You are on page 1of 21

International Journal of Production Research, 2018

Vol. 56, No. 13, 4565–4584, https://doi.org/10.1080/00207543.2017.1401234

A survey of semiconductor supply chain models part III: master planning, production
planning, and demand fulfilment
Lars Möncha*, Reha Uzsoyb and John W. Fowlerc
a
Department of Mathematics and Computer Science, University of Hagen, Hagen, Germany; bEdward P. Fitts Department of
Industrial and Systems Engineering, North Carolina State University, Raleigh, NC, USA; cDepartment of Supply Chain Management,
Arizona State University, Tempe, AZ, USA
(Received 23 December 2016; accepted 23 October 2017)

Part I of this three-part series described semiconductor supply chains from the decision-making and functional perspec-
tives, using this as a framework to review the industrial engineering (IE) and operations research (OR) literature on the
problems arising in these supply chains. Part I then reviewed the literature on Strategic Network Design, Supply Chain
Coordination, Sustainability and Semiconductor Supply Chain Simulation, while Part II reviewed Demand Planning,
Inventory Management, and Capacity Planning. This paper concludes the series, discussing Master Planning, Production
Planning, Demand Fulfilment, and Available to Promise (ATP) in semiconductor supply chains.
Keywords: supply chain management; semiconductor industry; production planning; survey; future research

1. Introduction
The constantly increase in use of integrated circuits in industrial, commercial and military products renders the semicon-
ductor industry of critical importance to the global economy. This industry has had a global footprint since very early in
its history, and deploys some of the most technologically complex production processes in industrial use today. While
the operations management problems of this industry have been studied extensively (Uzsoy, Lee, and Martin-Vega
1992, 1994; Mönch, Fowler, and Mason 2013), a comprehensive review of the literature on supply chain management
in this industry has not yet been attempted.
This series of three papers reviews the IE and OR literature related to the design, planning and control of the semi-
conductor supply chain using the functional decision-making perspective of Meyr, Wagner, and Rohde (2015), shown in
Figure 1. Part I of this series described the semiconductor supply chain from this perspective and reviewed the literature
on Strategic Network Design, supply chain coordination, sustainability, and simulation-based decision support. Part II
reviewed Demand Planning, Inventory Management, and Capacity Planning. This paper concludes the series by dis-
cussing Master Planning, Production Planning, and Demand Fulfilment in this industry. Boxes framed in red in Figure 1
indicate the planning functions discussed in the remainder of this paper.
The boundary between Master Planning, which seeks to determine capacitated requests for all end items sold by the
firm into the market, and Production Planning, which determines a capacity-feasible work release schedule for individual
production facilities that allows them to meet the output targets determined in the Master Plan, can be defined in differ-
ent ways. Rather than impose what would necessarily be a somewhat arbitrary separation between these domains, we
have elected to review this literature as a single body in Section 2. Section 3 considers models supporting the Demand
Fulfilment and Available to Promise (ATP) functions, which represent the interface between the supply chain and its
external customers. A summary and conclusions are given in Section 4.

2. Master planning and production planning


2.1 Problem setting
In the framework of Meyr, Wagner, and Rohde (2015), Master Planning takes a firm-wide, medium term view of the
problem of what to produce, where and how, and is usually performed at a level of aggregation considering product
families and aggregated capacity at potentially constraining tool groups. The principal output is the planned production

*Corresponding author. Email: Lars.Moench@fernuni-hagen.de

© 2017 Informa UK Limited, trading as Taylor & Francis Group


4566 L. Mönch et al.

Procurement Production Distribution Sales

long- Strategic Network Design


term

Capacity Planning
Demand
Planning
mid-
term
Master Planning

Inventory
Manage-
Material ment
Require- Production Planning
ments
short- Planning
Demand
term
Fulfillment

Transport
Scheduling
Planning

Figure 1. SC planning matrix for semiconductor supply chains.

and inventory quantities for each product or product family at each facility in each period. Important tasks of the Master
Planning function are determining how the supply chain will respond to seasonal demand fluctuations (e.g., by building
inventory ahead of demand peaks, backordering, outsourcing), how limited production capacity will be allocated among
competing product families and markets, and the selection of material sources and manufacturing processes from among
the available alternatives. Hence, to be effective, Master Planning must comprehend a planning horizon encompassing at
least one full seasonal demand cycle, making it a mid-term activity in Figure 1.
Master Planning takes firm orders and demand forecasts as inputs from Demand Planning, and provides input to
Demand Fulfilment, ATP, and Production Planning. It may also inform short-term capacity planning for resources with
relatively short procurement lead times, such as staffing. However, due to the long acquisition lead times for semicon-
ductor manufacturing equipment discussed in Part I, Figure 1 shows the Capacity Planning function taking place on a
longer time scale than Master Planning. Although seldom explicitly mentioned, the Master Planning activity must also
implement the firm’s safety stock strategy as discussed in Part II, specifying the locations and quantities of safety stocks
to be held.
The aggregate, medium-term nature of the Master Planning activity requires more detailed plans at the level of
specific end items for the individual production facilities; specific material and resource allocations decisions are
required within the facilities for production to be realised. In the framework of Meyr, Wagner, and Rohde (2015), the
Production Planning function seeks a capacity feasible release plan specifying how much of each item must be released
to production in each planning period to meet the production and inventory targets specified in the Master Plan. Once
an order authorising production of a specified quantity of an item is released to the production facility, control over its
progress passes to the Scheduling function (Uzsoy, Lee, and Martin-Vega 1992, 1994; Mönch et al. 2011), which is out-
side the scope of this paper.
There are significant differences between this perspective and the Manufacturing Planning and Control (MPC)
framework widely accepted in the Material Requirements Planning literature (Vollmann et al. 2005). In the latter frame-
work Master Production Scheduling is generally incapacitated, focuses on end items or major subsystems sold into the
market, and is usually conducted at the level of a single factory. In particular, the Bills of Material (BOMs) for the end
International Journal of Production Research 4567

items are not explored to obtain requirements for components and subassemblies; this is the task of the Material
Requirements Planning (MRP) function. The primary purpose of Master Planning, in contrast, is to coordinate the flow
of materials across multiple facilities making up the supply chain to meet demand in the most effective way. This
requires a longer planning horizon than traditional Master Production Scheduling, since it must span at least one sea-
sonal cycle and the cycle time of the entire supply chain being planned. The production of components and subassem-
blies required for the end items is explicitly addressed to ensure material availability. Master Planning is related to the
Supply Chain Operations Planning (SCOP) function of de Kok and Fransoo (2003) and the Goods Flow Problem of
Bertrand, Wortmann and Wijngaard (1990) in its purpose of coordinating material flow across the larger supply chain.
However, it differs from these approaches in that it does not specify release quantities for individual facilities. In this
aspect it resembles the approaches discussed in the literature under the name of Aggregate Production Planning
(Hax and Candea 1984). Thomas and McClain (1993) provide a useful comparison of this idea with the MRP-centred
perspective of Vollmann et al. (2005). Vollmann et al. (2005) use the term Production Planning to denote the aggregate
planning activity whose final result is the end-product level Master Production Schedule; Meyr, Wagner, and Rohde
(2015) use this term for the function that computes capacity-feasible release schedules for individual facilities.
The structure of the mathematical models for Master Planning and Production Planning are quite similar. The princi-
pal decision variables are either releases or production quantities of each product (or product family) in each period in
the planning horizon at each facility considered; under the assumption of fixed, workload-independent lead times the
two are equivalent (Missbauer and Uzsoy 2011). The models must include material balance constraints for all inventory
locations, capacity constraints for critical resources, and domain-specific constraints specifying detailed technological
and business policy constraints.
In the semiconductor supply chain there are good reasons for blurring the boundary between Master Planning and
Production Planning, as discussed by Fordyce et al. (2012). The presence of alternative manufacturing routings, both
within facilities and across different facilities, and BOMs discussed in Parts I and II results in significant loss of infor-
mation when individual machine capacities are aggregated into larger capacity pools, or individual products into broader
product families. As the level of detail in Master Planning models is increased to address these concerns, the structural
similarity of the mathematical models for the two problems can easily result in at least some portions of a Master Plan-
ning model acquiring the level of detail usually associated with the Production Planning function. Hence a lengthy
debate on the ‘correct’ boundaries between Master Planning and Production Planning is unlikely to be productive. We
take the view that the primary function of the combined problems is to coordinate the flow of material through the sup-
ply chain to best meet the firm’s objectives. This function can be decomposed in different ways, but the basic tasks must
take place in some form.

2.2 Master planning models for semiconductor supply chains


Despite the widely recognised importance of Master Planning to the success of any manufacturing firm (Albrecht,
Rohde, and Wagner 2015), there appears to be little literature explicitly addressing Master Planning in semiconductor
supply chains.
Ponsignon and Mönch (2012a) study Master Planning for several wafer fabs operating in parallel with no material
flow between them. Back-end facilities are not taken into account, capacity is modelled only for bottleneck machines,
fixed lead times are assumed and subcontracting is available. Firm orders must be fulfilled, and additional demand fore-
casts can be fulfilled if capacity is available. Since the problem is NP-hard, a genetic algorithm (GA) and a product-
based decomposition scheme are proposed. Ponsignon and Mönch (2012b) obtain improved results using an iterative
approach that alternates between solving the Master Planning problem based on a given fixed lead time and a simulation
model that estimates the realised cycle times implied by the Master Planning decisions. A subsequent paper (Ponsignon
and Mönch 2014) evaluates different heuristics for Master Planning under uncertain demand in a rolling horizon setting.
Simulation experiments highlight the trade-off between maximising profit and maintaining stable performance measures
(nervousness) achieved over the simulation horizon. Findings suggest that Master Planning approaches that are outper-
formed by other methods in a static, deterministic setting may yield more stable plans in a rolling horizon setting. A
number of papers have examined this issue of nervousness in production plans, most commonly in the context of lot-
sizing and Manufacturing Resource Planning (MRP) (Blackburn, Kropp, and Millen 1985; Sridharan and LaForge 1989;
Ho and Carter 1996; de Kok and Inderfurth 1997; Kazan, Nagi, and Rump 2000; Lin and Uzsoy 2016a, 2016b) and
many of the results are likely to be relevant to semiconductor supply chains (Braun and Schwartz 2012).
Lowe and Mason (2016) give a MIP formulation for a Master Planning problem in a network of wafer fabs and
back-end facilities. Safety stocks are taken into account to buffer against demand uncertainty. A robust optimisation ver-
sion of this model that allows manufacturers to make more informed decisions is presented by Lowe (2015).
4568 L. Mönch et al.

Denton, Forrest, and Milne (2006) describe a MIP formulation that minimises costs subject to constraints describing
material flows, available capacity at different workcenters in the supply chain, and discrete lot sizes. The model forms
part of the Central Planning Engine developed and implemented at IBM (Wang et al. 2008; Fordyce et al. 2011) to sup-
port strategic and operational planning. Decomposition heuristics are used to solve the large MIP problem instances.
Fordyce et al. (2011) give an extensive discussion of related approaches implemented in different IBM facilities, repre-
senting the evolution of several integrated decision support tools for Master Planning and Production Planning over sev-
eral decades.
Lyon et al. (2001) and Milne et al. (2012) combine MRP-type logic with optimisation to deal with binning and alter-
nate BOMs and manufacturing processes in an IBM wafer fab. First, a linear program (LP) is solved to choose the opti-
mal supply path for an end item, making a locally optimal selection among alternative manufacturing processes and/or
BOMs for the item. The decisions obtained from the LP are then used by an MRP heuristic that adjusts the material
plans. An important benefit of combining LPs with MRP heuristics is the ability to detect infeasibilities in the MRP
plans, indicating where improvements and human intervention are required.
Milne et al. (2015b) incorporate contractual arrangements in a Master Planning model. Minimum wafer starts con-
tracts require the foundry to release at least a minimum number of wafers as long as sufficient orders are placed by the
fabless company. Under complementary demands contracts, customer requirements are simultaneously applied to both
chips and modules in a linked manner, i.e. the foundry has to produce all the wafers required to fulfil the customer
requests, while some of the post-fab processing is done by the foundry and some by other firms contracted by the fab-
less company. A two-stage LP approach is proposed for both contract types. This paper highlights an interesting applica-
tion of optimisation techniques to ensure the compliance of master plans with corporate contractual obligations, a
process that can be very tedious with manual methods or spreadsheets, and which can expose the firm to considerable
financial liability in case of a violation.
The system developed and implemented by Leachman and his colleagues at Harris Semiconductor is a good exam-
ple of a planning system that integrates elements of the Demand Planning, Master Planning, and Production Planning
functions (Leachman 1993, 2001, Leachman and Raar 1994; Leachman et al. 1996). The system begins with a Demand
Quantification module whose task is to ensure that available supply is used to meet demand that is likely to be realised
before allocating scarce capacity to demands that may not materialise. Typically, firm commitments made to customers
and reliable demand forecasts have priority over safety stock replenishment, while less reliable demand forecasts have
the lowest priority.
After demand prioritisation, Requirements Planning for binning and substitution product structures (cf. Part I) is per-
formed. This determines which alternative production processes will be used to produce how much of each item, and
how available inventory and the output of the alternative production processes will be allocated among the different
demand classes. Leachman and Raar (1994) perform this function using LP models whose decision variables represent
the production of source products, i.e. products serving as input for binning, and the allocation of bins (different product
grades) to finished goods in each time period. Inventory balance constraints are imposed for each finished good type
and each bin for each time period. A separate LP is solved for each demand class where the backorder variables of each
class are constrained by the backorder level of the next higher class. Depending on the demand class, a profit- or cost-
based objective is considered. As in the Material Requirements Planning (MRP) logic (Vollmann et al. 2005; Voß and
Woodruff 2006), capacity constraints are not considered.
The Requirements Planning step determines the net requirement for die output by the front-end and back-end plants.
Capacitated loading then uses LP models to compute capacity-feasible release schedules for all production facilities to
provide this output at minimum cost, performing the Production Planning function of Figure 1 discussed in the follow-
ing section.
Ahmadi and Benson (2002) and Ahmadi, Benson, and Supernaw-Issen (1999) describe a planning system for AMD
based on an LP formulation that focuses on computing release quantities by end item at each stage of the manufacturing
process. They define corporate inventory points at different locations in the supply chain at which strategic inventory
can be held. In addition to the well-known inventory point at die bank between wafer fab and probe, they define another
inventory point within wafer fabrication, at the point in the process where a base wafer is customised to different speci-
fic end items. An additional inventory point is located within the final testing stage to allow substitution between end
items. Both binning and assembly are considered, and non-integer exogenous lead times are used in the LP models.
Implementation of the system at AMD is described.
Hung and Wang (1997) address the problem of alternative BOMs by proposing a LP formulation of the bin alloca-
tion problem arising from the presence of random co-production with downgrading where inventory of a higher-grade
product can be used to meet demand of a lower grade product (cf. Part I). Demand for each finished product can be
satisfied by a distinct set of alternative material types, representing different product grades from the binning process.
International Journal of Production Research 4569

The planning problem involves determining how much available material of different types (bins, product grades) to
allocate to finished products during the current planning period and how much to save against future demand. They
develop an alternative material set generation procedure that extends the procedure of Leachman and Carmon (1992) for
Production Planning with alternative machines discussed below. The proposed technique significantly reduces the size
of the LPs for bin allocation planning and their required computation time over conventional formulations.
Kalir and Grosbard (2017) consider the Master Planning problem for a semiconductor supply chain where demand
and capacity are flexible, i.e. can vary between pre-specified limits, as well as sales prices that can vary with the amount
sold to a particular customer. The endogenous prices result in a quadratic programming formulation, which the authors
solve for a simplified supply chain to illustrate the improvement in profit obtained by a global solution over local solu-
tions maximising utilisation at a specific fab.
Lee et al. (2006) formulate the Master Planning problem to implement different supply chain policies before and
after the customer order decoupling point (CODP). Noting that the segments upstream of the CODP, which they locate
at die bank, will operate in make to stock (MTS) mode and those downstream of the CODP, effectively the back end
operations, in make to order (MTO) mode, they extend the model developed by Lee and Lee (2003) for the LCD-TFT
production. They consider policies that maximise output, meeting output targets and minimising deviations from inven-
tory targets for the MTS segment. For the MTO segment they consider maximising revenue and minimising unmet
demand. They examine the performance of the supply chain under different combinations of objectives, concluding that
managing the MTS segment of the supply chain to meet appropriately set inventory targets yields superior performance.

2.3 Production planning models


Production Planning formulations receive demand information for one or more production facilities in the form of the
output targets developed by the Master Planning function. The time horizon considered is generally shorter than that for
Master Planning, and the capacity constraints are more detailed. In order to plan releases to match the output of the pro-
duction facility to the Master Plan, Production Planning models must consider cycle times, the delay between work
being released and its emerging as output, which are of the order of 10–15 weeks in most semiconductor fabs (Mönch,
Fowler, and Mason 2013). We shall refer to the exogenous estimates of cycle times used in production planning models
as lead times.

2.3.1 Models with fixed exogenous lead times


The prevalent approach in both the research literature and industrial practice is to treat cycle times as an exogenous
parameter, referred to as a lead time that is independent of workload and can be estimated from historical data or simu-
lation models. Most linear programming (LP) and mixed integer programming (MIP) models for Production Planning
(Billington, McClain, and Thomas 1983; Johnson and Montgomery 1974; Pochet and Wolsey 2006; Voß and Woodruff
2006) follow this approach, assuming lead times are an integer multiple of the basic planning period. Hackman and
Leachman (1989) and Hackman (2008) present LP models that incorporate fractional lead times independent of the
resource workload and are of the same computational complexity as models with integer lead times. The earliest model
we are aware of in semiconductor manufacturing is that of Smith (1965), who describes an LP model used for planning
the production of germanium alloy transistors at the semiconductor division of Raytheon in 1958.
The capacity loading modules of the IMPReSS system implemented at Harris Semiconductor by Leachman et al.
(1996) perform a capacity loading for all front-end facilities since the number of wafer types is small, and allocate the
resulting probed die output to the various back-end facilities. These network-wide decisions made by IMPReSS repre-
sent an integrated Master Planning solution across all front-end production facilities. Separate capacitated loading calcu-
lations for each back-end facility are performed based on the die allocation obtained from the solution to the front-end
problem. The LP models use fixed, fractional workload-independent lead times along with several other enhancements
described by Leachman (1993). Leachman (2001) discusses a Production Planning model for wafer fabs that accounts
for practical issues such as work in process (WIP) in the system at the start of the planning horizon, factory working
calendars, auxiliary resources such as reticles or test handlers, fractional lead times and variable-length planning periods,
and accommodates product structures featuring binning and substitution. Allison et al. (1997) use discrete-event simula-
tion to parameterize LP formulations in the IMPReSS system and the TYECIN Enterprise Optimization System (TEOS)
and to validate the corresponding production plans.
In the remainder of this paper, we shall assume the reader is familiar with the basic structure of production planning
models, as discussed by Hackman and Leachman (1989), Voß and Woodruff (2006) and Missbauer and Uzsoy (2011),
and focus on enhancements of the basic models.
4570 L. Mönch et al.

Horiguchi et al. (2001) propose a simple Production Planning algorithm based on forward scheduling for WIP and
backward scheduling for new orders. Simulation experiments demonstrate that explicitly modelling the finite capacity of
near-constraint work centres leads to the best results. Significant improvements in both on-time delivery performance
and predictability (nervousness) are obtained using a shop-floor scheduling procedure driven by the planned completion
date rather than customer due dates. Habenicht and Mönch (2002) use a beam search approach for a similar short-term
production planning problem, to which Habla, Mönch, and Drießel (2007) apply a Lagrangian relaxation approach.
These papers are of interest in its exploration of the interface between Production Planning and Shop-Floor Scheduling,
which is a significantly under-researched area.
Chou and Hong (2000) formulate the production planning problem facing a foundry with a mix of firm orders and
forecasts. They provide an interesting examination of the effects of the length of the planning period on the variability
of workloads on individual tools. They distinguish between order batches and release batches, where an order batch rep-
resents a quantity ordered by a customer and a release batch a portion of this order batch actually released to produc-
tion. They provide a MIP model whose decision variables select order batches for release subject to constraints on
capacity, variation in production quantities between periods and maximum release quantities. Several objectives such as
maximising utilisation, maximising output, or maximising profit are considered. Once the batch selection variables are
fixed, the resulting LP models can be used for sensitivity analysis. They also suggest using the model systematically for
bottleneck identification, taking into account the possibility that bottlenecks will shift as product mix changes over time.
This paper raises a number of very interesting issues of practical relevance that are well worthy of future study.
A common phenomenon in wafer fabs is the presence of alternative machines capable of performing a given process
step on a particular wafer. Since the efficiency of alternative machines for a given process step can differ substantially,
an accurate representation of capacity loading requires a planning model to allocate operations to specific machines.
However, specific operation-machine allocations lie within the domain of short-term scheduling, and hence are better
made in real time by shop-floor management based on changing conditions. Johri (1994) considers the problem of allo-
cating individual operations to specific subsets of machines, providing LP models for the optimal sharing of machines
among overlapping machine groups. Similar problems are addressed by Toktay and Uzsoy (1998), Akçalı, Üngör, and
Uzsoy (2005) and Ignizio (2009). Leachman and Carmon (1992) present a LP formulation that makes specific opera-
tion-machine allocations in each planning period, and use this to develop an aggregate formulation that ensures feasible
capacity loadings without using specific allocation variables assuming that the efficiency of alternative machines differs
by a fixed ratio for all shared process steps. This is accomplished by identifying sets of operations and alternative
machines whose capacity constraints are likely to be binding in the optimal solution (dominant operation-machine sets)
and writing capacity constraints only for these sets. Related modelling approaches for workcentres with non-identical
machines are proposed by Bermon and Hood (1999), Hung and Cheng (2002), and Liberopoulos (2002). The Produc-
tion Planning formulation of Leachman (2001) also addresses dynamic machine arrangement constraints where the set
of allowed machines for a specific process step depends on the machine type assigned at some previous process step.
Milne, Mahapatra, and Wang (2015a) propose a MIP model to determine values of the planned lead times used by
an MRP system that minimise certain costs and penalties. Computational results based on data from a DRAM manufac-
turer show that the proposed method of determining planned lead times outperforms conventional methods used to set
planned lead times in MRP systems. This paper raises an interesting problem that seems to be under-researched in gen-
eral – that of how to parameterise fixed lead time models to obtain the best results. Albey and Uzsoy (2015) address
this issue with a scaled-down model of a wafer fab, using simulation optimisation to obtain the lead time estimates that
result in the best performance. Kriett, Eirich, and Grunow (2017) propose a LP formulation for production planning in a
wafer fab where work is assigned a target cycle time when it is released, subject to the usual material flow balance and
capacity constraints. The formulation is closely related to the transportation formulation of Bowman (1956). An interest-
ing aspect of this model is that it is complemented in execution by an operation due date dispatching policy, where lots
on the shop floor are assigned operation due dates based on the current production plan, which is implemented on a
rolling horizon basis. The authors report promising results from an extensive simulation study based on the MIMAC-I
dataset.
Although exogenous, workload-independent lead times have been the prevalent approach to modelling cycle times
in production planning, this approach has theoretical disadvantages. Queueing theory (Gross and Harris 1974; Buzacott
and Shanthikumar 1993; Curry and Feldman 2009), simulation models (Atherton and Atherton 1995; Ankenman et al.
2011) and industrial experience all suggest that the average cycle time of a production facility increases nonlinearly with
the average resource utilisation, which, in turn, is determined by the release decisions made by the Production Planning
function. Hence, the use of workload-independent exogenous lead times in planning models may lead to poor perfor-
mance at high utilisation levels, where small fluctuations in workload may cause large changes in cycle time, especially
International Journal of Production Research 4571

when utilisation levels at key resources vary substantially over time. How to model these workload-dependent lead times
has been the focus of considerable research (Pahl, Voß, and Woodruff 2005a, 2005b, 2007).
Two principal approaches to this issue have emerged in the literature. The first of these, which we shall refer to as
Iterative Multimodel methods, decomposes the production planning problem into two subproblems. One subproblem,
usually implemented using a LP model, determines the optimal capacitated release schedule for a production facility
under a given set of cycle time estimates, while the other estimates the average cycle times that will be realised under a
given set of releases. The second approach uses non-linear clearing functions that estimate the average output of a tool
group in a planning period as a function of its available workload in that period. We now discuss these approaches, both
of which have been examined in the specific context of semiconductor manufacturing, in the following sections.

2.3.2 Iterative multimodel approaches


Although an iterative multimodel approach was described by Zäpfel (1984), the first iterative multimodel approach to be
widely discussed was that of Hung and Leachman (1996) developed for semiconductor wafer fabrication. These authors
propose an iterative algorithm that estimates the cycle times resulting from a given release plan using a simulation
model of the wafer fab being planned, and passes these estimates to an LP model. The LP model then computes a
release plan based on these revised cycle time estimates, and the simulation model is run again to estimate the new
cycle times. Iteration continues until some convergence criterion is satisfied. Similar approaches are proposed by Kim
and Kim (2001), Byrne and Bakir (1999), Byrne and Hossain (2005), Kim, Kim, and Lee (2014) and Albey and Bilge
(2014) for Production Planning and for Master Planning by Ponsignon and Mönch (2012b). Bang and Kim (2010) use
discrepancies in waiting times, WIP levels, and throughput at the bottleneck machines between the LP solution and the
simulation output to update the parameters of the LP within an iteration, terminating the algorithm if discrepancies
exceed a specified threshold. Hung and Hou (2001) examine the use of queueing and regression models as a substitute
for the simulation model in order to reduce computational burden. Irdem, Kacar, and Uzsoy (2010) study the Hung and
Leachman (1996) and Kim and Kim (2001) procedures, finding that the convergence behaviour of these methods differs
substantially.
Iterative multimodel approaches combine two solution methodologies – linear programming and simulation – that
are well-understood by many practitioners and for which excellent commercial software tools are available. However,
their convergence is not well understood. Some methods, for example that of Kim and Kim (2001), appear to converge
quite consistently, but Albey and Bilge (2014) find that the final solution is dependent on the initial solution. Other
methods of this type, such as that of Riaño (2003), have been found to either converge rapidly or cycle between a lim-
ited number of solutions. The second issue is the high computational burden of the detailed simulation models involved.
Efforts to replace the simulation model with steady state queueing models have not been successful, yielding highly
inaccurate cycle time predictions (Hung and Hou 2001), although replacing the simulation model with an efficient meta-
model as in Li et al. (2016) might offer a solution to this difficulty. Clearly the reduced simulation models discussed in
Part I are also relevant to this issue. As long as the computation of the production plan requires multiple replications of
a complex simulation model, computation time will be a major hurdle for these methods to overcome.

2.3.3 Clearing functions and simulation optimisation models


The Production Planning approaches described above either assume workload independent lead times or require compu-
tationally demanding simulation runs to estimate the cycle times implied by the release plans proposed by the LP mod-
els. In contrast, clearing functions (CFs) relate the expected output of a production resource in a planning period to
some measure of its expected workload, the amount of work available for the resource to process during the planning
period (Missbauer and Uzsoy 2011).
Several CFs proposed in the literature are shown in Figure 2. The ‘constant level’ CF imposes an upper bound on
output, assuming instantaneous production independent of WIP level. Graves (1986) proposes a CF of the form
Yt :¼ a Wt ; (1)

where the expected output Yt in period t is a linear function of the expected WIP Wt available at the start of the period.
This ‘constant proportion’ CF assumes that a fixed lead time of 1=a periods is maintained at all utilisation levels. As
workload increases, this behaviour cannot be sustained causing infeasible levels of output at high WIP levels. The con-
stant proportion CF can thus be complemented with a fixed upper bound on output, yielding the ‘combined’ CF. Non-
linear CFs that model output as a concave non-decreasing function of Wt are proposed by Srinivasan, Carey, and Morton
4572 L. Mönch et al.

Figure 2. Different types of clearing functions.

(1988) and Karmarkar (1989). Figure 2 depicts these CFs as the ‘effective’ CF (Missbauer and Uzsoy 2011). Orçun,
Uzsoy, and Kempf (2006) illustrate the differences between these different CFs using systems dynamics simulations.
CFs can be estimated using steady-state queuing analysis (Missbauer and Uzsoy 2011; Omar, Venkatadri, and Diallo
2016), or a curve-fitting procedure applied to empirical or simulated data (Kacar, Irdem, and Uzsoy 2012; Kacar and
Uzsoy 2014). Instead of trying to optimise the fit of the CFs to the data, Kacar and Uzsoy (2015) use a simulation opti-
misation scheme to estimate values of the CF parameters that optimise expected system performance when the fitted
CFs are used to develop release schedules.
Early CF-based Production Planning models had difficulty representing systems with multiple products due to the
possibility of creating capacity for one product by holding WIP of another. Asmundsson et al. (2009) address this prob-
lem with the Allocated Clearing Function (ACF) formulation where the aggregate output of the production resource is
estimated using an aggregate workload measure and then allocated to individual products. When the CF for each work-
centre is piecewise linearised, the ACF yields a LP formulation. Asmundsson, Rardin, and Uzsoy (2006) apply the ACF
formulation to Production Planning of a scaled-down wafer fab. Kacar, Irdem, and Uzsoy (2012) demonstrate that ACF-
based production planning formulations outperform the iterative LP-simulation schemes of Hung and Leachman (1996)
and Kim and Kim (2001). Kacar, Mönch, and Uzsoy (2013) use a simulation model of a large-scale wafer fab to show
that production plans determined by ACF-based formulations outperform those obtained using fixed integer-valued lead
times. Ziarnetzky et al. (2015) confirm these results in a rolling horizon setting under demand uncertainty. Kacar,
Mönch, and Uzsoy (2016) find that formulations based on appropriately chosen fractional lead times as suggested by
Leachman (2001) perform well under stationary demand, while under time-varying demand the CF-based formulations
perform slightly better. The use of exogenous fractional lead times requires a method for obtaining estimates of the lead
times before release quantities are known; the CF approach leads to much larger formulations. More complex CF mod-
els considering interactions between multiple products, especially setups, and different forms of manufacturing flexi-
biltity have been studied by Kang et al. (2014) and Albey, Bilge, and Uzsoy (2014, 2017).
CFs provide an intuitive approach to addressing the nonlinear dependency between cycle times and workload that
has been a persistent difficulty in the production planning area. Computational experiments on test instances have
yielded promising results, especially with the ACF formulation of Asmundsson, Rardin, and Uzsoy (2006), Asmundsson
et al. (2009). Another useful property of CF models is their ability to provide meaningful dual prices for resources
whose utilisation level is below 1, which classical fixed lead time LP models cannot do (Kefeli et al. 2011; Kefeli and
Uzsoy 2016). However, there are also a number of disadvantages. Multivariate CFs capable of predicting the output of
multi-product systems accurately (Albey, Bilge, and Uzsoy 2014, 2017) tend to result in non-convex optimisation mod-
els that are difficult to solve. Even the ACF formulation requires the definition of decision variables and constraints for
all operations performed on each tool group, resulting in much larger formulations than a conventional formulation of
similar scope (Kacar, Mönch, and Uzsoy 2016). Limited experimentation suggests that when the average utilisation of
resources does not vary drastically over time, the exogenous fixed-lead time models with fractional lead times proposed
by Leachman (1993) yield equally good performance as the substantially more complex CF models. Finally, there is as
International Journal of Production Research 4573

yet no general, rigorous methodology for estimating CFs from data, which constitutes a significant obstacle to their
widespread adoption (Gopalswamy and Uzsoy 2017). All these issues continue to be the subject of ongoing research.
The single-variable CFs on which the discussion above has focused can be viewed as a simple metamodel describ-
ing the behaviour of a production resource modelled as a queueing system. Richer, more complex metamodels can also
be used. Li et al. (2016) implement a metamodel-based simulation optimisation approach for Production Planning in a
scaled-down wafer fab system, using the metamodel to replace the simulation model and then applying simulation opti-
misation. The approach yields excellent results in very modest CPU times compared to the full-blown simulation optimi-
sation approach. The approach of Omar, Venkatadri, and Diallo (2016) can be extended quite directly to a data-driven
approach where detailed data on output and WIP levels is used to describe the behaviour of the production resources
without fitting an analytical function. However, the scalability of this approach to large problems remains to be
explored. Another direction of future research might consist in applying the concept of intrinsic ratios proposed by Wu
and McGinnis (2012, 2013) in production planning formulations for wafer fabs.
Liu et al. (2011) formulate the Production Planning problem for a single wafer fab as a multi-objective simulation
optimisation problem which is solved using the Nondominated Sorting Genetic Algorithm II (NSGA-II) of Deb et al.
(2002). The expected value and variance of the cost are considered as objectives to be minimised. This formulation does
not require explicit lead time information since the cycle time is implicitly represented by the simulation model. The
decision variables are the release quantities for each product in each period.
Ziarnetzky and Mönch (2016b) apply simulation-based optimisation to make backend capacity expansion decisions
that ensure an overall maximum cycle time in a simplified semiconductor supply chain. CFs are used in the production
planning formulation for the wafer fab part.

2.4 Linking production and inventory planning


Research on Inventory Management and Production Planning has evolved into two largely distinct bodies of literature.
Production Planning models take a primarily deterministic perspective, usually based on linear or mixed integer pro-
gramming, while inventory models emphasise the stochastic nature of demand. However, in practice, Production Plan-
ning is based on demand forecasts that are subject to random errors, requiring safety stocks to ensure adequate customer
service levels. The need to produce additional material for safety stocks increases the utilisation of production resources,
and hence the cycle times of production facilities. Changes in the mean and variance of the cycle time of the production
facility replenishing a particular inventory point will, in turn, affect the amount of safety stock needed. An effective
safety stock policy must specify both the amount of safety stock and where in the supply chain it should be held. The
safety stock models discussed in Part II of this review treat the longer term problem of locating and sizing safety stocks
based on the distribution of external demand and the topology of the supply chain network as part of the supply chain
design, without considering the tactical interaction between the capacitated Production Planning problem and the choice
of appropriate safety stocks.
Tian, Willems, and Kempf (2011) propose an iterative approach to this problem that solves a guaranteed service
model for a safety stock location problem (cf. Part II) based on the variability of period demand at each stage. A Pro-
duction Planning model formulated as an LP is solved to determine which demands will be met by which stages. The
period demand variability of each stage is then updated based on the result of the LP and a new safety stock location
problem is solved. Convergence of this iterative scheme for a two-stage supply chain can be proven under fairly mild
conditions. Orçun, Uzsoy, and Kempf (2009) propose a similar iterative approach using a chance-constrained formula-
tion whose constraints define the safety stock required for a specified service level. The algorithm iteratively updates the
safety stock targets based on the cycle times determined by the CF-based Production Planning model. Consistent con-
vergence is observed on a limited set of problem instances.
While the value of advanced demand forecast information is demonstrated by many studies in the inventory manage-
ment literature (Özer 2010), applications in Production Planning are rare. Higle and Kempf (2010) propose a multi-stage
stochastic programming formulation of a simplified semiconductor manufacturing system that is unique in including a
stochastic model of demand evolution over time, similar to the Martingale Model of Forecast Evolution (MMFE) (cf.
Part II). Albey et al. (2015) propose a model to plan releases into a wafer fab in the face of stochastic demand using
the additive MMFE model. The model manages inventory, backorder, and shortfall levels for each planning period using
chance constraints, which allow a subset of constraints to be violated with a specified probability (Charnes and Cooper
1959), to model the target service levels. Cost-minimising base stock levels are computed using a result from Toktay
and Wein (2001). The model is solved in a rolling horizon setting. The computational results, using data from Intel,
show that considering forecast evolution in Production Planning formulations leads to improved performance as long as
there is some excess capacity that can be exploited by the planning formulation. Ziarnetzky, Mönch, and Uzsoy (2017)
4574 L. Mönch et al.

propose similar formulations using the multiplicative MMFE. Similar models are explored in Aouam and Uzsoy (2012,
2015) and Ravindran, Kempf, and Uzsoy (2011), suggesting a promising direction for future work.

2.5 Discussion of master and production planning


Since the Master Plan determines how the firm will use its production resources to serve its markets in the best possible
way, a strong argument can be made that inventory policy decisions specifying the location and amount of safety stocks
to be held as well as the appropriate customer service levels belong in the domain of Master Planning. After all, the
Master Plan decides what we can produce with the resources at our disposal, and if we want to hold any safety stock at
all we have to produce that material as well. Thus, there is a strong need for research exploring the interactions between
safety stocks and capacitated Production Planning, especially for situations where the additional production required to
replenish safety stocks may impact the mean and variance of the factory cycle times.
An effective Master Plan must reflect a consensus between several different functional organisations, usually arrived
at through negotiations between the parties concerned. Sales and marketing may wish to produce small quantities of
some product that they believe will stimulate higher demand in the future; maintenance may wish to take certain key
pieces of equipment offline for preventive maintenance; product development groups may need to manufacture proto-
types to debug new designs; and product line organisations serving different markets often have conflicting requests.
The highly uncertain production environment and markets bring the need for risk management along multiple dimen-
sions. Thus in order to provide better support for Master Planning decisions, it appears necessary to appreciate its nego-
tiation-based multiobjective nature from the outset, and design mathematical and computational tools explicitly to
support this negotiation process. The need for field-based research (McKay 2011) examining how master planning is
done in different environments, such as a commodity manufacturer, a foundry, or a fabless design house, is also evident.
While there is some initial work applying relatively simple Master Planning formulations in a rolling horizon
(Ponsignon and Mönch 2014), more research is required to incorporate complex product structure and advance demand
information.
The problems of Production Planning in the semiconductor industry have attracted considerable attention over the
last two decades, and have led to the implementation of several decision support systems. The feasibility of constructing
large mathematical programming models has repeatedly been demonstrated by several award-winning industrial imple-
mentations (Leachman et al. 1996; Fordyce et al. 2011). As the data collection infrastructure and computational tools
for building and solving such models continue to improve, the development and solution of the models should only get
easier.
A number of rather mundane questions seldom addressed by academic research arise from the models themselves. A
repeated criticism of Production Planning models based on linear programming is their tendency to produce variable
release patterns over time, even when demand patterns are fairly smooth. This behaviour is at least partially due to the
fact that LP models produce extreme point solutions, the implications of which do not appear to have been investigated.
Similarly, many LP formulations of Production Planning problems are highly degenerate, with many alternative optimal
solutions. This behaviour emerges when management requests a change to a parameter after observing an initial solu-
tion; the formulation will often produce a very different solution with the same objective function value as before,
another undesirable behaviour. The combination of these factors suggests that the development of additional decision
support tools that can explain the results of the optimisation models in a manner intelligible to users without expertise
in optimisation would be highly desirable. Greenberg (1996) describes such a system for energy models that would be
extremely helpful for applications in this industry.
Some of these questions must also be addressed when Production Planning formulations are assessed in a rolling
horizon setting under uncertain demand. In such an environment, the value of safety stocks within integrated production
and inventory planning formulations must be investigated in more detail. More research is needed that investigates the
value of advance demand information in Production Planning formulations, as well as the impact of nervousness and
strategies to eliminate it. The work of Albey et al. (2015) and Ziarnetzky, Mönch, and Uzsoy (2017) integrating the
MMFE into a production planning model in a rolling horizon environment represents first steps in this direction.
Another future research direction is related to the fact that production and engineering lots share the same expensive
equipment in wafer fabs. While production lots will be shipped to customers, engineering lots are used to support pro-
duction and process development efforts. While there is some initial work considering engineering activities in Produc-
tion Planning formulations by modelling learning effects due to engineering activities (Kim and Uzsoy 2008, 2013;
Ziarnetzky and Mönch 2016a), more research is needed in this area. At the supply chain level, this type of model is also
related to the introduction of new products into existing factories, often with adverse impact on other products with
International Journal of Production Research 4575

which they must share capacity (Manda et al. 2016). The decision as to which facilities in the supply chain will produce
the new products is also of considerable interest.
Most semiconductor firms today have some form of Production Planning in place, generally using a commercial
platform such as those embedded in Enterprise Resource Planning (ERP) systems. Unless upgraded with a specialised
optimisation capability through Advanced Planning and Scheduling (APS) systems, the Production Planning capability
of most ERP systems is fairly basic. An interesting question is how Production Planning models based on optimisation
can show sufficient benefit to the semiconductor firms to justify the considerable effort of implementing them. The vol-
ume and nature of the data required, as well as the requirement that a Production Planning system interface effectively
with the Manufacturing Execution Systems used to implement and communicate its decisions, create a major tension
between the corporate IT staff who must provide high-quality data and keep daily systems running, and the planning
groups (Ovacik 2010).
Another important future research direction is related to designing next-generation planning systems for semiconduc-
tor supply chains. We believe that such systems must include agents and be service-oriented to permit better alignment
with business processes. Only prototypes based on the service-oriented paradigm (Cherbakov et al. 2007 for the IBM
semiconductor supply chain) and agent-based systems (Herding and Mönch 2016) are available at present. Avoiding
some of the limitations of the packaged APS systems that are currently in place in wafer fabs such as lack of trans-
parency and inability to model important business rules seems possible. A prerequisite for such systems is a standard
terminology to support the rich communication that is required for effective agent-based software systems.

3. Demand fulfilment
3.1 Problem setting and basic literature
Demand fulfilment involves tracking orders from order entry to delivery, encompassing order promising, due date set-
ting, and shortage planning (Stadtler 2005). This requires three categories of activities (Framinan and Leisten 2010):
order capture, production (order) planning, and order execution. This section will focus on order capture and some ele-
ments of order execution, but will only briefly discuss order planning which was covered in Section 2.3.
Order capture is generally supported by ATP software (Framinan and Leisten 2010), whose main decisions are order
acceptance/selection, due date assignment, and order scheduling. Once an order is accepted and a due date assigned, it
is the task of the Scheduling function (Mönch et al. 2011; Mönch, Fowler, and Mason 2013) to meet the specified due
date, typically by setting internal due dates for production lots as they make their way through the fab, sort, assembly,
and test stages. The interface between Planning and Scheduling functions is involved again here and represents an
under-researched area of considerable interest.
A key element in the order acceptance/selection process is estimating the available capacity to promise for a
requested order, in order to give reliable delivery date promises to customers. APICS (2016) defines ATP as: ‘the
uncommitted portion of a company’s inventory and planned production maintained in the master schedule to support
customer order processing’. They then define capable to promise (CTP) as ‘the process of measuring orders against
available capacity and inventory. CTP is used to determine when a new or unscheduled customer order can be delivered.
The process may involve multiple manufacturing or distribution sites’.
Pibernik (2005) indicates that: ‘“Conventional’ ATP, commonly implemented in ERP, merely determines the avail-
ability of finished goods at certain points of time in the future. ‘Advanced’ ATP provides additional functions such as
order quantity and due date quotation on the basis of available supply chain resources and alternative measures in case
of an anticipated shortage of finished goods or manufacturing resources.” He goes on to classify Advanced ATP (AATP)
according to availability level, operating mode, and interaction with MRP. Availability level indicates whether quantity
and due quotations are based on finished goods inventory (for MTS products) or supply chain resources (for MTO
products). Operating mode indicates whether AATP is done in real-time or batch mode. Interaction between AATP and
Manufacturing Resource Planning (MRP) can be passive or active. In passive AATP, information is shared from the
MRP system, but the system is not updated. In active AATP, there is integration and a two-way flow of information and
actions.
Ball, Chen, and Zhao (2004) offer a more comprehensive discussion of the ATP function. They note that most sup-
ply chains contain segments that are operated in MTS mode, and others operated in MTO mode. Activities in the MTS
or push segment are driven by forecasts, and those in the MTO or Pull segment by specific orders. They define the ATP
functionality as coordinating activities across these two areas. In the MTS segment of the supply chain, ATP involves
identifying production capacity that has been allocated to products or product families and has not yet been committed
to meeting a particular instance of customer demand. In the MTO segment, on the other hand, ATP involves identifying
4576 L. Mönch et al.

available inventory of particular products and their components that can be used to meet a realised or proposed customer
order within the customer lead time. Thus, the ATP activity interacts with Master Planning, Demand Management, Pro-
duction Planning, and Scheduling functions. The authors offer several different classifications of ATP methods, relating
them to the literature on inventory models, due date assignment and order acceptance, and shop-floor scheduling.
Chen, Zhao, and Ball (2002) note that customer quotations can be made in real time or in batch mode where multi-
ple customer requests are considered at the same time. Collecting information on multiple orders before making the quo-
tation yields a better overall solution. They develop a MIP model that maximises overall operations-based profit in a
multi-period batch ATP execution subject to inventory and capacity constraints for MTO and assemble-to-order (ATO)
systems. They also point out that it also possible to provide a rough quotation in real time and a refined quotation after
a batch execution. Zhao, Ball, and Kotake (2005) propose an integer programming model to be run in batch mode to
perform ATP functions for Toshiba, where the model decides on whether to accept orders, whether to split them if
accepted, and assigns due dates based on material and capacity availability.

3.2 Demand fulfilment for semiconductor manufacturing


Leachman and Raar (1994) and Leachman et al. (1996) mention a module of the Berkeley Planning System and its Har-
ris implementation (IMPReSS) that calculates planned product availability for the quotation and order entry system, but
give no further details. Soares, Azevedo, and de Sousa (2000) describe the requirements analysis and system specifica-
tion for an order promise (demand fulfilment and ATP) module for a semiconductor supply chain. Core elements of an
ontology for planning tasks in a semiconductor supply chain are derived. They also discuss the integration of global
planning and order promising. In their description of the PROFIT system implemented at IBM, Lyon et al. (2001) men-
tion the existence of a division ATP application without providing details. However, the basic function must take place
in any operating firm, so it is likely that many semiconductor firms use commercial APS systems for this task.
Chiang and Hsu (2014) describe a three-stage integrated order fulfilment model with periodic review for MTO semi-
conductor manufacturing (specifically foundries) to maximise profits. They propose a periodic review mechanism to
reallocate unused ATP quotas, demonstrating enhanced capacity utilisation and profitability on a simplified but realistic
data-set. The model can be used in real time or batch mode, and includes allocation planning, order promising, and peri-
odic allocation review. During allocation planning, an optimisation model is run to identify capacity as committed or
uncommitted based on forecasted and firm demand along with pre-specified yield and cycle time parameters. The multi-
period capacity allocation model maximises profit by assigning orders to available capacity on bottleneck machines. The
remaining uncommitted ATP capacity becomes an input to the next stages; the committed capacity is hard pegged to
specified customers and cannot be consumed by other customers. In the order promising stage, a multi-period ATP con-
sumption model maximises the total profit associated with the amount of committed ATP capacity assigned to specific
order items plus the profit associated with the amount of unconsumed ATP capacity assigned to specific order items.
The profit margin depends on the customer, factory, technology group, and time period.
In the third stage, the status of the ATP capacity consumption is reviewed periodically to account for demand uncer-
tainty, since it is not feasible to check it in real time. If customers confirm that their allocation will not be consumed,
committed ATP quantities are released to a central pool and the multi-period consumption model rerun to allocate the
newly released capacity to unsatisfied order items. The objective function at this stage is to maximise the profit obtained
from the committed ATP capacity minus the cost of completing orders early. A case study of a MTO semiconductor
firm shows significant improvement from the allocation review mechanism and, interestingly, only slight improvement
from using the model in batch mode over real-time processing.
Framinan and Perez-Gonzalez (2016) conduct experiments with the allocation planning formulation from Chiang
and Hsu (2014) to understand the trade-off between forecast accuracy and bias and allocation planning rules. The results
reveal the sensitivity of ATP allocation to the accuracy and bias of the forecasts.
Tsai and Wang (2009) present a similar three-stage model for TFT-LCD assemble-to-order manufacturing, which
consists of three processes: the Array process (simpler than, but similar to, wafer fabrication), the Cell process, and the
Module process. They assume that the CODP (cf. Part II) (push-pull boundary) lies between the Cell and Module pro-
cesses. The first stage of their model assigns orders to module factories, the second stage allocates ATP within module
factories, and the third stage reallocates unsatisfied orders to module factories. Several other papers on ATP for the
TFT-LCD industry have appeared over the last several years. Jeong et al. (2002) develop an ATP model that estimates
the promise delivery date for new orders and a CTP model that determines the unused capacity at the shop-floor level
for given production schedules. They also propose a heuristic to schedule final assembly to effectively use the uncom-
mitted capacity. Lin and Chen (2005) propose an order promising mechanism that uses a MIP model to allocate manu-
facturing resources to high profit products or important customers subject to material and capacity constraints. Chen,
International Journal of Production Research 4577

Lin, and Wu (2008) discusses the need for a rolling horizon approach to order promising that reallocates materials and
capacity considering updated information. Lin et al. (2010) develop a batch mode MIP model for allocation planning
that includes alternative BOMs and grade transitions (similar to binning) and study the sensitivity of plant performance
to changes in order batching intervals. Yang et al. (2010) proposes a new method for order promising in wafer fabs
called Capacity Reserved for Multi-Priority Orders (CRMPO) that divides orders into several priority classes and
reserves capacity for each class. Han, Dong, and Liu (2014) develop a near real-time allocation policy that considers
demand upgrade substitution.
Seitz and Grunow (2017) propose a procedure to perform the ATP function in a semiconductor supply chain with
alternative manufacturing flows and BOMs. The central idea is the use of a cumulative ATP calculation, which estimates
the total amount of material available to meet a particular product based on all alternative BOMs and process flows
available to provide that product. Simple knapsack-like LPs, similar to those suggested by Leachman and Raar (1994),
are used to allocate the cumulative ATP quantities among customer orders.

3.3 Lot to order matching


There are two small bodies of literature that deal with assigning wafer fabrication lots to specific customer orders, one
concerned with MTO products and the other with MTS products. The main difference between the two is where the lots
are when the assignment to an order is made. For MTO products, the lots are ‘pegged’ to orders while in the wafer fab-
rication facility. In the case of MTS products, the lots are generally in a die bank inventory located at the assembly
facility.
Kim et al. (2008) consider a hard pegging strategy and three soft pegging strategies for MTO products. Hard peg-
ging assumes that a production lot is assigned to a single customer order and cannot be reassigned. Soft pegging permits
reassignment (repegging) in the face of urgent orders and machine failures. Repegging is an option when many products
share many of their initial process steps, with material being committed to a specific product at a branching point in the
process flow. Lots can be repegged to different orders if they have not yet reached the branching point or are of the
same product type. Unsurprisingly, they find that soft pegging significantly outperforms hard pegging. Kim et al. (2010)
consider the problem of pegging and order release into the wafer fab by formulating a mixed integer program and a
simulated annealing algorithm to minimise the total tardiness of customer orders. Kim and Lim (2012) develop more
efficient solution methods for this problem.
For MTS products, Knutson et al. (1999) tackle the daily problem of assigning different sized wafer lots of a single
product in a die bank serving a single assembly facility to customer orders of different sizes. The objectives are to max-
imise the number of die sent to customers (DTO) and the number of orders delivered on time (OTD), and minimise the
number of die sent to the warehouse as excess inventory (DTW). A wafer fab lot is only used in a single order (for
traceability reasons) and any extra dice will be considered excess inventory. Fowler, Knutson, and Carlyle (2000) and
Carlyle, Knutson, and Fowler (2001) propose additional heuristics for this problem. Boushell et al. (2008) consider the
lot to order matching problem when binning yields multiple product classes on each wafer, which is related to the
inventory management problem with random co-production. Ng, Sun, and Fowler (2010) consider a similar problem that
looks at under- or over-filling customer orders with uncertain lot sizes using robust optimisation. Sun, Fowler, and
Shunk (2011) study a generalised version of the problem that allows downward product substitution when demand
exceeds supply.

3.4 Discussion of demand fulfilment and future research directions


There has been relatively little work in the IE and OR communities on Demand Fulfilment for semiconductor manufac-
turing. There is still a need to improve the use of unused capacity and to develop algorithms for swapping ATP capacity
for different order items in both MTO and MTS environments. An important aspect of these problems is that they are
conducted on a rolling horizon basis, which creates the possibility of undesirable nervous behaviours, such as a given
production lot or unit of inventory being repeatedly reallocated to different customer orders and only shipped after sev-
eral reallocations which, in retrospect, were not all necessary. Finally, the interaction between Production Planning, Mas-
ter Planning, and Order Fulfilment needs to be better understood. Much existing work is based on decompositions of
these tasks that are assumed to be given. Developments in information technology may well enable alternative
approaches in which the different levels can take into account the needs of others to yield a more effective overall oper-
ation. It is interesting to examine whether relatively minor modifications to the Master Planning logic might make
Demand Fulfilment substantially easier.
4578 L. Mönch et al.

4. Summary and conclusions


This series of three papers has sought to provide a unified review of the IE and OR literature on supply chain manage-
ment in the semiconductor industry. The papers have covered a wide range of research areas and methods, which we
hope will provide a useful resource for both researchers and practitioners in this domain. Having adopted the functional
perspective of Meyr, Wagner, and Rohde (2015), we have examined the research in each functional area, discussing the
state of the art and suggesting directions for future research. We now conclude with some broader conclusions from this
review, and their implications for future work in this area.
It is immediately apparent that the SCM problems in the semiconductor industry have attracted a great deal of
research attention from the academic community, as well as considerable interest from various industry research groups.
The degree to which this research has led to implemented systems, and the extent to which these systems have managed
to survive and evolve over time, as well as spread beyond the firms that originated them would be a very interesting
topic for an extensive field study, and would yield some interesting insights into the nature of industry-academic collab-
orations and the effective transfer of technology in this area. In the 1990s a number of companies, notably IBM, Harris
Semiconductor, AMD, Intel, and Samsung among others, were willing to invest substantially in the implementation of
different types of planning systems. However, there is rather less evidence of this type of work in the open academic lit-
erature in recent years, which may be due to a number of factors. One such might be the broader trend in supply chain
planning practice where commercial APS and ERP systems have achieved dominance in the market, creating substantial
barriers to entry for novel solutions whose benefits are often much harder to assess than their risks (Ovacik 2010). In
some cases it may be simply that much ongoing work, which by its nature must be closely linked to the rapidly chang-
ing IT environment, is taking place in companies who are unwilling to compromise intellectual property by publishing
in the open literature.
From an academic perspective, the complexity of the problems encountered in semiconductor supply chains renders
them a challenging testbed for many solution methods developed in this area, attracting many academic researchers who
have developed models of a relatively generic nature to this sector. Almost all the models reviewed in this survey are
applicable, at least in broad terms, to a wide range of other industries, giving some justification to the claim that work
in this industry may benefit other industrial sectors.
Comparison of the work reviewed in this paper with that reviewed in Uzsoy, Lee, and Martin-Vega (1992, 1994) is
interesting in several ways. A rather larger proportion of the early work originated in industry, or in fairly explicit indus-
try-academia collaborations. A far larger proportion of the work reviewed in the current series originates outside the
U.S. suggesting shifts in interest by both industry and governments in the different countries as well as significant
growth in the industry outside the U.S. While it is hard to draw firm conclusions from the research papers alone, one is
left with the impression that industry-academia collaborations are better established on an institutional basis in countries
such as Taiwan, Korea, and Germany than is currently the case in the United States. In support of this point, a review
of the papers in this series seems to yield approximately the same number of U.S. academic research groups working in
the area of semiconductor supply chains as was the case in the early 1990s, if not actually fewer. Limited funding from
both government and industrial sources makes it very difficult for academics at the start of their careers to invest in
developing industrial collaborations and the knowledge of this industry required to support advances in this area. A rig-
orous examination of these hypotheses would be of considerable benefit to several constituencies, but would require
substantial resources.
A recurring theme of this review has been that many of the SCM problems faced in this industry are also faced to a
greater or lesser degree by others, creating opportunities for leveraging research originally conducted in other industries.
Notable in this aspect are the development of demand forecasting models for products with short life cycles, production
and inventory management with random co-production and alternative BOMs; parsimonious simulation models for large
complex systems; and the strategic design and coordination of supply chains under different types of uncertainties.
This review also highlights significant gaps in our knowledge that also apply to a broader range of industries beyond
the semiconductor sector. The generally close match between the functional areas and the research literature has led to
very limited research at the boundaries of functional areas, or spanning several functional areas. The systems developed
and implemented at IBM (Fordyce et al. 2011) and Harris (Leachman et al. 1996; Leachman 2001) are among the few
to span multiple functional areas, and even in these cases the interfaces between functional areas are not well described.
Several papers hint at the benefits of models linking Shop-floor Scheduling and Production Planning (Horiguchi et al.
2001; Asmundsson, Rardin, and Uzsoy 2006), Demand Forecasting and Production Planning (Albey et al. 2015), and
Production Planning and Inventory Management (Orçun, Uzsoy, and Kempf 2009; Tian, Willems, and Kempf 2011;
Albey et al. 2015), but none of these research efforts has yet led to a clear understanding of how to structure the interac-
tions between these areas under different operating conditions. As such, research in the semiconductor supply chain
International Journal of Production Research 4579

may well continue to generate new developments that will advance our knowledge of the broader fields of production
and supply chain management.

Acknowledgements
The authors would like to thank the attendees of the Dagstuhl Seminar 16062 ‘Modeling and Analysis of Semiconductor Supply
Chains’ held in 7–12 February 2016. This paper profits by the insights that are a result of the active contribution of every attendee.
The seminar would not have been possible without the generous support of the Leibniz Center for Informatics, which the authors also
gratefully acknowledge. We would like to thank Kenneth Fordyce and John Milne who read and commented on an earlier draft of
this paper.

Disclosure statement
No potential conflict of interest was reported by the authors.

References

Ahmadi, J., and R. Benson. 2002. “Supply Chain Production Planning Using Linear Optimization.” Scientia Iranica 9 (3): 283–288.
Ahmadi, J., R. Benson, and D. Supernaw-Issen. 1999. “Planning of Production in Semiconductor Manufacturing.” Operational
Research in Industry, Chapter 10, edited by T. A. Ciriani, S. Gliozzi, and E. Johnson. Basingstoke: Macmillan.
Akçalı, E., A. Üngör, and R. Uzsoy. 2005. “Short-term Capacity Allocation with Tool and Setup Constraints.” Naval Research Logis-
tics 52: 754–764.
Albey, E., and U. Bilge. 2014. An Improved Iterative Linear Programming-simulation Approach for Production Planning. Istanbul:
Department of Industrial Engineering, Ozyegin University.
Albey, E., and R. Uzsoy. 2015. “Lead Time Modeling in Production Planning.” Proceedings of the 2015 Winter Simulation Confer-
ence, 1996–2007, Huntington Beach, CA.
Albey, E., U. Bilge, and R. Uzsoy. 2014. “An Exploratory Study of Disaggregated Clearing Functions for Multiple Product Single
Machine Production Environments.” International Journal of Production Research 52 (18): 5301–5322.
Albey, E., U. Bilge, and R. Uzsoy. 2017. “Aggregate Capacity Modeling in Multi-stage Production Systems.” International Journal
of Production Research 55 (14): 4164–4179.
Albey, E., A. Norouzi, K. Kempf, and R. Uzsoy. 2015. “Demand Modeling With Forecast Evolution: An Application to Production
Planning.” IEEE Transactions on Semiconductor Manufacturing 28 (3): 374–384.
Albrecht, M., J. Rohde, and M. Wagner. 2015. “Master Planning.” In Supply Chain Management and Advanced Planning – Concepts,
Models and Software. 5th ed. Chapter 8, H. Stadtler, C. Kilger, and H. Meyr, 155–176. Berlin: Springer.
Allison, R., J. Yu, L. Tsai, C. Liu, M. Drummond, D. Kayton, T. Sustae, and J. Witte. 1997. “Macro Model Development as a Bridge
between Factory Level Simulation and LP Enterprise Systems.” In Proceedings of the IEEE/CPMT International Electronics
Manufacturing Technology Symposium, Austin, TX, 408–416.
Ankenman, B., J. Bekki, J. Fowler, G. Mackulak, B. Nelson, and F. Yang. 2011. “Simulation in Production Planning: An Overview
with Emphasis on Recent Developments in Cycle Time Estimation.” Planning Production and Inventories in the Extended
Enterprise: A State of the Art Handbook, vol. 1, edited by K. G. Kempf, P. Keskinocak, and R. Uzsoy, Chapter 17, 565–591.
Berlin: Springer.
Aouam, T., and R. Uzsoy. 2012. “An Exploratory Analysis of Production Planning in the Face of Stochastic Demand and Workload-
dependent Lead Times.” In Decision Policies for Production Networks, edited by K. Kempf, and D Armbruster, Chapter 8,
173–208. Boston, MA: Springer.
Aouam, T., and R. Uzsoy. 2015. “Zero-order Production Planning Models with Stochastic Demand and Workload-dependent Lead
Times.” International Journal of Production Research 53 (6): 1661–1679.
APICS. 2016. APICS Operations Management Body of Knowledge (OMBOK) Framework. 3rd ed. Last accessed November 27.
http://www.apics.org/sites/apics-supply-chain-council/research-and-pulicaions/ombok
Asmundsson, J., R. Rardin, and R. Uzsoy. 2006. “Tractable Nonlinear Production Planning Models for Semiconductor Wafer Fabrica-
tion Facilities.” IEEE Transactions on Semiconductor Manufacturing 19 (1): 95–111.
Asmundsson, J., R. Rardin, C. Turkseven, and R. Uzsoy. 2009. “Production Planning Models with Resources Subject to Congestion.”
Naval Research Logistics 56: 142–157.
Atherton, L. F., and R. W. Atherton. 1995. Wafer Fabrication: Factory Performance and Analysis. Norwell, MA: Kluwer.
Ball, M., C. Y. Chen, and Z. Y. Zhao. 2004. “Available to Promise.” In Handbook of Quantitative Supply Chain Analysis: Modeling
in the E-business Era, edited by D. Simchi-Levi, S. D. Wu, and W. M. Shen, Chapter 11, 447–483. Boston, MA: Kluwer Aca-
demic Publishers.
Bang, J., and Y. Kim. 2010. “Hierarchical Production Planning for Semiconductor Wafer Fabrication Based on Linear Programming
and Discrete-event Simulation.” IEEE Transactions on Automation Science and Engineering 7 (2): 326–336.
4580 L. Mönch et al.

Bermon, S., and S. Hood. 1999. “Capacity Optimization Planning System (CAPS).” Interfaces 29 (5): 31–50.
Bertrand, J. W. M., J. C. Wortmann, and J. Wijngaard. 1990. Production Control: A Structural and Design Oriented Approach.
Amsterdam: Elsevier.
Billington, P., J. McClain, and L. Thomas. 1983. “Mathematical Programming Approaches to Capacity-constrained MRP Systems:
Review, Formulation and Problem Reduction.” Management Science 29 (10): 1126–1141.
Blackburn, J., D. Kropp, and R. Millen. 1985. “MRP System Nervousness: Causes and Cures.” Engineering Costs and Production
Economics 9 (1–3): 141–146.
Boushell, T. G., J. W. Fowler, A. Keha, K. Knutson, and D. C. Montgomery. 2008. “Evaluation of Heuristics for a Class-constrained
Lot-to-Order Matching Problem in Semiconductor Manufacturing.” International Journal of Production Research 46 (12):
4143–3166.
Bowman, E. B. 1956. “Production Scheduling by the Transportation Method of Linear Programming.” Operations Research 4 (1):
100–103.
Braun, M., and J. Schwartz. 2012. “A Control Theoretic Evaluation of Schedule Nervousness Suppression Techniques for Master
Production Scheduling.” In Decision Policies for Production Networks, edited by K. Kempf, and D. Armbruster, Chapter 7,
143–171. New York: Springer.
Buzacott, J. A., and J. G. Shanthikumar. 1993. Stochastic Models of Manufacturing Systems. Englewood Cliffs, NJ: Prentice Hall.
Byrne, M., and M. Bakir. 1999. “Production Planning Using a Hybrid Simulation – Analytical Approach.” International Journal of
Production Economics 59: 305–311.
Byrne, M., and M. Hossain. 2005. “Production Planning: An Improved Hybrid Approach.” International Journal of Production Eco-
nomics 93–94: 225–229.
Carlyle, W., K. Knutson, and J. Fowler. 2001. “Bin Covering Algorithms in the Second Stage of the Lot to Order Matching Prob-
lem.” Journal of the Operational Research Society 52 (11): 1232–1243.
Charnes, A., and W. Cooper. 1959. “Chance-constrained Programming.” Management Science 6 (1): 73–79.
Chen, J. H., J. T. Lin, and Y. S. Wu. 2008. “Order Promising Rolling Planning with ATP/CTP Reallocation Mechanism.” Industrial
Engineering and Management Systems 7 (1): 57–65.
Chen, C., Z. Zhao, and M. Ball. 2002. “A Model for Batch Advanced Available to Promise.” Production and Operations Manage-
ment 11 (4): 424–440.
Cherbakov, L., G. Meissner, C. Osipov, and L. Walker. 2007. “Service Oriented Architecture – SOA: IBM Microelectronics ‘Factory
in the Box’.” IBM. Last accessed on August 13, 2016. http://www-01.ibm.com/software/solutions/soa/newsletter/june07/arti
cle_microelectronics_case_study.html
Chiang, C., and H. Hsu. 2014. “An Order Fulfillment Model With Periodic Review Mechanism in Semiconductor Foundry Plants.”
IEEE Transactions on Semiconductor Manufacturing 27 (4): 489–500.
Chou, Y.-C., and I.-H. Hong. 2000. “A Methodology for Product Mix Planning in Semiconductor Foundry Manufacturing.” IEEE
Transactions on Semiconductor Manufacturing 13 (3): 278–285.
Curry, G., and R. M. Feldman. 2009. Manufacturing Systems Modeling and Analysis. Berlin: Springer.
Deb, K., A. Pratap, S. Agarwal, and T. Meyarivan. 2002. “A Fast and Elitist Genetic Algorithm: NSGA-II.” IEEE Transactions on
Evolutionary Computation 6 (2): 182–197.
Denton, B., J. Forrest, and R. Milne. 2006. “IBM Solves a Mixed-integer Program to Optimize its Semiconductor Supply Chain.”
Interfaces 36 (5): 386–399.
Fordyce, S., C. Wang, C. Chang, A. Degbotse, B. Denton, P. Lyon, R. Milne, R. Orzell, R. Rice, and J. Waite. 2011. “The Ongoing
Challenge: Creating and Enterprise-wide Detailed Supply Chain Plan for Semiconductor and Package Operations.” In Planning
Production and Inventories in the Extended Enterprise: A State of the Art Handbook, vol. 2, K. Kempf, P. Keskinocak, and R.
Uzsoy, Chapter 14, 313–387. Springer, New York.
Fordyce, K., J. Fournier, R. J. Milne, and H. Singh. 2012. “Tutorial: Illusion of Capacity – Challenge of Incorporating the Complexity
of Fab Capacity (Tool Deployment and Operating Curve) into Central Planning for Firms with Substantial Non-fab Complex-
ity.” In Proceedings of the 2012 Winter Simulation Conference, Berlin, 2302–2317.
Fowler, J., K. Knutson, and W. Carlyle. 2000. “Comparison and Evaluation of Lot-to-Order Matching Policies for a Semiconductor
Assembly and Test Facility.” International Journal of Production Research 38 (8): 1841–1853.
Framinan, J., and R. Leisten. 2010. “Available-to-Promise (ATP) Systems: A Classification and Framework for Analysis.” Interna-
tional Journal of Production Research 48 (11): 3079–3103.
Framinan, J. M., and P. Perez-Gonzalez. 2016. “Available-to-Promise Systems in the Semiconductor Industry: A Review of Contribu-
tions and a Preliminary Experiment.” In Proceedings of the 2016 Winter Simulation Conference, Arlington, VA, 2652–2663.
Gopalswamy, K., and R. Uzsoy. 2017. “An Iterative Refinement Approach to Fitting Clearing Functions to Data from Simula-
tion Models of Production Systems.” Prcoeedings of the 2017 Winter Simulation Conference, Las Vegas, NV. Accepted for
publication.
Graves, S. C. 1986. “A Tactical Planning Model for a Job Shop.” Operations Research 34: 522–533.
Greenberg, H. J. 1996. “The Analyze Rulebase for Supporting LP Analysis.” Annals of Operations Research 65: 91–126.
Gross, D., and C. M. Harris. 1974. Fundamentals of Queuing Theory. New York: Wiley.
International Journal of Production Research 4581

Habenicht, I., and L. Mönch. 2002. “A Finite-capacity Beam-search-algorithm for Production Scheduling in Semiconductor Manufac-
turing.” In Proceedings of the 2002 Winter Simulation Conference, San Diego, CA, 1406–1413.
Habla, C., L. Mönch, and R. Drießel. 2007. “A Finite Capacity Production Planning Approach for Semiconductor Manufacturing.” In
Proceedings IEEE Conference on Automation Science and Engineering (CASE), Scottsdale, AZ, 82–87.
Hackman, S. 2008. Production Economics. Berlin: Springer.
Hackman, S., and R. Leachman. 1989. “A General Framework for Modeling Production.” Management Science 35 (4): 478–495.
Han, G., M. Dong, and S. Liu. 2014. “Yield and Allocation Management in a Continuous Make-to-stock System with Demand
Upgrade Substitution.” International Journal of Production Economics 156: 124–131.
Hax, A. C., and D. Candea. 1984. Production and Inventory Management. Englewood Cliffs, NJ: Prentice-Hall.
Herding, R., and L. Mönch. 2016. “S²CMAS: An Agent-based System for Planning and Control in Semiconductor Supply Chains.”
In Proceedings MATES 2016 LNCS 9872, Klagenfurt, 115–130.
Higle, J., and K. Kempf. 2010. “Production Planning Under Supply and Demand Uncertainty: A Stochastic Programming Approach.”
In Stochastic Programming: The State of the Art, edited by G. Infanger, Chapter 14, 297–315, Berlin: Springer.
Ho, C., and P. Carter. 1996. “An Investigation of Alternative Dampening Procedures to Cope with MRP System Nervousness.” Inter-
national Journal of Production Research 34 (1): 137–156.
Horiguchi, K., N. Raghavan, R. Uzsoy, and S. Venkateswaran. 2001. “Finite Capacity Production Planning Algorithms for a Wafer
Fabrication Facility.” International Journal of Production Research 39: 825–842.
Hung, Y., and G. Cheng. 2002. “Hybrid Capacity Modeling for Alternative Machine Types in Linear Programming Production Plan-
ning.” IIE Transactions 34: 157–165.
Hung, Y. F., and M. C. Hou. 2001. “A Production Planning Approach Based on Iterations of Linear Programming Optimization and
Flow Time Prediction.” Journal of the Chinese Institute of Industrial Engineers 18 (3): 55–67.
Hung, Y., and R. Leachman. 1996. “A Production Planning Methodology for Semiconductor Manufacturing Based on Iterative Simu-
lation and Linear Programming Calculations.” IEEE Transactions on Semiconductor Manufacturing 9: 257–269.
Hung, Y., and Q. Wang. 1997. “A New Formulation Technique for Alternative Material Planning – An Approach for Semiconductor
Bin Allocation.” Computers & Industrial Engineering 32 (2): 281–297.
Ignizio, J. P. 2009. “Cycle Time Reduction via Machine-to-operation Qualification.” International Journal of Production Research
47 (24): 6899–6906.
Irdem, D., N. Kacar, and R. Uzsoy. 2010. “An Exploratory Analysis of Two Iterative Linear Programming-simulation Approaches for
Production Planning.” IEEE Transactions on Semiconductor Manufacturing 23: 442–455.
Jeong, B., S.-B. Sim, H.-S. Jeong, and S.-W. Kim. 2002. “An Available-to-promise System for TFT LCD Manufacturing in Supply
Chain.” Computers & Industrial Engineering 43: 191–212.
Johnson, L., and D. Montgomery. 1974. Operations Research in Production Planning, Scheduling and Inventory Control. New York:
Wiley.
Johri, P. 1994. “Overlapping Machine Groups in Semiconductor Wafer Fabrication.” European Journal of Operational Research 74:
509–518.
Kacar, N., and R. Uzsoy. 2014. “A Comparison of Multiple Linear Regression Approaches for Fitting Clearing Functions to Empirical
Data.” International Journal of Production Research 52 (11): 3164–3184.
Kacar, N., and R. Uzsoy. 2015. “Estimating Clearing Functions for Production Resources Using Simulation Optimization.” IEEE
Transactions on Automation Science and Engineering 12 (2): 539–552.
Kacar, N., D. Irdem, and R. Uzsoy. 2012. “An Experimental Comparison of Production Planning Using Clearing Functions and Itera-
tive Linear Programming-Simulation Algorithms.” IEEE Transactions on Semiconductor Manufacturing 25: 104–117.
Kacar, N., L. Mönch, and R. Uzsoy. 2013. “Planning Wafer Starts Using Nonlinear Clearing Functions: A Large-scale Experiment.”
IEEE Transactions on Semiconductor Manufacturing 26 (4): 602–612.
Kacar, N., L. Mönch, and R. Uzsoy. 2016. “Modeling Cycle Times in Production Planning Models for Wafer Fabrication.” IEEE
Transactions on Semiconductor Manufacturing 29 (2): 153–167.
Kalir, A. A., and D. I. Grosbard. 2017. “Global Optimization of a Semiconductor IC Supply Chain Network.” Journal of Service
Science and Management 10: 338–352.
Kang, Y., E. Albey, S. Hwang, and R. Uzsoy. 2014. “The Impact of Lot-sizing in Multiple Product Environments with Congestion.”
Journal of Manufacturing Systems 33: 436–444.
Karmarkar, U. 1989. “Capacity Loading and Release Planning with Work-in-Progress (WIP) and Lead-times.” Journal of Manufactur-
ing and Operations Management 2: 105–123.
Kazan, O., R. Nagi, and C. Rump. 2000. “New Lot-sizing Formulations for Less Nervous Production Schedules.” Computers &
Operations Research 27 (13): 1325–1345.
Kefeli, A., and R. Uzsoy. 2016. “Identifying Potential Bottlenecks in Production Systems Using Dual Prices from a Mathematical
Programming Model.” International Journal of Production Research 54 (7): 2000–2018.
Kefeli, A., R. Uzsoy, Y. Fathi, and M. Kay. 2011. “Using a Mathematical Programming Model to Examine the Marginal Price of
Capacitated Resources.” International Journal of Production Economics 131 (1): 383–391.
Kim, B., and S. Kim. 2001. “Extended Model for a Hybrid Production Planning Approach.” International Journal of Production
Economics 73: 165–173.
4582 L. Mönch et al.

Kim, J., and S. Lim. 2012. “Order-lot Pegging for Minimizing Total Tardiness in Semiconductor Wafer Fabrication Process.” Journal
of the Operational Research Society 63 (9): 1258–1270.
Kim, S., and R. Uzsoy. 2008. “Integrated Planning of Production and Engineering Process Improvement.” IEEE Transactions on
Semiconductor Manufacturing 21 (3): 390–398.
Kim, S., and R. Uzsoy. 2013. “Modeling and Analysis of Integrated Planning of Production and Engineering Process Improvement.”
IEEE Transactions on Semiconductor Manufacturing 26 (3): 414–422.
Kim, Y., J. Bang, K. An, and S. Lim. 2008. “A Due-date-based Algorithm for Lot-order Assignment in a Semiconductor Wafer Fabri-
cation Facility.” IEEE Transactions on Semiconductor Manufacturing 21 (2): 209–216.
Kim, S. H., J. W. Kim, and Y. H. Lee. 2014. “Simulation-based Optimal Production Planning Model Using Dynamic Lead Time Esti-
mation.” The International Journal of Advanced Manufacturing Technology 75: 1381–1391.
Kim, J., S. Lim, S. Shim, and S. Choi. 2010. “Order-lot Pegging Heuristics for Minimizing Total Tardiness in a Semiconductor Wafer
Fabrication Facility.” In Proceedings of the 2010 IEEE International Conference on Industrial Engineering and Engineering
Management (IEEM), Macao, 1224–1229.
Knutson, K., K. Kempf, J. Fowler, and M. Carlyle. 1999. “Lot-to-order Matching for a Semiconductor Assembly & Test Facility.”
IIE Transactions 31 (11): 1103–1111.
de Kok, A. G., and J. C. Fransoo. 2003. “Planning Supply Chain Operations: Definition and Comparison of Planning Concepts.”
In Handbooks in Operations Research and Management Science: Supply Chain Management, edited by A. G. de Kok, and
S. C. Graves Vol. 11, Chapter 12, 597–675. Amsterdam: Elsevier.
de Kok, A. G., and K. Inderfurth. 1997. “Nervousness in Inventory Management: Comparison of Basic Control Rules.” European
Journal of Operational Research 103 (1): 55–82.
Kriett, P. O., S. Eirich, and M. Grunow. 2017. “Cycle Time-oriented Mid-term Production Planning for Semiconductor Wafer Fabrica-
tion.” International Journal of Production Research 55 (16): 4662–4679.
Leachman, R. 1993. “Modeling Techniques for Automated Production Planning in the Semiconductor Industry.” In Optimization in
Industry, edited by T. A. Ciriani, and R. C. Leachman, Chapter 1, 1–30. Chichester: Wiley.
Leachman, R. 2001. “Semiconductor Production Planning.” In Handbook of Applied Optimization, edited by P. Pardalos, M. Resende,
Chapter 18.6, 746–762, New York: Oxford University Press.
Leachman, R., and T. Carmon. 1992. “On Capacity Modeling for Production Planning with Alternative Machine Types.” IIE Transac-
tions 24 (4): 62–72.
Leachman, R., R. Benson, C. Liu, and D. Raar. 1996. “IMPReSS: An Automated Production-planning and Delivery-quotation System
at Harris Corporation – Semiconductor Sector.” Interfaces 26 (1): 6–37.
Leachman, R., and D. Raar. 1994. “Optimized Production Planning and Delivery Quotation for the Semiconductor Industry.” In Pro-
ceedings Optimization in Industry 2: Mathematical Programming and Modeling Techniques in Practice, edited by T. Ciriani,
and R. Leachman, Chapter 6, 63–72. Chichester: Wiley.
Lee, Y. H., S. Chung, B. Lee, and K. H. Kang. 2006. “Supply Chain Model for the Semiconductor Industry in Consideration of Man-
ufacturing Characteristics.” Production Planning & Control 17 (5): 518–533.
Lee, Y. H., and B. J. Lee. 2003. “Push-pull Production Planning of the Re-entrant Process.” International Journal of Advanced Manu-
facturing Technology 22: 922–931.
Li, M., F. Yang, R. Uzsoy, and J. Xu. 2016. “A Metamodel-based Monte Carlo Simulation Approach for Responsive Production Plan-
ning of Manufacturing Systems.” Journal of Manufacturing Systems 38: 114–133.
Liberopoulos, G. 2002. “Production Capacity Modelling of Alternative, Nonidentical, Flexible Machines.” International Journal of
Flexible Manufacturing Systems 14: 345–359.
Lin, J., and J. Chen. 2005. “Enhance Order Promising with ATP Allocation Planning Considering Material and Capacity Constraints.”
Journal of the Chinese Institute of Industrial Engineers 22 (4): 282–292.
Lin, P., and R. Uzsoy. 2016a. “Chance-constrained Formulations in Rolling Horizon Production Planning: An Experimental Study.”
International Journal of Production Research 54 (13): 3927–3942.
Lin, P., and R. Uzsoy. 2016b. “Estimating the Costs of Planned Changes Implied by Freezing Production Plans.” In Heuristics, Meta-
heuristics and Approximate Methods in Planning and Scheduling, edited by G. Rabadi, Chapter 2, 17–44. New York: Springer.
Lin, J., I. Hong, C. Wu, and K. Wang. 2010. “A Model for Batch Available-to-promise in Order Fulfillment Processes for TFT-LCD
Production Chains.” Computers & Industrial Engineering 59 (4): 720–729.
Liu, J., C. Li, F. Yang, H. Wan, and R. Uzsoy. 2011. “Production Planning for Semiconductor Manufacturing via Simulation Opti-
mization.” In Proceedings of the 2011 Winter Simulation Conference, Phoenix, AZ, 3617–3627.
Lowe, J. 2015. “Modeling and Analysis of Semiconductor Master Production Planning.” PhD thes., Industrial Engineering Depart-
ment, Clemson University.
Lowe, J., and S. Mason. 2016. “Integrated Semiconductor Supply Chain Production Planning.” IEEE Transactions on Semiconductor
Manufacturing 29 (2): 116–126.
Lyon, P., R. Milne, R. Orzell, and R. Rice. 2001. “Matching Assets with Demand in Supply-chain Management at IBM Microelec-
tronics.” Interfaces 31 (1): 108–124.
International Journal of Production Research 4583

Manda, A. B., R. Uzsoy, K. G. Kempf, and S. Kim. 2016. “Modeling the Impact of New Product Introduction on the Output
of Semiconductor Wafer Fabrication Facilities.” In Proceedings of the 2016 Winter Simulation Conference, Arlington, VA,
2547–2558.
McKay, K. 2011. “Field-based Research on Production Control.” In Planning Production and Inventories in the Extended Enterprise:
A State of the Art Handbook, vol. 1, edited by K. G. Kempf, P. Keskinocak, and R. Uzsoy, Chapter 10, 205–232. Berlin:
Springer.
Meyr, H., M. Wagner, and J. Rohde. 2015. “Structure of Advanced Planning Systems.” In Supply Chain Management and Advanced
Planning – Concepts, Models and Software, 5th ed. H. Stadtler, C. Kilger, and H. Meyr, Chapter 5, 99–106. Berlin: Springer.
Milne, R., S. Mahapatra, and C. Wang. 2015a. “Optimizing Planned Lead Times for Enhancing Performance of MRP Systems.” Inter-
national Journal of Production Economics 167: 220–231.
Milne, R., C. Wang, B. Denton, and K. Fordyce. 2015b. “Incorporating Contractual Arrangements in Production Planning.” Comput-
ers & Operations Research 53: 353–363.
Milne, R., C. Wang, C. A. Yen, and K. Fordyce. 2012. “Optimized Material Requirements Planning for Semiconductor Manufactur-
ing.” Journal of the Operational Research Society 63: 1566–1577.
Missbauer, H., and R. Uzsoy. 2011. “Optimization Models of Production Planning Problems.” In Planning Production and Inventories
in the Extended Enterprise: A State of the Art Handbook, vol. 1, edited by Kempf, K., Keskinocak, P., and Uzsoy, R., Chapter 16,
437–507, Berlin: Springer.
Mönch, L., J. Fowler, S. Dauzère-Pérès, S. Mason, and O. Rose. 2011. “A Survey of Problems, Solution Techniques, and Future
Challenges in Scheduling Semiconductor Manufacturing Operations.” Journal of Scheduling 14 (6): 583–599.
Mönch, L., J. Fowler, and S. Mason. 2013. Production Planning and Control for Semiconductor Wafer Fabrication Facilities: Model-
ing, Analysis, and Systems. New York: Springer.
Ng, T., Y. Sun, and J. Fowler. 2010. “Semiconductor Lot Allocation Using Robust Optimization.” European Journal of Operational
Research 205 (3): 557–570.
Omar, R., U. Venkatadri, and C. Diallo. 2016. “Estimating the Clearing Function for a Multi-product Production Network Based on
Mean Value Analysis.” Proceedings International Federation of Automatic Control, Troyes.
Orçun, S., R. Uzsoy, and K. G. Kempf. 2006. “Using System Dynamics Simulations to Compare Capacity Models for Production
Planning.” In Proceedings of the 2006 Winter Simulation Conference, Monterey, CA, 1855–1862.
Orçun, S., R. Uzsoy, and K. Kempf. 2009. “An Integrated Production Planning Model with Load-dependent Lead-times and Safety
Stocks.” Computers & Chemical Engineering 33 (12): 2159–2163.
Ovacik, I. 2010. “Advanced Planning and Scheduling Systems: The Quest to Leverage ERP for Better Planning.” In Planning Pro-
duction and Inventories in the Extended Enterprise: A State of the Art Handbook, vol. 1, edited by K. Kempf, P. Keskinocak,
and R. Uzsoy, Chapter 3, 33–44. New York: Springer.
Özer, O. 2010. “Inventory Management: Information, Coordination and Rationality.” In Planning Production and Inventories in the
Extended Enterprise: A State of the Art Handbook, vol. 1, edited by K. Kempf, P. Keskinocak, and R. Uzsoy, Chapter 13,
321–366. New York: Springer.
Pahl, J., S. Voß, and D. Woodruff. 2005a. “Production Planning With Load Dependent Lead Times.” 4OR 3: 257–302.
Pahl, J., S. Voß, and D. L. Woodruff. 2005b. “Load Dependent Lead Times – From Empirical Evidence to Mathematical Modeling.”
In Research Methodologies in Supply Chain Management, edited by H. Kotzab, S. Seuring, M. Müller, and G. Reiner, Chap-
ter 34, 539–554. Heidelberg: Physica.
Pahl, J., S. Voß, and D. Woodruff. 2007. “Production Planning With Load Dependent Lead Times: An Update of Research.” Annals
of Operations Research 153: 297–345.
Pibernik, R. 2005. “Advanced Available-to-promise: Classification, Selected Methods and Requirements for Operations and Inventory
Management.” International Journal of Production Economics 93–94: 239–252.
Pochet, Y., and L. Wolsey. 2006. Production Planning by Mixed Integer Programming. New York: Springer.
Ponsignon, T., and L. Mönch. 2012a. “Heuristic Approaches for Master Planning in Semiconductor Manufacturing.” Computers &
Operations Research 39 (3): 479–491.
Ponsignon, T., and L. Mönch. 2012b. “Using Iterative Simulation to Incorporate Load-dependent Lead Times in Master Planning
Heuristics.” In Proceedings of the 2012 Winter Simulation Conference, Berlin, Vol. 188, 1–11.
Ponsignon, T., and L. Mönch. 2014. “Simulation-based Performance Assessment of Master Planning Approaches in Semiconductor
Manufacturing.” Omega 46: 21–35.
Ravindran, A., K. Kempf, and R. Uzsoy. 2011. “Production Planning with Load-dependent Lead Times and Safety Stocks for a Single
Product.” International Journal of Planning and Scheduling 1 (1/2): 58–86.
Riaño, G. 2003. Transient Behavior of Stochastic Networks: Application to Production Planning with Load-dependent Lead Times.
Atlanta, GA: School of Industrial and Systems Engineering.
Seitz, A., and M. Grunow. 2017. “Increasing Accuracy and Robustness of Order Promises.” International Journal of Production
Research 55 (3): 656–670.
Smith, S. 1965. “Planning Transistor Production by Linear Programming.” Operations Research 13 (1): 132–139.
Soares, A., A. Azevedo, and J. de Sousa. 2000. “Distributed Planning and Control Systems for the Virtual Enterprise: Organizational
Requirements and Development Life-cycle.” Journal of Intelligent Manufacturing 11: 253–270.
4584 L. Mönch et al.

Sridharan, V., and R. LaForge. 1989. “The Impact of Safety Stock on Schedule Instability, Cost and Service.” Journal of Operations
Management 8 (4): 327–347.
Srinivasan, A., M. Carey, and T. Morton. 1988. Resource Pricing and Aggregate Scheduling in Manufacturing Systems. Technical
Report, Graduate School of Industrial Administration, Carnegie-Mellon University.
Stadtler, H. 2005. “Supply Chain Management and Advanced Planning – Basics, Overview and Challenges.” European Journal of
Operational Research 163 (3): 575–588.
Sun, Y., J. Fowler, and D. Shunk. 2011. “Policies for Allocating Product Lots to Customer Orders in Semiconductor Manufacturing
Supply Chains.” Production Planning & Control 22 (1): 69–80.
Thomas, L. J., and J. O. McClain. 1993. “An Overview of Production Planning.” In Handbooks in Operations Research & Manage-
ment Science, vol. 4: Logistics of Production and Inventory, edited by S. C. Graves, A. H. G. Rinnooy Kan, and P. H. Zipkin,
Chapter, 7, 333–370. Amsterdam: Elsevier.
Tian, F., S. Willems, and K. Kempf. 2011. “An Iterative Approach to Item-level Production and Inventory Planning.” International
Journal of Production Economics 133 (1): 439–450.
Toktay, L. B., and R. Uzsoy. 1998. “Capacity Allocation with Integer Side Constraints.” European Journal of Operational Research
109: 170–182.
Toktay, L., and L. Wein. 2001. “Analysis of a Forecasting-production-Inventory System with Stationary Demand.” Management
Science 47 (9): 1268–1281.
Tsai, K., and S. Wang. 2009. “Multi-site Available-to-promise Modeling for Assemble-to-order Manufacturing: An Illustration on
TFT-LCD Manufacturing.” International Journal of Production Economics 117: 174–184.
Uzsoy, R., C. Lee, and L. Martin-Vega. 1992. “A Review of Production Planning and Scheduling Models in the Semiconductor
Industry Part I: System Characteristics, Performance Evaluation and Production Planning.” IIE Transactions on Scheduling and
Logistics 24: 47–60.
Uzsoy, R., C. Lee, and L. Martin-Vega. 1994. “A Review of Production Planning and Scheduling Models in the Semiconductor
Industry Part II: Shop-floor Control.” IIE Transactions on Scheduling and Logistics 26: 44–55.
Vollmann, T. E., W. L. Berry, D. C. Whybark, and F. R. Jacobs. 2005. Manufacturing Planning and Control for Supply Chain Man-
agement. New York: McGraw-Hill.
Voß, S., and D. Woodruff. 2006. Introduction to Computational Optimization Models for Production Planning in a Supply Chain.
2nd ed. New York: Springer.
Wang, C., K. Fordyce, R. Milne, and R. Orzell. 2008. “The IBM Advanced Planning System for Managing Next Generation
Demand-supply Networks.” International Journal of Integrated Supply Management 4 (1): 125–140.
Wu, K., and L. McGinnis. 2012. “Performance Evaluation for General Queueing Networks in Manufacturing Systems: Characterizing
the Trade-off Between Queue Time and Utilization.” European Journal of Operational Research 221 (2): 328–339.
Wu, K., and L. McGinnis. 2013. “Interpolation Approximations for Queues in Series.” IIE Transactions 45 (3): 273–290.
Yang, L., K. Chen, Z. Lin, R. Qiang, X. Huang, and R. Lin. 2010. “Order Promising with Capacity Reserved for Multi-Priority
Orders.” In Proceedings 2010 International Conference on Management and Service Science (MASS), Wuhan, 1–4.
Zäpfel, G. 1984. “Systemanalytische Konzeption der Produktionsplanung und –steuerung für Betriebe der Fertigungsindustrie.” In
Sozio-technologische Systemgestaltung als Zukunftsaufgabe, edited by K. J. Zink, Chapter 3, 73–92. München: Carl Hanser
Verlag.
Zhao, Z. Y., M. O. Ball, and M. Kotake. 2005. “Optimization-based Available-to-Promise with Multi-stage Resource Availability.”
Annals of Operations Research 135: 65–85.
Ziarnetzky, T., and L. Mönch. 2016a. “Incorporating Engineering Process Improvement Activities into Production Planning Formula-
tions Using a Large-scale Wafer Fab Model.” International Journal of Production Research 54 (21): 6416–6435.
Ziarnetzky, T., and L. Mönch. 2016b. “Simulation-based Optimization for Integrated Production Planning and Capacity Expansion
Decisions.” In Proceedings of the 2016 Winter Simulation Conference, Arlington, VA, 2992–3003.
Ziarnetzky, T., N. Kacar, L. Mönch, and R. Uzsoy. 2015. “Simulation-based Performance Assessment of Production Planning Formu-
lations for Semiconductor Wafer Fabrication.” In Proceedings of the 2015 Winter Simulation Conference, Huntington Beach,
CA, 2884–2895.
Ziarnetzky, T., L. Mönch, and R. Uzsoy. 2017. Rolling Horizon Production Planning for Wafer Fabs with Chance Constraints and
Forecast Evolution. Submitted for publication.
Copyright of International Journal of Production Research is the property of Taylor & Francis
Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv
without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.

You might also like