑖𝑖 integer linear integer programming method can be used to reduce or
eliminate waste in cutting stock problem.

Table 2. Results of optimal solutions.

Optimal solutions
Cut patterns Cut 1 log Cut 6 logs Cut 2 Cut 16 Cut 11 logs Cut 9 logs Cut 25 logs Cut 1 log
with with logs logs with with with pattern with pattern with
pattern pattern with pattern pattern pattern
pattern

Number of 10 cuts of 5 cuts of 4 cuts 1 cut of 1 cut of 1 cut of length 3 cuts of 7 cuts of
cut lengths length 10 length 20 of length 10 length 10 12 and 1 cut of length 12 and length 12
length and 3 cuts and 5 cuts length 16 and 4 cuts of and 1 cut of
24 of length of length 3 cuts of length 16 length 16
30 18 length 24

Waste 0 0 24 0 0 0 0 0
generated
394 Ilesanmi Daniyan et al. / Procedia CIRP 126 (2024) 390–395

Table 3. Quantity demanded vs quantity produced. Research, 1990, 44:145-159.

[7] Waescher, G. & Gau, T. (1996). Heuristics for the integer one-
S/N Length of sheet Quantity Quantity
dimensional cutting stock problem. A Computational Study. OR
metal demanded demanded produced Spektrum 18(3):131-144.
(mm) [8] Gradiar, M., Kljaji M., Resinovi, G. & Jesenko J. A sequential heuristic
1 10 29 37 procedure for one-dimensional cutting. European Journal of
2 12 90 91 Operational Research, 1999, 114(3):557-568.
3 16 110 110 [9] Kang, M. & Yoon, K. An improved best-first branch-and-bound algorithm
for unconstrained two-dimensional cutting problems. Int. J. Prod. Res.
4 18 54 55
2011, 49, 4437–4455.
5 20 30 30 [10] Lu, H. & Huang, Y. An efficient genetic algorithm with a corner space
6 24 35 35 algorithm for a cutting stock problem in the TFT-LCD industry. Eur. J.
7 30 46 48 Oper. Res. 2015, 246, 51–66.
Total 130 394 406 [11] Haessler, R. & Sweeney, P. Cutting stock problems and solution
procedures. Eur. J. Oper. Res. 1991, 54, 141–150.
[12] Wäscher, G. & Gau, T. Heuristics for the integer one-dimensional
cutting stock problem: A computational study. Oper. Res. Spektrum
1996, 18, 131–144.
4. Conclusion [13] Wu, Z., Zhang, L. & Wang, K. An Ant Colony Algorithm for one-
dimensional cutting-stock problem. Mech. Sci. Technol. Aerosp. Eng.
2008, 27, 1681–1684.
The aim of this study is to solve a cutting stock problem of large [14] Guan, W., Gong, J. & Xue, H. A hybrid Heuristic Algorithm for the one-
stock in a time and cost effective manner. To achieve this, a linear dimensional cutting stock problem. Mach. Des. Manuf. 2018, 8, 237–
integer programming technique with a robust algorithm was 239.
employed. First mathematical formulation was formulated and a list [15] Scheithauer, G., Terno, J., Muller, A. & Belov, G. Solving one-
of all possible cuts combination of configuration and patterns were dimensional cutting stock problems exactly with a cutting plane
algorithm. Journal of the Operational Research Society, 2001, 52, 1390-
derived. The linear integer program was solved in the MATLAB
1401.
2022 environment. The results indicated that the large rolls can be [16] Johnston, R. E. & Sadinlija, E. A new model for complete solutions to
cut into the desired length, number, patterns and configuration in a one-dimensional stock problems. European Journal of Operational
time effective manner using the developed linear integer Research, 2004, 153, 176-183.
programming technique. This implies that the proposed solution is [17] Mukhacheva, E. A., Belov, G., Kartak, V. & Mukhacheva, A. S. One-
feasible in solving large cutting stock problem in a time and cost dimensional cutting stock problem: Numerical experiments with the
effective manner. The 24 waste generated can be traced to the item sequential value correction method and a modified branch-and-bound
material’s length. When the material’s length is relatively long with method. Pesquisa Operacional, 2001, 2:153-168.
[18] Umetani, S., Yagiura, M. & Ibaraki, T. One-dimensional cutting stock
high quantity demand, the waste generated may be high. However,
problem to minimize thenumber of different patterns. European Journal
this waste can be further reduced by implementing the generalized of Operational Research, 2003, 146, 388-402.
assortment model can help reduce waste. All the solutions generated [19] Belov, G. & Scheithauer, G. A cutting plane algorithm for the one-
were optimal and feasible. Thus, this study provides theoretical and dimensional cutting stock problem with multiple stock lengths.
empirical findings that can assist the manufacturing industries in European Journal of Operational Research, 2002, 141:274-294.
minimising time and cost variables while addressing stock cutting [20] Kellerer, H., Pferschy, U. & Pisinger, D. Knapsack Problems. Springer-
problems. It is therefore recommended that manufacturing industries Verlag Berlin Heidelberg, 2004.
faced with cutting stock problem consider the implementation of the [21] Holthaus, O. Decomposition approaches for solving the integer one-
dimensional cutting stock problem with different types of standard
linear integer programming method to reduce or eliminate waste
lengths. European Journal of Operational Research, 2002, 141: 295–
during cutting operations. It can also be employed as a decision 312.
support technique for addressing cutting stock related problems. [22] Zhu, S. The Research on Optimization Algorithms for one-Dimensional
Future works can consider the implementation of the Cutting Stock Problems. Master’s Thesis, Huazhong University of
generalized assortment model can help reduce or eliminate Science and Technology, Wuhan, China, 2013.
waste. [23] Cui, Y., Song, X. & Chen, Y. New model and heuristic solution
approach for one-dimensional cutting stock problem with usable
leftovers. J. Oper. Res. Soc. 2017, 68, 269–280.
[24] Ma, J., Han, Z., Luo, D., Xiao, H. Research on one-dimensional cutting
stock problem based on Recursive Matrix Column Generation
Algorithm. Mach. Des. Manuf. 2022, 117–119.
References [25] Belov, G. & Scheithauer, G. Setup and open-stacks minimization in one-
dimensional stock cutting. INFORMS J. Comput. 2007, 19, 27–35.
[1] Daniyan, I. A., Adeodu, A. O., Mpofu, K., Maladhzi, R & Kana-kana [26] Cao, J., Cui, Y. & Li, D. Study on the solution of one-dimensional
cutting stock for multiple stock lengths with variable cross-section.
Katumba. M. G. Application of lean Six Sigma methodology using
Forg. Stamp. Technol. 2017, 42, 161–165.
DMAIC approach for the improvement of bogie assembly process in the
[27] Cerqueira, G., Aguiar, S. & Marques, M. Modified Greedy Heuristic for
railcar industry. Heliyon, 2022, (8) e09043:1-14.
the one-dimensional cutting stock problem. J. Comb. Optim. 2021, 42,
[2] Adeodu, A. O., Maladzhi, R., Kana-kana Katumba, M. G. & Daniyan, I.
657–674.
A. Development of an improvement framework for warehouse processes
[28] Ravelo, S., Meneses, C. & Santos, M. Meta-heuristics for the one-
using lean Six Sigma (DMAIC) approach. A case of third party logistics
dimensional cutting stock problem with usable leftover. J. Heuristics
(3pl) services. Heliyon, 2023 (9) e14915: 1-19.
2020, 26, 585–618.
[3] Sibanda, V., Mpofu, K., Daniyan, I. A. & Ale, F. Reconfiguration ramp-
[29] Pimenta, Z., Sakuray, F. & Hoto, R. A heuristic for the problem of one-
up cost analysis for a reconfigurable guillotine shear and bending press
dimensional steel coil cutting. Comput. Appl. Math. 2021, 40(39):1-18.
machine. Cogent Engineering, 2022, 9 (2085002):1-24.
[30] Tian, S., Lv, L.& Cai, Y. Design and implementation of a simple
[4] Gilmore, P. C. & Gomory, R. E. A linear programming approach to the
algorithm for solving one dimensional cutting block problem based on
cutting-stock problem. Operations Research, 1961, 9(6):849-859.
Lingo. Ind. Sci. Trib. 2021, 20, 45–47.
[5] Fang, J.; Rao, Y.; Luo, Q.; Xu, J. Solving one-dimensional cutting stock
[31] Zhang, C., Song, W., Cao, Z., Zhang, J., Tan, P. & Xu, C. Learning to
problems with the Deep Reinforcement Learning. Mathematics 2023,
dispatch for job shop scheduling via Deep Reinforcement Learning.
11(4):1028.
Adv. Neural Inf. Process. Syst. 2020, 33, 1621–1632.
[6] Dyckhoff, H. A typology of cutting and packing problems. European
[32] Park, J., Chun, J., Kim, S. & Kim, Y. Learning to schedule job-shop
Journal of Operational
 Ilesanmi Daniyan et al. / Procedia CIRP 126 (2024) 390–395 395

problems: Representation and policy learning using graph neural

network and reinforcement learning. Int. J. Prod. Res. 2021, 59, 3360–
3377.
[33] Li, J., Ma, Y., Gao, R., Cao, Z., Lim, A., Song, W. & Zhang, J. Deep
Reinforcement Learning for solving the heterogeneous capacitated
vehicle routing problem. IEEE Trans. Cybern. 2022, 52, 13572–13585.
[34] Xin, L., Song, W., Cao, Z. & Zhang, J. Step-Wise Deep Learning
Models for solving routing problems. IEEE Trans. Ind. Inform. 2021,
17, 4861–4871.
[35] Kool, W., Van, H. & Welling, M. Attention, learn to solve routing
problems. In Proceedings of the 7th International Conference on
Learning Representations, ICLR 2019, New Orleans, LA, USA, 6–9
May 2019.
[36] Xin, L., Song, W., Cao, Z., Zhang, J. NeuroLKH: Combining Deep
Learning Model with Lin-Kernighan-Helsgaun Heuristic for
solving the traveling salesman problem. Adv. Neural Inf. Process. Syst.
2021, 34, 7472–7483.
[37] Ivanov, D., Kiselev, M. & Larionov, D. Neural Network optimization for
Reinforcement Learning tasks using sparse computations. arXiv 2022,
arXiv:2201.02571.
[38] Zhou, R., Tian, Y., Wu, Y. & Du, S. Understanding curriculum learning
in policy optimization for solving combinatorial optimization problems.
arXiv 2022, arXiv:2202.05423.
[39] Peng, B., Wang, J. & Zhang, Z. A Deep Reinforcement Learning
Algorithm Using Dynamic Attention Model for Vehicle Routing
Problems. Commun. Comput. Inf. Sci. 2020, 1205, 636–650.
[40] Pitombeira-Neto, A.R.; Murta, A.H.F. A reinforcement learning
approach to the stochastic cutting stock problem. Eur. J. Comput.
Optim. 2022, 10, 100027.
[41] Zhang, W., Tang, S., Su, J., Xiao, J. & Zhuang, Y. Tell and guess:
Cooperative learning for natural image caption generation with
hierarchical refined attention. Multimed. Tools Appl. 2021, 80, 16267–
16282.
[42] Xia, B., Wong, C., Peng, Q., Yuan, W. & You, X. CSCNet: Contextual
semantic consistency network for trajectory prediction in crowded
spaces. Pattern Recognit. 2022, 126, 108552.
[43] Song, J., Kim, S. & Yoon, S. AligNART: Non-autoregressive Neural
Machine Translation by Jointly Learning to Estimate Alignment
and Translate. In Proceedings of the 2021 Conference On Empirical
Methods in Natural Language Processing (EMNLP 2021), Punta Cana,
Dominican Republic, 7–11 November 2021; pp. 1–14.
[44] Vinyals, O., Fortunato, M. & Jaitly, N. Pointer networks. Adv. Neural
Inf. Process. Syst. 2015, 28, 2692–2700.
[45] Bello, I., Pham, H., Le, Q., Norouzi, M. & Bengio, S. Neural
combinatorial optimization with reinforcement learning. In Proceedings
of the 5th nternational Conference on Learning Representations, ICLR
2017 Workshop Track Proceedings, Toulon, France, 24–26 April 2017.
[46] Raffensperger, J. F. The generalized assortment and best cutting stock
length problems. International Transactions in Operational Research,
2010, 17: 35–49.
[47] Vance, P., Barnhart, C., Johnson, E. L. & Nemhauser, G. L. Solving
binary cutting stock problems by column generation and branch-
and-bound. Computational Optimization and Applications, 1994, 3, 111-
130.
[48] Vance, P. Branch-and-price algorithms for the one-dimensional cutting
stock problem. Computational Optimization and Applications, 1998, 9,
211-228.
[49] Vanderbeck, F. Computational study of a column generation algorithm
for bin packing and cutting stock problems. Mathematical Programming
A, 1999, 86, 565-594.
[50] Vanderbeck, F. On DantzigWolfe decomposition in integer
programming and ways to perform branching in a branch-and-price
algorithm. Operations Research, 2020, 48:111-128.
[51] Valerio de Carvalho, J. M. (1999). Exact solution of bin-packing
problems using column generation and branch-and-bound. Annals of
Operation Research, 86:629-659.