A review of discrete-time optimization models for tactical production planning
Summary (5 min read)
1 Introduction
- Production planning is related to managing the productive resources required to perform transformation from raw materials to final products to satisfy customers in the most efficient way (Pochet, 2001) .
- These authors also contemplate the inclusion of setup times, multi-level product structures and overtime.
- Moreover, the authors extend these classification criteria by adding the following categories: problem type, modeling approach, solution method, development tool, application, benefits and limitations (Mula et al., 2010a) .
3.1 Problem type
- The MPS establishes an optimal production plan which meets customers' orders, and provides release dates and amounts of final products to manufacture by minimizing production, holding and set up costs.
- Given the increase of integration and coordination among suppliers, manufacturers and distributors, multi-site production planning or SCP has become a very important issue in recent decades.
- Besides, the distinction of different planning levels based on the time period and the amount of detail in the plans is known as HPP, which decomposes the global production planning problem into a number of subproblems that correspond to different levels of a hierarchy of plans.
- Of the reviewed works, 142 of them address MPS problems, followed by a group of 47 references which deal with SCP and 39 references which consider MRP problems.
3.2 Number of products and number of levels
- The complexity of not only the production planning problem, but also its modeling, may be influenced by the number of items manufactured in the production system.
- Among the papers analyzed, 71 references address production planning for a single product, while the remaining 172 correspond to multi-item models.
- Another important characteristic that can affect the complexity of production planning problems is products structure and its number of levels.
- The former corresponds to production systems where only final products are manufactured according to the demand obtained directly from customer orders or market forecasts.
3.3 Time period
- In terms of the time period terminology, multi-item production planning problems fall into the big or small bucket problems categories (Karimi et al., 2003) .
- In small bucket models, only one type of item can be manufactured, or at most one item can be set up per period, while the time period is long enough to produce multiple items in big bucket models.
- Table 4 presents the time period and the product structure details of the references addressing multi-product planning problems.
- With regard to small bucket models, only Stadtler (2011) deals with multi-level planning problems, while those references which present both small bucket and big bucket models (Transchel et al., 2011; Van Vyve and Wolsey, 2006) only consider mono-level product structures.
3.4 Nature of demand
- Demand acts as a typical parameter of production planning models and its nature can affect their complexity.
- If demand levels are known exactly, demand is called deterministic.
- In production planning models, uncertainty is modeled by using probability distributions, fuzzy sets, stochastic approaches based on stochastic values, or several scenarios and robust approaches.
- On the other hand, 207 references consider deterministic demand and 37 uncertain demand levels.
3.5 Capacities or resource constraints
- Capacity constraints may increase the complexity of the production planning models and their resolution, but enable more realistic models.
- The authors identify the constraints related to inventory limitations (48 references), supply of parts and raw materials from suppliers (24 references), productive resources such as machines and workforce (194 references) and transportation resources (21 references).
- A model may have only production capacity constraints, while another might also include limitations related to inventory capacity constraints and/or supply from suppliers.
- Table 6 shows the different combinations associated with each capacity constraint class.
- Most of the analyzed papers (143 references) present only capacity constraints related to productive resources.
3.6.a Demand
- In order to obtain production planning models that come closer to reality, in addition to considering price-dependent demand levels, several extensions related to demand are identified.
- The ability to meet demand through product substitution, the existence of time windows, the option of backlogs to meet demand in following periods, and modeling lost sales if demand cannot be met during the corresponding period or during the subsequent one.
- Table 7 presents the different extensions related to demand considered in this work.
- There is a group of 12 papers that considers product substitutions with several approaches.
- Moreover, one of the reviewed papers (Wolsey, 2006) presents models with production and time windows separately.
Lost sales 30
- Substitution 14 Price-dependent levels 10 Time windows 7 3.6.b Setups Generally, setup activities are included in production planning models by considering the setup costs and/or setup times which model the production changeovers between different products.
- Moreover according to Karimi et al. (2003) , three other setup types of complex setups can be contemplated: setup carry-overs; sequence-dependent setups; family setups.
- The inclusion of setup carry-overs reduces the setup times needed as compared to standard production planning models, which use a setup for each product produced per period.
- Table 8 provides details of the number of references dealing with the considered setup extensions.
- Moreover, Menezes et al. (2011) incorporate sequence-dependent and period-overlapping setup times.
3.6.c Production times
- In order to adjust the capacity usage level of productive resources, production planning models include overtime, subcontracting and undertime decisions.
- If during a period production capacity is less than customer demand, the decision maker may choose to produce in overtime or to outsource part of the production to meet demand without backlogs.
- Of the reviewed works, 35 references consider overtime decisions, 19 include the possibility of subcontracting production, and only 4 references (Fandel and Stammen-Hegene, 2006; Lusa et al., 2009; Mula et al., 2008; Peidro et al., 2010) contemplate modeling idle time.
- Besides, all the references that consider outsourcing decisions model them in terms of the amounts of products to manufacture by subcontractors.
- Table 9 shows the different combinations associated with extensions on production times.
3.6.d Multiple and parallel machines
- According to Quadt and Kuhn (2007) , standard production planning models can represent the existence of parallel machines by augmenting the production variables and the capacity parameters by an additional index indicating the individual machines.
- Finally, 2 references (Jozefowska and Zimniak, 2008; Mateus et al., 2010) deal with production planning problems with parallel unrelated machines (with no particular relationship between the processing times in the different machines).
- Finally, 2 references present conjoined supply chains formed by several suppliers, one manufacturer and several customers (Zolghadri et al., 2008) , and by several plants, a distribution center and several customers (Romero and Vermeulen, 2009) .
- Table 12 details the number of references dealing with the considered remanufacturing extensions.
- The quality of the returned products to be remanufactured is an important aspect to consider when organizing and planning remanufacturing activities.
3.7 Modeling approach
- Since the 1950s, mathematical programming formulations have been proposed for a wide range of production-related problems to address problems of aggregate production planning, lot sizing and detailed short-term production scheduling, among others (Missbauer and Uzsoy, 2011) .
- In general terms, in order to solve multi-objective problems using a standard MILP solver, multi-objective programs are converted into an equivalent MILP model with goal programming or fuzzy programming approaches and their variants.
- Qu and Williams (2008) use a commercial NLP solver with default settings to solve their proposed nonlinear model, whereas Fandel and Stammen-Hegene (2006) present neither a solution procedure nor results.
3.8 Solution approach
- According to Buschkühl et al. (2010) , the approaches to solve different types of production planning or capacitated lot-sizing models can be classified into five groups: mathematical programming-based (MP-based) approaches, Lagrangian heuristics; decomposition and aggregation heuristics; metaheuristics; problem-specific and greedy heuristics.
- Based on the idea that most variables are nonbasic and assume a value of zero in the optimal solution, in theory it is necessary to consider only one subset of variables when solving the problem.
- Column generation can be hybridized with the branch-andbound algorithm to generate a solution method called branch-and-price.
- Metaheuristics have emerged as a result of the extensive application of these heuristic-type algorithms to many optimization problems.
3.9 Development tool
- Sixty-three references do not provide any implementation or development details.
- The second most frequently used tool considered in the reviewed papers is the C programming language and its variants, such as C++ or Visual C, which appears in 49 references.
- Besides, AMPL and GAMS are used mainly with CPLEX solver, and are presented in 11 and 9 references, respectively, while Xpress-MOSEL is presented as the main development tool in Akartunali and Miller (2009) , Akbalik and Pochet (2009) and Transchel et al. (2011) .
3.10 Application
- The proposed models can be validated by using data from real-world production systems or by carrying out numerical experiments based on artificially generated instances.
- Of the 250 papers analyzed, 71 were validated by practical applications in real-world environments and 160 by numerical experiments, 18 of which were inspired in real practices from several industrial sectors.
- Moreover, 14 references do not present any application result.
- Tables 17 and 18 present the industrial areas in which each reference was validated with a practical application or with a numerical experiment inspired in real environments, respectively.
- These tables show the variety of industries in which the reviewed models were validated; sawmills, wood and furniture, automobile and semiconductor and electronic devices industries in the case of practical applications, and processed food, beverages and dairy and pulp and paper industries with regard to numerical experiments, are highlighted.
3.11 Limitations
- Some of the limitations pointed out by the authors of the proposals are related to the solution method used, the considered production systems, demand issues, capacities, the non consideration of uncertain parameters, product properties, applications in non real-world environments, supply chain issues and costs.
- A hundred and two references present limitations related to solution methods.
3.12 Benefits
- Table 20 summarizes the main benefits pointed out by the reviewed references reported by their authors.
- The vast majority (187 references) obtain good solutions in terms of either the CPU time needed or optimality, or they present solution procedures that outperform previous methods in the literature.
- Flexibility in lead times (Bjork and Carlsson, 2007) , routing and processes (Ahkioon et al., 2009) , transport capacity (Hwang et al., 2010) , related to uncertainty or different scenarios (Erromdhani et al., 2012; Leung and Chan, 2009; Schütz and Tomasgard, 2011) and to modeling new constraints (Helber and Sahling, 2010) are an important advantage in dynamic production environments.
- Moreover, the capability of extending their proposed models is emphasized by Li and Meissner (2011) and Stadtler (2011) .
4 Discussion
- After reviewing the selected papers on tactical production planning, this section provides some relevant streams and limitations in the literature on tactical production planning.
- Among them, extensions related to demand and setup properties are those more included in the reviewed models.
- In general, industrial practitioners look for tools whose general purpose is to solve production problems easily without having to learn new modeling or programming languages.
5 Conclusions
- This work surveys 250 articles related to tactical planning in relevant operations research and management journals.
- To study the analyzed works, a classification based on the analysis of the following aspects is proposed: problem type, aim, number of products, time period, nature of demand, capacities constraints, extensions, modeling approach, solution approach, development tool, application, limitations and benefits.
- Like Mula et al. (2010a) , one can confirm the need for optimization models and tools for the production and procurement transport planning processes which contemplate different forms of long-and short-distance transport (railway, air, full truck load, grouping, milk round, routes, etc.) and different characteristics (legal or environmental restrictions).the authors.
- Analytical models based on these conceptual models are a forthcoming work; (2) growing customer requests and increasing competition make demand management an important part of the success and applicability of tactical production planning models.
Did you find this useful? Give us your feedback
Citations
138 citations
Cites background from "A review of discrete-time optimizat..."
...The majority of these surveys deal with multi-item and multi-level problems, some of which are cited here in chronological order: Bahl et al. (1987), Karimi et al. (2003), Pochet and Wolsey (2006), Jans and Degraeve (2008), Buschkühl et al. (2010), Diaz-Madroñero et al. (2014)....
[...]
50 citations
46 citations
36 citations
34 citations
References
41,772 citations
18,439 citations
"A review of discrete-time optimizat..." refers methods in this paper
...…1986), ant colony optimization (Dorigo, Maniezzo, and Colorni 1996), bee colony optimization (Pham et al. 2005), particle swarm optimization (Kennedy and Eberhart 1995; Shi and Eberhart 1998), greedy randomised adaptive search procedure (GRASP) (Feo and Resende 1989), scatter searches and…...
[...]
12,584 citations
"A review of discrete-time optimizat..." refers methods in this paper
...Examples of metaheuristics algorithms include genetic algorithms (Holland 1975), memetic algorithms (Moscato 1989), variable neighbourhood searches (Mladenović and Hansen 1997), simulated annealing (Černý 1985; Kirkpatrick, Gelatt, and Vecchi 1983), tabu searches (Glover 1989, 1990; Glover and…...
[...]
11,224 citations
"A review of discrete-time optimizat..." refers methods in this paper
...…annealing (Černý 1985; Kirkpatrick, Gelatt, and Vecchi 1983), tabu searches (Glover 1989, 1990; Glover and McMillan 1986), ant colony optimization (Dorigo, Maniezzo, and Colorni 1996), bee colony optimization (Pham et al. 2005), particle swarm optimization (Kennedy and Eberhart 1995; Shi and…...
[...]
[...]
5,883 citations
"A review of discrete-time optimizat..." refers methods in this paper
...…variable neighbourhood searches (Mladenović and Hansen 1997), simulated annealing (Černý 1985; Kirkpatrick, Gelatt, and Vecchi 1983), tabu searches (Glover 1989, 1990; Glover and McMillan 1986), ant colony optimization (Dorigo, Maniezzo, and Colorni 1996), bee colony optimization (Pham et al.…...
[...]
Related Papers (5)
Frequently Asked Questions (15)
Q2. What are the future works mentioned in the paper "A review of discrete-time optimization models for tactical production planning" ?
To study the analyzed works, a classification based on the analysis of the following aspects is proposed: problem type, aim, number of products, time period, nature of demand, capacities constraints, extensions, modeling approach, solution approach, development tool, application, limitations and benefits. After conducting this review, the authors indicate some gaps in the literature with some proposed future research lines: ( 1 ) it is important to underline that they found no work that examines multi-level tactical production problems by considering not only the existence of near and offshore suppliers of parts and components, but also the impact that procurement transport may involve on accomplishing production plans. In this sense, like Mula et al. ( 2010a ), the authors can confirm the need for optimization models and tools for the production and procurement transport planning processes which contemplate different forms of long- and short-distance transport ( railway, air, full truck load, grouping, milk round, routes, etc. ) and different characteristics ( legal or environmental restrictions ). Thus, consideration of demand-driven tools and mass customization practices can be an important extension to bear in mind ; ( 3 ) applying tactical planning models in real-world production environments in which uncertain conditions can also be considered ; ( 4 ) real-world industrial problems often have several conflicting objectives.
Q3. What are some of the used approaches to model uncertainty in production planning problems?
Stochastic programming (SP), fuzzy programming (FP), robust optimization (RO) and stochastic dynamic programming (SDP) are some of the most used approaches to model uncertainty in production planning problems (Sahinidis, 2004).
Q4. What are the popular decomposition and aggregation approaches to solve production planning problems?
time-based and resource-based are the most popular decomposition and aggregation approaches to solve production planning problems.
Q5. What is the main reason for the inclusion of setup times in production planning models?
The inclusion of setup times involves reducing the production capacity available per period and increases the models’ complexity because they are usually modeled by introducing zero-one variables.
Q6. What type of heuristics are proposed by Sahling et al.?
Another class of MP-based heuristics called fix-and-optimize heuristics is proposed by Sahling et al. (2009) and Helber and Sahling (2010).
Q7. What are the common solutions proposed in the analyzed papers?
Mathematical programming-based solution procedures and specific solution methods such as heuristic algorithms are proposed in most of the analyzed papers, and in a lesser extent metaheuristics.
Q8. What are the main reasons for the increase in the use of heuristics?
the impossibility of discovering the exact solutions corresponding to optimization problems and the need to respond to the practical situations considered in many real-world cases have led to an increased use of heuristic-type algorithms, which have proven to be valuable tools that provide solutions where exact algorithms do not (Verdegay et al., 2008).
Q9. How can the production planning model represent the existence of parallel machines?
3.6.d Multiple and parallel machinesAccording to Quadt and Kuhn (2007), standard production planning models can represent the existence of parallel machines by augmenting the production variables and the capacity parameters by an additional index indicating the individual machines.
Q10. How many papers address mono-level production planning problems?
among the 169 papers presenting big bucket models, 99 address mono-level production planning problems, while the 70 remaining ones consider multi-level production systems.
Q11. How many references address multi-level product planning?
Of the 177 references addressing production planning for multiple items, 106 references correspond to final products and 71 references consider multi-level product structures.
Q12. What are the main types of remanufacturing activities?
These include the collection of used products, dismantlement or disassembly of returned products, incorporation of remanufacturing activities into the overall production planning (Guide and Van Wassenhove, 2002), and the recycling or disposal of unused products.
Q13. What are the main benefits of the extended lot sizing models?
these papers focus mainly on developing efficient algorithms for typical lot-sizing extensions, such as inclusion of backlogs, setup times, sequence-dependent setups, etc.
Q14. What is the main aspect to consider when planning remanufacturing activities?
The quality of the returned products to be remanufactured is an important aspect to consider when organizing and planning remanufacturing activities.
Q15. What are the main reasons for the classification of production planning problems?
Several authors, such as Anthony (1965), Salomon et al. (1991), McDonald and Karimi (1997), Min and Zhou (2002) and Gupta and Maranas (1999, 2003), among others, classify production planning problems into strategic, tactical and operational problems.