6 G.M. Komaki et al.
Lagodimos, Mihiotis, and Kosmidis (2004) modelled commercial refrigerator manufacturing line as an AF with several
machines in the assembly stage. In this application, components such as sheets and frames are processed by the metal
cutter and press machines and they are assembled together to form three major sub-assemblies: the basin, the cabin, and
doors. Yokoyama (2004) referred to the manufacturing process in the garment industry as a suitable application for AF
models. Yokoyama (2004) also formulated a cloth making shop floor as AF. In this model, a cutting machine produces
cloth pieces and a sewer knits the pieces and add parts such as buttons to make jackets with different sizes and styles.
Blocher and Chhajed (2008) pointed out applications of AF models in the supply chain management. Hwang and Lin (2012)
modelled a food and fertiliser production facility as AF problem where there are both common and unique ingredients in
every product. In this process, basic ingredients are made in batches and then these ingredients are blended in assembly
stage according to customer’s recipe. Terekhov et al. (2012) illustrated supply chain for Alcatel-Lucent Corporation as
an AF problem with two manufacturing facility and one assembly facility. Alcatel-Lucent’s products are shipped to a
regional facility where they are packaged with necessary ancillary equipment before they are being shipped to the customer’s
location. Cheng (2012) investigated the application of lot streaming in two-stage assembly lines. This AF line consists of
multiple suppliers at the first stage and one or more assembly locations in the second stage. The model is utilised by several
companies such as Dell where the customer chooses different components of a laptop from the Dell website. As soon as
Dell receives the order, it obtains sub-assemblies from suppliers (first stage) and then assembles the laptop (second stage)
upon receiving all sub-assemblies. Author’s rationale is that AF can reduce operational costs while at the same time, it
increases customer satisfaction. Industrial examples of AF problem with sequence-dependent set-up times are studied by
Liao, Lee, and Lee (2015) s uch as motor gear factory that produces and assembles main parts of the motor (stator and
rotor). Chen et al. (2015) modelled a paint manufacturing case study as an AF problem with performance measures of
production simultaneity and shipment punctuality. Lalami, Frein, and Gayon (2017) discussed an AF case study in PSA
Peugeot Citroen Company. The machining stage in Peugeot Citroen Company is responsible for making engine block,
cylinder head, crankshaft, and connecting rods. The presented model satisfies multiple performance measures including
safety stock threshold, stock balance and stability among products, and levelling car engines production or planning stability.
Application of AF problem in service industry was addressed by Zhang, Zhou, and Liu (2010) where authors modelled a
commercial multi-page invoice company that is composed of a single stencil preparation line in the first stage, parallel
printing lines in the second stage, and parallel assembly lines for making final invoices.
Another important application of AF problems is in semi-conductor manufacturing, see Hayrinen et al. (2000), Jin et al.
(2002), Bard et al. (2015), and Gholami-Zanjani et al. (2017). The semi-conductor manufacturing process starts from wafer
fabrication (a.k.a. front-end operations). Wafers are then sent to assembly and test facility (a.k.a. back-end operations) where
they are converted to chips, and then packaged and tested. In testing stage, a single machine can be used to perform different
testing steps. Sawik, Schaller, and Tirpak (2002) presented an application of AF problems for printed wiring board assembly
in surface mount technology lines. Sarin, Yao, and Trietsch (2011) proposed two applications of AF problem with single
batch; one in fabricating integrated circuits (ICs) and power supplies and IC assembly on printed circuit boards and the other
in producing and assembling engines and transmissions.
2.2. Complexity of AF problems
In AF scheduling problems, there is precedence constraint between machining and assembly operation that increases the
complexity of these problems comparing with standard flow shops. Lenstra and Rinnooy Kan (1978) provide a summary
of complexity of problems in the presence of precedence constraints and shows that most of the cases are NP-Complete.
Potts et al. (1995) showed that AF(m,1)||C
max
for m ≥ 2 with simple products is NP-hard; therefore, any extension of this
problem is NP-hard due to increase in the number of constraints. Koulamas and Kyparisis (2004) showed that minimising
makespan of distributed flow shop is NP-hard, therefore, DAF is also NP-hard. It is also known that F2||
C
i
and F2||L
max
are NP-hard; therefore, adding assembly operations to any of these problems will make them NP-hard in strong sense. On
the other hand, AF2||TT, and AF2||TC are known to be NP-hard according to Lee, Cheng, and Lin (1993), and Tozkapan,
Kırca, and Chung (2003). Based on these arguments, extension of the problems including multi-objective AF, AF with more
than two stages, HAF, and DAF are also NP-hard.
2.3. Search methodology
For this review paper, WEB OF SCIENCE™ (2018) and Scopus (2018) are selected as the database. Authors performed
independent searches on these databases and crosschecked their results for higher accuracy. The keywords used to identify
AF journal papers, conference papers, and books are based on the prior experience of the authors and the keywords in the
cited articles. We proposed three level keyword structure that captures AF publications. Levels 1 and 2 define keywords that