You are on page 1of 10

The International Journal of Advanced Manufacturing Technology

https://doi.org/10.1007/s00170-018-2958-x

ORIGINAL ARTICLE

Human, AGV or AIV? An integrated framework for material handling


system selection with real-world application in an injection molding
facility
William Hellmann1 · Dean Marino1 · Mohamed Megahed1 · Merlin Suggs1 · Jeffrey Borowski1 ·
Ashkan Negahban1

Received: 31 July 2018 / Accepted: 29 October 2018


© Springer-Verlag London Ltd., part of Springer Nature 2018

Abstract
Motivated by a real-world material handling system selection problem, this paper proposes a framework that allows
for quantifying safety and incorporating it in multi-criteria decision-making processes that involve both quantitative and
qualitative measures. In the proposed framework, the results of failure mode and effect analysis (FMEA) for each alternative
are converted into a quantitative measure of total safety and reliability associated with that alternative. A modified
analytic hierarchy process (AHP) that differentiates between subjective and objective measures is then used to compare the
alternatives at hand. In this modified AHP, experts’ judgments are used for pairwise comparison of alternatives with respect
to qualitative measures, while for quantitative criteria, measured or estimated performance is directly used to obtain the
required pairwise comparisons. An Excel-based decision support tool that implements the proposed framework is developed
and made available online for researchers and practitioners. An application based on a real-world problem in an injection
molding facility is also presented.

Keywords Multi-criteria decision-making · Analytic hierarchy process · Failure mode and effect analysis ·
Automated guided vehicle · Automated intelligent vehicle

1 Introduction or carousels running through aisles/racks horizontally


and/or vertically.
Material handling is a critical component in many systems – Industrial robots that utilize an end of arm tooling
ranging from manufacturing [1, 2] and distribution centers (EOAT) attachment to hold, manipulate, and move parts
[3] to mining [4] and container terminals [5]. In manufac- and are directed by a combination of programmable
turing, material handling activities account for a significant software, sensors, and other controls.
portion of total manufacturing costs (different estimates – Automated guided vehicles (AGV) are computer-
have been reported that range from 15 to 70% [1, 6]). Mate- controlled and wheel-based that travel along the floor
rial handling is also known to play a crucial role in enhanc- of a facility without an on-board operator and their
ing productivity, flexibility, agility, and resource utilization movement is directed by a combination of software and
[1]. Over the past few decades, material handling technolo- sensor-based guidance systems.
gies have evolved from traditional hand trucks, carts, lift – Automated intelligent vehicles (AIV) are among the
trucks, overhead cranes, and conveyors into more advanced more recent technologies that do not have infrastructure
forms including but not limited to: requirements such as the guided paths made with wires,
magnetic tapes, or laser paths on the shop floor that
– Automated storage and retrieval systems (AS/RS) AGVs typically need. Therefore, AIVs can be deployed
that typically consist of computer-controlled cranes rapidly without the need for workspace retrofits. AIVs
can function autonomously (not just automatically) and
 Ashkan Negahban react to obstacles as opposed to traditional AGVs that
anegahban@psu.edu follow fixed pre-programmed paths and cannot drive
around obstacles. Moreover, AIVs can communicate
1 School of Graduate Professional Studies, The Pennsylvania and adapt to dynamic environments allowing them to
State University, Malvern, PA 19355, USA plan alternate routes and work collaboratively in a fleet
Int J Adv Manuf Technol

that enables automatic and dynamic selection of the proposed framework and currently being used by the
optimal vehicle to dispatch. For a sample of recent company. Finally, Section 5 provides the concluding
papers that study the use of AIVs as material handling remarks and potential avenues for future extensions.
equipment, see [7–9].
In many cases, when making decisions about a new or
2 Literature review
improved facility, multiple alternatives exist, among which,
the most appropriate material handling technology needs
There is a massive body of literature on optimization and
to be selected. However, this is a difficult decision-making
evaluation of the design, operation, and control of material
problem as there are many quantitative (e.g., load capacity,
handling systems, namely flow path layout, number and
cost, and energy consumption) and qualitative criteria (e.g.,
location of pickup and delivery points, collision avoidance
flexibility, reliability, and safety) to be considered, making and deadlocks, and vehicle requirements, dispatching,
this a multi-criteria decision-making (MCDM) problem. routing, and idle position. Several survey papers provide
Due to the importance of material handling as a determinant a review of these studies [1, 2, 10]. Esmaeilian et
of system performance and their high implementation al. [1] classify the literature on material handling for
costs, effective decision-making requires a careful and manufacturing systems into four main categories, namely
joint consideration and evaluation of both quantitative and design, operation, control, and selection. This paper deals
qualitative criteria. with the latter, i.e., the technology selection problem. In this
Motivated by a real-world project at an injection mold- section, we focus on recent studies on the application of
ing facility in Pennsylvania, USA, this paper proposes a FMEA and MCDM approaches in manufacturing systems
framework to be used during the design stage for a new and the material handling selection problem in particular
or improved facility and help select an appropriate mate- with the goal to illustrate the strengths and gaps in the
rial handling technology among a list of options when existing literature and highlight the contributions of this
the optimal design, operation and control policies for each work.
option have already been determined. This novel framework
integrates failure modes and effects analysis (FMEA) and 2.1 FMEA in manufacturing system design
analytic hierarchy process (AHP) to evaluate and consider and operation
both quantitative and qualitative criteria and support effec-
tive decision-making. To the best of our knowledge, this The traditional FMEA is a commonly used tool for
is the first study in the context of material handling selec- assessing the risk of possible design and process failures in
tion that explicitly evaluates safety and reliability (using manufacturing systems. For instance, Ozkok [11] considers
FMEA) and incorporates these considerations in a modified the hull structure production process of a shipyard and uses
AHP that differentiates between qualitative and quantita- historical data to determine the probability and severity of
tive criteria, where experts’ judgments are used to compare the failure modes identified. Ocampo et al. [12] provide
alternatives with respect to qualitative criteria (as in the tra- an analysis of various analysis tools on competitiveness of
ditional AHP), while measured/estimated performance are advanced manufacturing with a special focus on the apparel
used for pairwise comparisons with respect to quantita- industry in Central America. They recommend FMEA as a
tive criteria. Moreover, we develop an Excel-based decision useful tool for improving delivery time, cost, environmental
support tool that implements the proposed methodology. impacts, and flexibility.
The tool is made available online and hopes to trigger and The majority of FMEA applications in manufacturing,
facilitate future applications of the proposed framework by however, involve integrated frameworks that combine mul-
researchers and practitioners in any other decision-making tiple tools. Mejjaouli and Babiceanu [13] propose a holistic
problem that requires integration of FMEA and AHP. There- simulation-based monitoring and fault-recovery system that
fore, the contributions of this paper are based on both combines sensor network and distributed computing tech-
methodological and practical grounds. nologies with FMEA to improve agility and response to
The remainder of the paper is organized as follows. real-time uncertainties in a manufacturing environment.
Section 2 provides a critical analysis of the literature. Kuo and Huang [14] propose an expert system integrating
Section 3 describes the real-world material handling sys- a colored timed Petri net (CTPN), FMEA, and statistical
tem selection problem that motivated this work. Section 4 process control to enhance monitoring of flexible manufac-
presents the integrated methodology and illustrates its effec- turing systems (FMS). Shah et al. [15] propose a general
tiveness in supporting decision-making through its appli- value-risk based decision-making tool that combines per-
cation in the injection molding facility under study. The formance and risk with the decision-making process, where
section also describes a tool developed based on the FMEA is used to assess and prioritize the risk factors
Int J Adv Manuf Technol

identified in a manufacturing system. Chemweno et al. [16] programming model to account for the different constraints
propose a methodology combining FMEA, Fault Tree Anal- that need to be considered by the decision-maker.
ysis, Bayesian Networks, and Analytic Network Process Our analysis of this stream of research shows that
(ANP) for decision-making related to maintenance oper- an integrated framework combining MCDM methods and
ations and illustrate the application of their approach in FMEA (which is known to be a powerful tool for assessing
a process industry and an assembly line. Almannai et al. safety and reliability) is missing. This is an important gap
[17] propose an integrated tool combining FMEA and Qual- since the safety of personnel, equipment, and products is
ity Function Deployment (QFD) for selection of automa- often among the top priorities (if not the highest priority)
tion technologies and present an application of the tool in in virtually any company. Therefore, this paper contributes
the Rolls-Royce’ compression systems plant in Inchinnan, to this stream of research by proposing a framework that
Scotland for selection of chip forming machines. integrates FMEA and AHP for evaluating and incorporating
It is widely accepted that FMEA is useful (and often rec- safety and reliability considerations in the decision-making
ommended) prior to implementation of complex systems, process for selecting an appropriate material handling
such as a material handling system, to assess safety and reli- technology. It is worth noting that, while other MCDM
ability issues, identify potential effects that might arise from techniques could be adopted, the authors chose AHP as it
component/system malfunctions and failures, and deter- is the most commonly used MCDM method for material
mine critical system components and potential risks early handling selection [6].
in the design stage [18, 19]. In fact, the rise in focus on
safety, reliability, and risk management and mitigation has
made FMEA an integral part of Six Sigma, Production 3 Problem description: human, AGV or AIV?
Part Approval Process (PPAP), Advanced Product Quality
Planning (APQP), Tooling and Equipment (TE 9000), ISO We present the proposed methodology in the context of
9000, QS 9000, and ISO/TS16949 standards, all requiring a real-world material handling system selection problem
or recommending the use of FMEA [20]. Nevertheless, our at a manufacturing company in Pennsylvania, USA, that
literature review reveals that there is no article that use motivated this work in the first place. The company’s
FMEA, either by itself or as part of an integrated method- name and details of the facility layout as well as
ology, specifically to solve a material handling selection production and financial data are disguised to protect
problem—a gap in this stream of research that this paper the company’s proprietary information. The names of the
aims to address. vendor companies that supply and install the alternative
material handling technologies are also disguised.
2.2 MCDM methods for material handling system The company is a premier manufacturer of hinges, latch
selection and lock systems, swinghandle and multi-point systems,
rivets, and display mounts and has facilities in North
Our focus here is on studies that use MCDM approaches America, Europe and Asia Pacific. For one of their plants,
for selection of an appropriate material handling system. the company has a goal of improving the material handling
There are only a few studies that use only a single MCDM process between the injection molding facility and the
approach (this is also confirmed by the literature review storage area. Figure 1 provides a simplified layout of the
in [6]). For example, Chakraborty and Banik [21] use injection molding facility under study. Each cell in the
AHP and identify the most critical and robust criteria in facility consists of multiple machines (there are more than
the material handling equipment selection process through sixty machines distributed among these cells). The products
sensitivity analysis. The majority of the studies in this produced by the machines in each cell are placed in boxes
context, however, employ integrated MCDM approaches. that may contain hundreds or thousands of parts depending
To name a few, Onut et al. [6] use a fuzzy Technique for on the size of the product being produced. These boxes are
Order Preference by Similarity to Ideal Solution (TOPSIS) then placed in the corresponding pickup location(s) for that
to make a selection by using the criteria weights assigned by cell once they are ready to be picked up and transported to
a fuzzy ANP method. In a similar study, an integrated fuzzy the storage area.
AHP and fuzzy TOPSIS is proposed for AS/RS selection There is currently no automated process for transporting
[22]. Chan et al. [23] propose a tool consisting of three products between the two areas or for loading/unloading of
modules, namely a database to store specifications of the materials. These are performed by an operator using a pow-
different alternatives, a knowledge-based expert system, and ered jack truck. There is an interest to determine whether
an AHP module to select an appropriate equipment type. there might be advantages in using robotic solutions. The
For another study that uses a similar methodology, see [24]. motivation and potential benefit for improving this pro-
Braglia et al. [25] use AHP in conjunction with an integer cess is that while the material is being transported from
Int J Adv Manuf Technol

Table 1 Estimated performance of the alternatives considered by the


company with respect to quantitative measures

Alternative Number of Surplus capacity Payback period


transporters (boxes per day) (months)
required

AGV 1 (vendor 1) 3 460 52


AGV 2 (vendor 2) 3 270 43
AGV 3 (vendor 3) 4 220 75
AIV 1 (vendor 4) 5 300 84
AIV 2 (vendor 5) 5 180 65
Additional material 2 100 72
handlers

production and financial data provided by the company


Fig. 1 The layout, pickup points, and picking routes for the injection cannot be disclosed, we consider a scenario were these
molding facility. As for the picking route, the material handling estimates are already available through some preliminary
operator/equipment would go to a specific pickup location and return analysis.
to the storage area with a package/pallet. They would then move to The above represents a multi-criteria decision-making
a subsequent pickup location to retrieve the material. The company
refers to this approach as star picking problem consisting of both quantitative and qualitative
criteria that also requires a method to quantify safety and
reliability to include it directly in the decision-making
one portion of the facility to another via an AGV or AIV, process. Figure 2 illustrates the hierarchy for selection of the
the current material handler can focus on putting materi- most appropriate material handling equipment using the five
als away in the storage facility. The candidate solutions criteria and six alternatives specified by the company. The
(alternatives) identified by the company are (1) hiring addi- company consulted a research team from a local university
tional material handlers, (2) install an AGV system, or (3) (our team) to help with decision-making on the most
install an AIV system. The company has also determined a appropriate solution. Due to the high priority given to safety
set of objectives/criteria for evaluating the alternatives: (1) and reliability, and the lack of an approach that explicitly
payback period, (2) surplus capacity defined as the poten- evaluates and incorporates these considerations into the
tial to handle increased demand/production in the future, decision-making process for selection of material handling
(3) congestion, (4) ease of implementation and control, and technology, the following methodology is proposed.
(5) safety and reliability. In particular, safety and reliability
of the system to be installed is extremely important to the
company executives. 4 The proposed methodology
The company has identified a set of AGV and AIV and its implementation
vendors, who have provided cost estimates for acquisition
and maintaining the respective technology as well as the Figure 3 summarizes the proposed framework integrating
optimal design and operating policy for their proposed a modified AHP with FMEA. The following subsections
solution if selected and implemented. Table 1 provides the describe the different components of the methodology and
number of transporters required, surplus capacity, and the its implementation in an Excel-based tool to solve the
estimated payback period for the six alternatives under above material handling selection problem. For the sake of
investigation. The surplus capacity for each alternative conciseness, an extensive explanation of FMEA and AHP
represents the (nominal) maximum number of boxes that is avoided. For a detailed description on the theory and
can be transported per day using the material handling justification for these tools, see [18] and [26].
system minus the current average daily production (boxes
per day). The company uses the number of transporters 4.1 Modified AHP: Differentiating between
as a measure of congestion in the facility caused by qualitative and quantitative criteria
material handling operations. The payback period for each
alternative is estimated based on its initial cost, estimated The modified AHP specifically distinguishes between
monthly operating and maintenance costs, and expected qualitative and quantitative criteria. A standard 1–9 scale
revenue increase due to improved throughput. Since the proposed by Saaty [26] is used by the experts for pairwise
Int J Adv Manuf Technol

Fig. 2 The AHP for the material Goal: Selecting the most appropriate
handling equipment selection material handling system
problem

Safety Control Surplus capacity Congestion Payback period

Additional AGV 1 AGV 2 AGV 3 AIV 1 AIV 2


material handlers (Vendor 1) (Vendor 2) (Vendor 3) (Vendor 4) (Vendor 5)

comparisons involving qualitative criteria. For quantitative The modified AHP presented here alleviates these issues
criteria, however, measured or estimated performance is (at least partially). In this modified AHP, the pairwise
used for comparisons. This approach is justified because comparison matrix A is still defined as in the original AHP
of the following problems that may arise from using the so that A = (aij ) is an n × n matrix with positive entries
commonly used standard scale for quantitative criteria [27]: that satisfy the reciprocal property aij = 1/aj i , ∀i, j ∈
,  = {1, 2, ..., n}. For qualitative measures, aij values
– The standard integer scale can be easily misused. For
are based on the experts’ judgments using the standard
example, the value 3 that indicates a weak preference
scale. However, when comparing quantitative criteria for
essentially assigns three times more weight to the
two elements i and j with measured/estimated performance
slightly preferred component.
values wi and wj , the preference of element A to B will be
– Valid quantitative performance data will be ignored.
aij = wi /wj . Therefore, for a quantitative criterion where
Suppose alternative A costs 1.1 times more than B. It is
a larger performance value is preferable (e.g., profit), the
not clear what judgment value in the standard scale can
pairwise comparison matrix of the n alternatives is given by
best represent this difference since 1 represents equiva-
⎡ ⎤
lence and 3 essentially indicates three times difference. w1 /w1 w1 /w2 ... w1 /wn
Therefore, there is a need to use the available mea- ⎢ w2 /w1 w2 /w2 ... w2 /wn ⎥
⎢ ⎥.
sured/estimated data directly in the decision-making ⎣ ... ⎦
process. wn /w1 wn /w2 ... wn /wn
– The experts may have different preferences that may
compromise their true judgment. In the case presented The pairwise comparison of two alternatives i and
here, some of the stakeholders were completely for and j with respect to a negative criterion such as cost or
some were completely against a robotic solution. These congestion, where a smaller performance value is desirable,
differences can easily lead to bias in the final recom- is given by aij = wj /wi , where wi and wj represent
mendation if stakeholders’ judgment matrices are used the measured/estimated performance for alternatives i and
for comparing alternatives with respect to quantitative j , respectively. The pairwise comparison matrix for n
measures. alternatives with respect to a negative criterion is formed by
the inverse of the ratios as follows
⎡ ⎤
w1 /w1 w2 /w1 ... wn /w1
⎢ w1 /w2 w2 /w2 ... wn /w2 ⎥
⎢ ⎥.
⎣ ... ⎦
w1 /wn w2 /wn ... wn /wn

Therefore, pairwise comparisons in this modified AHP


can be summarized as follows:
– For qualitative criteria, use experts’ judgment and the
standard 1–9 scale for the pairwise comparison of
alternatives.
– For a quantitative criterion where a larger performance
value is preferable, use the ratios of estimated/measured
performance to form the pairwise comparison matrix.
– For a quantitative criterion where a smaller perfor-
mance value is desirable, use the inverse ratios of
estimated/measured performance to form the pairwise
Fig. 3 The proposed methodology comparison matrix.
Int J Adv Manuf Technol

Fig. 4 Specifying the criteria and alternatives in the Excel-based tool. Safety or FMEA is a required name in order for the FMEA results to be
automatically included in the AHP analysis

4.2 FMEA: Quantifying safety and reliability the results of the FMEA in quantitative terms. For any
alternative α, we have
In order to evaluate and incorporate safety and reliability 

in the decision-making process, we use the total risk TRPNα = α
Sm × Om
α
× Dm
α
, (1)
priority number (TRPN) metric proposed in [28] to express m=1

Fig. 5 Pairwise comparison of criteria and alternatives in the Excel- respectively. For more information on the description and justifica-
based tool. The modified AHP differentiates between quantita- tion of the 1–9 scale used for subjective pairwise comparisons in AHP,
tive and qualitative criteria indicated by gray and orange cells, see [26, 29]
Int J Adv Manuf Technol

where Mα represents the number of failure modes identified for free via the tool’s webpage. We hope that this not only
for alternative α, and for any failure mode m, Sm α denotes enables replicability of the results presented in this paper,
the severity, Omα is the probability/frequency of occurrence, but also allows researchers and practitioners to use the
α
and Dm is the probability of detection. The TRPN index proposed methodology in any other multi-criteria decision-
is then treated as a quantitative measure in the modified making problem that requires FMEA and AHP integration.
AHP to compare the alternatives in terms of safety and It is worth noting that a different version of this tool is
reliability. Since a smaller TRPN is desirable, the inverse currently used by the company studied in this work at its
ratio of TRPN values will be used to form the corresponding various facilities around the world.
pairwise comparison matrix. The first step in using the Excel-based tool is to set up
the workbook by specifying the criteria and alternatives for
4.3 Implementation and results the decision-making problem at hand as shown in Fig. 4.
When defining the criteria, the user needs to specify their
The goal of this section is twofold: (1) present the applica- type (qualitative or quantitative). Also, the name Safety or
tion of the proposed methodology in the material handling FMEA is required in order for the FMEA analysis to be
system selection problem discussed in Section 3 and (2) automatically included in the AHP. Once the criteria and
provide a tutorial for an Excel-based tool that implements alternatives are defined, the user needs to press the Set
the proposed framework. The tool is made available online up Workbook button to create a worksheet called Input

Fig. 6 A sample FMEA analysis for AGV 1 supplied by vendor 1. only include additional verbal descriptions and do not affect the TRPN
The columns for potential effect(s), cause(s), process control, recom- calculation). The list of failure modes are based on AGV FMEA anal-
mended action, and other standard fields in an FMEA form are not ysis in the literature [30] with a few additional modes defined by the
shown for enhanced representation (note that these hidden columns company
Int J Adv Manuf Technol

Table 2 Total risk priority number (TRPN) for the alternatives engineers to determine these subjective pairwise compar-
determined by their respective FMEA results isons. Looking at the pairwise comparison of alternatives
Alternative Total risk priority with respect to Control, we observe that the panel were
number (TRPN) especially skeptical about the level of control over the oper-
ation of AIV systems. The panelists’ judgment suggests
AGV 1 (vendor 1) 2014 that human material handlers provide maximum control,
AGV 2 (vendor 2) 2516 followed by AGV 1 that involves taped/wired paths,
AGV 3 (vendor 3) 3160 then AGV 2 and 3 that involve wireless directed paths.
AIV 1 (vendor 4) 4182 Pairwise comparison of alternatives with respect to quan-
AIV 2 (vendor 5) 3772 titative criteria, however, are indicated by gray cells and
Additional material handlers 1748 determined as described in Section 4.1 according to the
measured/estimated performance of the alternatives as pro-
vided in Table 1. For instance, the pairwise comparison
containing blank pairwise comparison matrices required for of “AGV 1 vs AGV 2” with respect to Payback Period is
the AHP analysis as well as worksheets containing blank computed as the inverse of the ratio of their estimated pay-
back periods given by ( 52 −1 = 0.83. Note that the inverse
FMEA forms for each alternative. 43 )
The next step is to complete the pairwise comparison ratio is used for quantitative measures where a smaller
matrices in the newly created Input worksheet. In this value is preferable (that is, for payback period, congestion,
worksheet, the orange cells indicate pairwise comparison and FMEA’s total risk priority number referred to as Safety
for qualitative measures, which will be based on experts’ index in this example), while the actual ratio is used for
judgment. The orange cells in Fig. 5 indicate that, for this quantitative measures where a larger value is desirable as for
case study, the pairwise comparison of criteria (shown on Surplus Capacity in this case. Therefore, using the estimates
the left) and the pairwise comparison of alternatives with in Table 1, the pairwise comparison of “AGV 1 vs AGV 2”
respect to criterion Control (shown in the table on the right) 270 ) = 1.70 as shown
with respect to Surplus Capacity is ( 460
are based on experts’ judgments. The research team worked in Fig. 5.
with a panel consisting of top executives (including the As described in Section 4.2, safety is quantified based
Director of Operational Excellence), shop floor managers on the total risk priority number (TRPN) obtained from the
(including production and quality managers), and process FMEA analysis and calculated by Eq. 1. In other words,

Fig. 7 A screenshot of the AHP tab crated by the decision support tool comparison matrices and their consistency ratio (the figure only shows
containing the final results. The AHP tab also contains a chart of the the details related to weighting of criteria)
AHP hierarchy (similar to Figure 2) and details of all other pairwise
Int J Adv Manuf Technol

each pairwise comparison of alternatives with respect to and reliability and obtain a total risk priority number
safety, shown in the table on the right in Fig. 5, is the (TRPN) as a global measure of risk. Once the risk associated
inverse ratio of the TRPN for the two alternatives being with each alternative is quantified using TRPN, the results
compared. The research team worked with the same panel of are fed into a modified AHP that differentiates between
representatives from the company to perform a comprehen- quantitative and qualitative criteria. This allows for direct
sive FMEA analysis considering both design and process use of available data on measured/estimated performance
failure modes for a material handler operator, AGVs, and of alternatives with respect to the quantitative criteria in
AIVs. In the Excel-based tool, the FMEA results are speci- the decision-making process. It also prevents misuse and
fied in separate FMEA tabs (one worksheet per alternative), inaccuracies that would be introduced if experts’ judgments
which are then used by the tool to compute the TRPN values were used instead of actual data. This is especially
for the alternatives and automatically complete the corre- important since the experts may have different preferences
sponding pairwise comparisons in the Input tab under the that may result in a bias toward certain alternatives. For
Safety column. Figure 6 presents a sample FMEA for AGV qualitative criteria, however, using experts’ judgments is
1 by Vendor 1 and Table 2 summarizes the total risk prior- inevitable due to the lack of data. An Excel-based tool
ity number for all alternatives. Since a smaller total TRPN that implements the proposed framework is developed and
is desirable, the pairwise comparisons are obtained by the made available online via the tool’s webpage. A real-world
inverse of the TRPN ratios. For example, the pairwise com- material handling equipment selection problem is used to
parison for “AGV 1 vs AGV 2” with respect to Safety is illustrate the applicability of the proposed methodology and
given by ( 2014 −1 = 1.25 as shown in Fig. 5.
2516 ) accompanying tool. There are a few important things to
Once all pairwise comparisons are specified, the user note:
needs to click on the Calculate Results button (see Fig. 5)
and the tool generates a new tab called AHP containing the – While we use TRPN as a measure of overall safety
final scores and ranks for the alternatives as well as the and reliability risks [28], many other possibilities exist
details of the AHP analysis including the consistency ratio for quantifying FMEA results. For example, one could
(CR) for each pairwise comparison matrix (a CR ≤ 0.1 use the maximum RPN value. Note that each of these
is generally considered acceptable in the AHP literature). metrics has its advantages and disadvantages. The
Therefore, the newly created AHP worksheet contains maximum RPN can be an effective measure when there
all information necessary for decision-making. Figure 7 is a significant difference between the RPN values
summarizes the final results reported by the decision for the identified failure modes or when one of the
support tool. As suggested by the table on top, the modified alternatives has a failure mode with an extremely high
AHP recommends hiring additional material handlers as the RPN. In such cases, using the average (or sum) of RPN
solution with the highest overall score followed by AGV 1 values could result in a biased metric that ignores such
and AGV 2. This is primarily due to the fact that a human extreme failure modes. Future research could focus on
operator is determined to result in higher safety and control determining the most appropriate metric under various
while minimizing congestion as it requires the smallest scenarios. The proposed tool can then be extended so
number of material handler equipment (2 compared to 3–5 that the user specifies what metric to use (average, sum,
for other alternatives as shown in Table 1). This is mainly maximum, median, etc.).
because the powered jack trucks used by operators are – We chose AHP for the initial implementation of the
larger than the candidate AGV and AIV devices considered methodology mainly because it is the most commonly
in the study. Therefore, fewer jack trucks are needed to used tool in this context. However, any other MCDM
keep up with the production rate, reducing congestion in tool (e.g., data envelopment analysis [31]) could be
the facility. Similar results were obtained for the actual used.
project. – Here, the original FMEA was adopted since many prac-
titioners and companies use this method and are famil-
iar with it. However, any other failure/safety assessment
5 Summary and conclusions tool or variants of FMEA could be used. Future exten-
sions can investigate the efficacy of these alternative
Motivated by a real-world material handling system approaches. The tool can also be extended to support
selection problem, this paper proposes a novel and these different methods.
integrated methodology to explicitly incorporate safety and – A limitation of the proposed methodology is that it does
reliability considerations into the decision-making process not account for uncertainties in the measured/estimated
while distinguishing between qualitative and quantitative performance and/or experts’ judgments. This can be
criteria. More specifically, FMEA is used to assess safety addressed by replacing the traditional AHP with the
Int J Adv Manuf Technol

enhanced Monte Carlo AHP approach proposed in [32] 13. Mejjaouli S, Babiceanu RF (2014) Holonic condition monitoring
to support uncertainties in pairwise comparison matri- and fault-recovery system for sustainable manufacturing enter-
prises. In: Borangiu T, Trentesaux D, Thomas A (eds) Service ori-
ces in an individual or group decision-making setting entation in Holonic and Multi-Agent Manufacturing and Robotics.
while still maintaining consistency requirements. Springer International Publishing, pp 31–46
– The proposed methodology and tool are valid and appli- 14. Kuo C-H, Huang H-P (2000) Failure modeling and process
cable in any decision-making problem where FMEA monitoring for flexible manufacturing systems using colored
timed Petri nets. IEEE Trans Robot Autom 16(3):301–312
is determined to be an important factor to be directly
15. Shah LA, Etienne A, Siadat A, Vernadat F (2016) Decision-
included in the decision-making process along with making in the manufacturing environment using a value-risk
other quantitative and qualitative measures, regardless graph. J Intell Manuf 27(3):617–630
of what specific alternatives are under consideration. 16. Chemweno P, Pintelon L, Van Horenbeek A, Muchiri P (2015)
Therefore, there are many possible applications beyond Development of a risk assessment selection methodology for asset
maintenance decision making An analytic network process (ANP)
material handling system selection. The Excel-based approach. Int J Prod Econ 170:663–676
tool also significantly facilitates sensitivity analysis to 17. Almannai B, Greenough R, Kay J (2008) A decision support tool
assess the effect of inclusion/exclusion of certain crite- based on QFD and FMEA for the selection of manufacturing
ria and/or changing the pairwise comparison matrices. automation technologies. Robot Comput Integr Manuf 24(4):501–
507
All the user needs to do is set up the workbook using 18. Stamatis DH (2003) Failure mode and effect analysis: FMEA from
the new set of inputs and rerun the calculations. theory to execution, 2nd edn. ASQ Quality Press, Milwaukee
19. Sikich J (1998) Development and implementation of an auto-
Publisher’s Note Springer Nature remains neutral with regard to mated wafer transport system. In: Proceedings of the IEEE/SEMI
jurisdictional claims in published maps and institutional affiliations. Advanced Semiconductor Manufacturing Conference and Work-
shop, pp 400–404
20. Shoults LW (2016) Implementation of design failure modes and
effects analysis for hybrid vehicle systems. MS thesis, Virginia
References Polytechnic Institute and State University
21. Chakraborty S, Banik D (2006) Design of a material handling
equipment selection model using analytic hierarchy process. Int J
1. Esmaeilian B, Behdad S, Wang B (2016) The evolution and future
Adv Manuf Technol 28(11):1237–1245
of manufacturing: a review. J Manuf Syst 39:79–100
22. Aktan HE, Tosun O (2013) An integrated fuzzy AHP – fuzzy
2. Negahban A, Smith JS (2014) Simulation for manufacturing
TOPSIS approach for AS/RS selection. Int J Prod Qual Manag
system design and operation: literature review and analysis. J
11(2):228–245
Manuf Syst 33(2):241–261
3. Roodbergen KJ, Vis IFA (2009) A survey of literature on 23. Chan FTS, Ip RWL, Lau H (2001) Integration of expert system
automated storage and retrieval systems. Eur J Oper Res with analytic hierarchy process for the design of material handling
194(2):343–362 equipment selection system. J Mater Process Technol 116(2):137–
4. Burt CN, Caccetta L (2014) Equipment selection for surface 145
mining: a review. Interfaces 44(2):143–162 24. Chan FTS (2002) Design of material handling equipment selection
5. Lau HYK, Zhao Y (2008) Integrated scheduling of handling system: an integration of expert system with analytic hierarchy
equipment at automated container terminals. Int J Prod Econ process approach. Integr Manuf Syst 13(1):58–68
112(2):665–682 25. Braglia M, Gabbrielli R, Miconi D (2001) Material handling
6. Onut S, Kara SS, Mert S (2009) Selecting the suitable material device selection in cellular manufacturing. J Multi-Criteria Decis
handling equipment in the presence of vagueness. Int J Adv Manuf Anal 10(6):303–315
Technol 44(7):818–828 26. Saaty TL (1980) The analytic hierarchy process. McGraw-Hill,
7. Zaghdoud R, Collart-Dutilleul S, Ghedira K, Mesghouni K, Zidi New York
K (2013) A multi-objective approach for assignment containers 27. Weber SF (1993) A modified analytic hierarchy process for
to AIVs in a container terminal. In: Proceedings of the IEEE automated manufacturing decisions. Interfaces 23(4):75–84
International Conference on Systems, Man, and Cybernetics, pp 28. Silvestri A, DeFelice F, Petrillo A (2012) Multi-criteria risk
2460–2466 analysis to improve safety in manufacturing systems. Int J Prod
8. Mhalla A, Collart-Dutilleul S, Zhang H (2016) Robust control Res 50(17):4806–4821
under uncertainty for seaport handling equipments. Transportation 29. Meesariganda BR, Ishizaka A (2017) Mapping verbal AHP scale
Research Procedia 14:203–212 to numerical scale for cloud computing strategy selection. Appl
9. Mhalla A, Benrejeb M, Zhang H (2016) Maintenance processes Soft Comput 53:111–118
for container handling equipment using P-time petri nets. J Eng 30. Yan R, Dunnett SJ, Jackson LM (2016) Reliability modelling of
Res 4(4):99–112 automated guided vehicles by fault tree analysis. In: Proceedings
10. Vis IFA (2006) Survey of research in the design and control of of the 5th Conference on Operational Research, pp 3–10
automated guided vehicle systems. Eur J Oper Res 170(3):677– 31. Azadeh A, Moghaddam M, Asadzadeh SM, Negahban A (2011)
709 An integrated fuzzy simulation-fuzzy data envelopment analysis
11. Ozkok M (2013) Risk assessment in ship hull structure production algorithm for job-shop layout optimization: the case of injec-
using FMEA. J Mar Sci Technol 22(2):173–185 tion process with ambiguous data. Eur J Oper Res 214(3):768–
12. Ocampo JR, Hernandez-Matias JC, Vizan A (2017) A method 779
for estimating the influence of advanced manufacturing tools on 32. Negahban A (2018) Optimizing consistency improvement of
the manufacturing competitiveness of Maquiladoras in the apparel positive reciprocal matrices with implications for monte carlo
industry in Central America. Comput Ind 87:31–51 analytic hierarchy process. Comput Ind Eng 124:113–124

You might also like