Journal of Manufacturing Systems · the modeling of product design [8,9]. Product portfolio design opti-mization, incorporating reuse, remanufacturing and recycling, to bal-ance tradeoffs

Contents lists available at ScienceDirect

Journal of Manufacturing Systems

journal homepage: www.elsevier.com/locate/jmansys

Quantifying impacts of product return uncertainty on economic andenvironmental performances of product configuration design

Ridvan Aydina,⁎, Adam Browna,b, Fazleena Badurdeena,b, Wei Lia,b, Keith E. Roucha,b,I.S. Jawahira,b

a Institute for Sustainable Manufacturing, Lexington, KY, 40506, United StatesbDepartment of Mechanical Engineering, University of Kentucky, Lexington, KY, 40506, United States

A R T I C L E I N F O

Keywords:End-of-life recoveryProduct return rateUncertaintyMonte Carlo simulation

A B S T R A C T

The product returns involve considerable uncertainties that have an impact on the economic (i.e., total cost) andenvironmental (i.e., global warming potential, water use and energy use) performance measures of a productconfiguration design. This is because it directly affects the number of reusable, remanufacturable and/or re-cyclable components/items. However, impact of the uncertainty of product return rate on the economic andenvironmental performance of new product configuration designs has not been addressed in literature. In thisstudy, a methodology is proposed to quantify the impact of product return rate uncertainty using Monte Carlosimulation. The proposed methodology is implemented on an industrial case study for quantifying the impact ofproduct return rate uncertainty on the economic and environmental performance of toner cartridge config-uration designs. The results of this study can provide useful information on the variation of total lifecycle cost,global warming potential, total water use, and total energy use of product configuration designs due to theuncertainty of product return rate.

1. Introduction

Growing global consumption and faster product retirement duerapid technological advancement, have led to a majority of productsbeing discarded after their use. These wasted products often lead toenvironmental pollution and the loss of the remaining value of the usedproducts, particularly if they have not reached their end-of-life (EoL).Traditional manufacturing practices have not focused on recovering thevalue from EoL products (as well as end-of-use (EoU) products that havenot reached their EoL after the use stage of the product lifecycle [1].Sustainable manufacturing practices by adopting a 6R (Reduce, Reuse,Recycle, Redesign, Recover and Remanufacture) methodology enables atotal lifecycle-based closed-loop material flow [1,2]. Implementingproduct EoL recovery strategies, such as reuse, remanufacturing, andrecycling, can help companies mitigate environmental impact andconform to strict regulations, while increasing global manufacturingcompetitiveness and promoting sustainable economic growth [3,4].

A closed-loop supply chain involves collecting used products fromcustomers and performing product recovery strategies, such as reuse,remanufacturing, and recycling as well as disposing unrecoverablecomponents/materials safely [5]. In closed-loop systems, products,components and materials can be utilized multiple times over multiple

lifecycles before landfilled [6,7]. However, recycled materials arecommonly used in different applications that leads to a challenge toclose the loop in industrial practices [7].

EoL recovery strategies and lifecycle issues have been considered inthe modeling of product design [8,9]. Product portfolio design opti-mization, incorporating reuse, remanufacturing and recycling, to bal-ance tradeoffs between cost, reliability, and environmental impact ofproducts were studied by Mangun and Thurston [10]. A multi-objectiveevolutionary algorithm was proposed by Jun et al. [11] to select EoLproduct recovery options (i.e., reuse, remanufacturing, reconditioning,or disposal and replacement). Mixed-integer programming for productfamily design profitability when selling used products [12] and a de-cision-support model to determine the optimal design of new and re-manufactured products simultaneously and the number of returnedproducts while investigating the trade-off between total profit and en-vironmental impact [13] have been studied.

However, there are considerable uncertainties associated with EoL(and EoU) product returns (which collectively will be referred to as‘product returns’ henceforth) that can have an effect on the economic(e.g., cost, profit) and environmental (e.g., emissions, energy use) im-pacts of recovering EoL products and using them in subsequent lifecycleproducts. Thus, the non-deterministic parameters related to product

https://doi.org/10.1016/j.jmsy.2018.04.009Received 28 November 2017; Received in revised form 26 February 2018; Accepted 15 March 2018

⁎ Corresponding author.E-mail address: [email protected] (R. Aydin).

Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

0278-6125/ © 2018 Published by Elsevier Ltd on behalf of The Society of Manufacturing Engineers.

Please cite this article as: Aydin, R., Journal of Manufacturing Systems (2018), https://doi.org/10.1016/j.jmsy.2018.04.009

returns lead to the uncertainty relating to product recovery issues [14].Some previous studies focused on the uncertainties associated withproduct returns in terms of quantity, timing and quality of returns[8,15,16,17]. Akcali and Cetinkaya [15] reviewed previous studies ondeterministic and stochastic product return models for inventory andproduction planning in closed-loop supply chains. Kim and Goyal [18]examined the effect of recovery rate of used products on the profit-ability of closed-loop supply chains under different recovery policies.Aydin et al. [8] studied the uncertainties in the quantity and quality ofused products to determine the optimal product returns for re-manufacturing. Ma and Kim [19] developed a predictive model to im-prove forecasting about future product return quantities and timing ofreturns. Uncertainties in the quantity and quality of product returnshave been studied using a two-stage scenario-based optimization ap-proach [16], and a stochastic optimization model [20], and with a fuzzymixed integer algorithm to optimize the reverse logistics network [21].A mathematical model was developed to explore the effects of adver-tising the return rate of used products on the total profit achieved by theclosed-loop supply chain considering demand and return rate un-certainty [22]. A mathematical model was formulated by Chen et al.[23] to help determine the optimal return rate and new product re-plenishment quantities that would minimize the total cost per productreplenishment cycle.

Some have considered the cost of product returns in closed-loopsupply chain models. The cost of product returns was considered in amulti-echelon reverse logistic network optimization [24]. Companiescan minimize the reverse logistics cost by selecting the optimal locationfor collection points and centralized return centers using a genetic al-gorithm approach. Shi et el. [25] studied optimal production planningfor a closed-loop system to examine the relationship between acquisi-tion price and the stochastic return quantity under uncertain demandand return. Ghoreishi et al. [26] examined the effect of financial in-centives, transportation and advertisement on return rate and the totalcost of used product recovery.

A few studies have used Monte Carlo simulation to deal with un-certainties related to input data and parameters in reverse logistics andclosed-loop supply chain models. Lo et al. [27] studied how to quantifyand reduce uncertainties associated with greenhouse gas emissions andglobal warming potential in life cycle assessment (LCA) using Bayesian-based Monte Carlo simulation method. Hung and Ma [28] later devel-oped a Monte Carlo simulation based methodology to quantify theuncertainties related to life cycle impact assessment and life cycle in-ventory data on municipal waste management. Similarly, Vinodh andRathod [29] applied a Monte Carlo simulation model to investigate theuncertainties in data and parameters of LCA. Diaz and Marsillac [30]used Monte Carlo methods to study the effect of data uncertainty in

supply and demand to help companies make decisions in re-manufacturing supply chain operations. Ewertowska et al. [31] de-monstrated the effect of uncertainties in environmental and eco-effi-ciency analysis on the evaluation of different electricity generationtechnologies.

Although quite a number of previous studies investigated the un-certainties in product return issues, they have not addressed the effectof product return rate uncertainty on the economic and environmentalperformances resulting from a product configuration design. In thisstudy, the impact of product return rate uncertainty on the total life-cycle performance of product configuration designs is quantified byperforming a Monte Carlo simulation-based methodology. Such ananalysis will enable selecting the product configuration design(s) thatare most robust in their ability to meet desired performance goals.

The remainder of this paper is organized as follows. Section 2 de-scribes the proposed methodology for quantifying product return rateuncertainty on the economic and environmental performances of aproduct configuration design. Section 3 presents an industrial casestudy to demonstrate the proposed approach by quantifying the un-certainty of toner cartridge return rates on the total lifecycle perfor-mance of the toner cartridge configuration design. Results and discus-sion of the Monte Carlo simulation are shown in Section 4. Section 5provides the conclusions and future research directions.

2. Proposed methodology

Sustainable product design requires being responsible for the pro-ducts’ entire life from extracting materials to disposal of retired pro-ducts. A closed-loop material flow system considers the total productlifecycle that includes the pre-manufacturing, manufacturing, use, andpost-use stages [2]. Fig. 1 shows the total lifecycle-based closed-loopmaterial flow system considered in this study. The straight-line anddashed-line arrows indicate the forward and reverse flow of materials/products in the supply chain, respectively. In this closed-loop system,EoL and EoU products can be collected and recovered through furtherpost-use activities (i.e., reuse, remanufacturing and recycling) that en-hance overall product sustainability. In this study, component reuse andremanufacturing are considered instead of product reuse and/or re-manufacturing due to the focus on product configuration design.Components which are not reused or remanufactured can be recycledfor material recovery or sold to third-party recyclers to gain revenueand reduce overall environmental impact. There could also be somecomponents and materials that are disposed in the post-use stage.

Designing products considering total lifecycle-based closed-loopmaterial flow is much more complex than the traditional, open-loopmaterial flow based approach which does not require collecting back

Fig. 1. Total lifecycle-based closed-loop material flow system [32].

R. Aydin et al. Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

2

and recovering products at the end of the use stage for use in sub-se-quent lifecycles. The complexities arise because the quantity andquality of product returns are unknown and random.

One approach to analyze the impacts of product return rate un-certainty on the economic and environmental performances of productconfiguration design is to first identify Pareto optimal solutions for thedesign of the product under consideration and then evaluate the ro-bustness of those designs. To follow this approach presented here, weuse the optimal solutions obtained in our previous study [32] whichfocused on identifying the optimal product configuration design (i.e.,specific variants to be used for each component) considering multi-lifecycle material flow (with component reuse and remanufacturing,and material recycling) and several objectives. The multi-objectiveoptimization problem was solved using a non-dominated sorting ge-netic algorithm (NSGA-II). In that study [32], issues related to all life-cycle stages, from extracting raw materials to product EoL recovery(i.e., pre-manufacturing, manufacturing, use, and post-use) were con-sidered; the entire demand cycle and the changes in the quantity ofproduct sold over that period is also considered. The three conflictingobjectives of the first study [32] are minimizing total lifecycle cost,global warming potential, and total water use. And, the objectives ofthe second study [32] are maximizing total lifecycle profit, and mini-mizing total energy use, and total water use. We evaluated the results ofthe product configuration obtained in the first study [32] and in-corporated total energy use functions into the proposed methodology inthis study. Cost is one of the most important objectives when designingproducts. In addition, GWP, water use and energy use are also con-sidered as objectives to assess environmental impacts. Industry haswidely used LCA tools (i.e., Simapro, Ecoinvent or any other software)in order to measure GWP, water use and energy use and support de-cision making in developing sustainable products [33].

In this study, a methodology is proposed to assess the impact of thepotential variations and uncertainties in the product return rate. Fig. 2shows the proposed methodology for quantifying the effect of productreturn rate uncertainty on the economic and environmental perfor-mances of product configuration design using Monte Carlo simulation.The approach uses the following as inputs for the analysis: productconfiguration design; demand estimates; cost, global warming poten-tial, water use, and energy use data throughout the entire lifecycle; andproduct return rate (as a distribution function). Monte Carlo simulationis then conducted to quantify the impact of the product return ratevariation on total lifecycle cost, global warming potential, total wateruse, and total energy use.

Further details of the proposed methodology are described in thefollowing sub-sections.

2.1. Estimation of product return and post-use processing quantities

Product returns are highly dependent on the number of previouslysold products [8,34]. The number of products returned after EoU andEoL at the end of each period can be estimated considering productdemand and the return rate parameter as follows:

= = …+R i D i for t T( ) β( ), 1, 2, 3, , .t u t (1)

= = …R i for t u( ) 0, 1, , .t (2)

where Dt is the product demand in time t, iβ( ) denotes the randomdistribution function for return rate of previously sold products in the i-th simulation run, and +R i( )t u is the number of products returned intime t + u in the i-th simulation run in which u is the use life of theproduct (in use stage). T is the length of the demand cycle. Eq. (2)ensures that there will be no product returns until products completetheir use stage.

After used products are collected, they are disassembled to recoversome components for reuse, remanufacturing or recycling. Componentswhich are not reused or remanufactured are recycled for material

recovery or sold to third-party recyclers to gain revenue and reduceoverall environmental impact. Unrecoverable components are disposedwithout a cost. The number of components reused, remanufactured,recycled, sold, and disposed at the end of each time period can be es-timated using the relative percentage of components processed fol-lowing each of post-use strategies, respectively, at the componentvariant level as shown below:

= +n i δ R i( ) ( )tklpu

tklpu

t u (3)

=for pu reu rmf rcy sol and dis1, 2, 3, 4, 5 ( , , , , )

∑ ==

δ 1pu

tklpu

1

5

(4)

where n i( )tklpu represent the number of the l-th variant of the k-th com-

ponent processed following different post-use strategies (i.e., reused,remanufactured, recycled, sold, and disposed) at time t for the i-th si-mulation run; δtkl

pu denotes the percentage of the l-th variant of the k-th

Fig. 2. Proposed methodology for quantifying uncertainty impacts.

R. Aydin et al. Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

3

component processed following different post-use strategies (i.e., re-used, remanufactured, recycled, sold, and disposed) at time t. Thesepost-use percentages may not necessarily change from on time period tothe other. However, they may change as a product goes through theintroduction, growth, maturity and decline phases of its demand cycle.Eq. (4) ensures that the sum of the post-use percentages is equal to 1.

Products can be (re)manufactured using new, reused, and re-manufactured components, as well as those made from recycled ma-terials. The product return rate directly impacts the number of returnedproducts and also the quantities of components available for reuse,remanufacturing, recycling, and selling. These quantities can be re-presented as shown below.

= = …D D λ for t T, 1, 2, 3, ,tkl t k (5)

= − − −n i D n i n i n i( ) ( ) ( ) ( )tklnew

tkl tklreu

tklrmf

tklrcy (6)

where Dtkl is the demand for the l-th variant of the k-th component attime t; λk is the number of units of component k needed for each pro-duct (as per the bill of materials); n i( )tkl

new , n i( )tklreu , n i( )tkl

rmf , and n i( )tklrcy

represent the quantity of the l-th variant of the k-th component that isnew, reused, remanufactured, and recycled, respectively, at time t forthe i-th simulation run. For each component variant, the quantity ofnew items needed in each period (see Eq. (6)) to satisfy the productdemand can be estimated by subtracting the sum of reused, and re-manufactured component variants, and those made from recycled ma-terials from the total demand for the corresponding component variantsin that period. For any component variant, if there are more reusable,remanufacturable, and recyclable items collected than the demand ofthe corresponding component variant in that time period (i.e., over-collection), excess items are being sold to the third-party recyclers andnot kept in inventory.

2.2. Estimation of economic and environmental performance measures

This study investigates the impact of uncertainty of product returnrate on the economic and environmental performance measures ofproduct configuration design considering the total lifecycle approach.Economic performance is measured by total lifecycle cost, and en-vironmental performance measures are total energy use, globalwarming potential, and total water use.

The total lifecycle cost includes two cost components, fixed cost(c fix) and variable cost. The fixed cost includes assembly cost, setupcost, and all other overhead costs. The variable cost is the cost elementaffected by the selection of component variants for the product designconfiguration, and the usage of new, reused, and/or remanufacturedcomponents, as well as those made from recycled materials and the costof collecting those EoL products. The components which are sold tothird-party recyclers are also considered in the cost computation.Disposal cost is considered negligible and not included in the model.Hence, the total lifecycle cost can be estimated as follows:

∑ ∑ ∑= + +

+ + − + +

= ==

TLCC i c x n i c n i c

n i c n i c n i r D c R i c

( ) ( ( ) ( )

( ) ( ) ( ) ) ( )

fixt

T

k

K

l

L

kl tklnew

klnew

tklreu

klreu

tklrmf

klrmf

tklrcy

klrcy

tklsol

klsol

tuse

t tret

1 11

k

(7)

where TLCC i( ) is the estimated total lifecycle cost in the i-th simulationrun; xkl denotes the parameters for component variants equal to 1 if thel-th variant of the k-th component is selected and 0 otherwise; ckl

new, cklreu,

cklrmf , and ckl

rcy represent the unit costs of the l-th variant of the k-th new,reused, and remanufactured components, and those made from re-cycled materials, respectively; n i( )tkl

sol is the number of components soldfor the l-th variant of the k-th component at time t at the i-th simulationrun; rkl

sol is the unit revenue gained from selling the l-th variant of the k-th component to a third-party company; cuse corresponds to the use cost(per unit) of a product; and ct

ret is the unit cost of collecting used

products in the t-th period.The total energy use includes energy use during pre-manufacturing,

manufacturing, use (if any) and post-use stages of the product, can beestimated as follows:

∑ ∑ ∑= + +

+ +

== =

TEU i x n i e n i e n i e

n i e D e

( ) ( ( ) ( ) ( )

( ) )

t

T

k

K

l

L

kl tklnew

klnew

tklreu

klreu

tklrmf

klrmf

tklrcy

klrcy

tuse

11 1

k

(8)

where TEU i( ) is the estimated total energy use at the i-th simulationrun; ekl

new, eklreu, ekl

rmf , and eklrcy represent energy use of the l-th variant of

the k-th new, reused, and remanufactured components, and those madefrom recycled materials, respectively; and euse is associated with theunit energy consumption during the use stage of a product.

The global warming potential, as considered in this study, accountsfor the greenhouse gas emissions caused by the following activities: theprocessing of virgin materials for the component variants used in theproduct design; the energy used for manufacturing and/or re-manufacturing processes as well as for product assembly and energyconsumed during product use. Since components/materials sold tothird-party recyclers may potentially return to the closed-loop materialflow as recycled materials, it helps offset the overall environmentalimpact (GWP) in the loop. The total number of units of energy con-sumed for the product over all the lifecycles is assessed using TEU. TheGWP provides another perspective of the environmental impact due tothe emissions generated through the various sources of energy usedacross the different lifecycle stages. This can be estimated as follows:

∑ ∑ ∑= + +

− +

== =

GWP i x n i g n i g n i g

n i g D g

( ) ( ( ) ( ) ( )

( ) )

t

T

k

K

l

L

kl tklnew

klnew

tklreu

klreu

tklrmf

klrmf

tklsol

klsol

tuse

11 1

k

(9)

where GWP i( ) is the global warming potential estimated in the i-thsimulation run; gkl

new, gklreu, and gkl

rmf denote the global warming potentialimpacts of the l-th variant of the k-th new, reused, and remanufacturedcomponents, respectively; gkl

sol is the global warming potential impactsaved (per unit) by selling the l-th variant of the k-th component to athird-party company; and guse is associated with the unit globalwarming potential impact during the use stage of a product.

The total water use includes water consumption during pre-manu-facturing, manufacturing, use (if any) and post-use stages of the pro-duct, can be estimated as follows:

∑ ∑ ∑= + +

+ +

== =

TWU i x n i w n i w n i w

n i w D w

( ) ( ( ) ( ) ( )

( ) )

t

T

k

K

l

L

kl tklnew

klnew

tklreu

klreu

tklrmf

klrmf

tklrcy

klrcy

tuse

11 1

k

(10)

where TWU i( ) is the total water use estimated at the i-th simulationrun; wkl

new, wklreu, wkl

rmf , and wklrcy represent water use of the l-th variant of

the k-th new, reused, and remanufactured components, and those madefrom recycled materials, respectively; and wuse is associated with theunit water consumption during the use stage of a product.

3. Industrial case study

To illustrate the applicability and effectiveness of the proposedmethodology, an industrial case study is presented here for a tonercartridge configuration design. The OEM, headquartered in NorthAmerica, manufactures laser printers and cartridges (company namenot disclosed due to confidentiality). The company is planning to de-velop a new laser toner cartridge following a closed-loop material flowapproach by recovering EoL products. However, there is uncertaintyassociated with the product returns which affects the number of reu-sable and remanufacturable components that can be used in manu-facturing the toner cartridge in each time period of the demand cyclefor the configuration design identified through the optimization model.

R. Aydin et al. Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

4

Our case company does not perform in-house recycling; instead all re-cyclable materials are sold to third-part recyclers that may return re-cycled materials to the same material flow.

In the industry case study, we assume that there is no cost incurredwhen the product return rate is below or equal to 0.5; a step-wise in-creasing cost will be incurred to achieve a product return rate higherthan 0.5. In fact, the collection cost has already been considered as areverse logistic cost and integrated into reuse and remanufacturing costand off-set against revenue for materials sold to recyclers. However, thecompany’s current reverse logistic system is able to collect about 50%of previously sold products in each period. As indicated by their en-gineers/managers, additional costs may have to be incurred to increasethe return rate further. Hence the consideration of a step-wise in-creasing cost when return rate is above 0.5. Also, if the quantity ofproduct returns in any period is higher than the product demand in thatperiod (over-collection), we assume the excess quantity is sold by thecompany. Fig. 3shows the ten variable components, as well as theirvariants, and two fixed components of the laser toner cartridge selectedfrom the bill of materials for this study. Inclusion of components withhazardous materials are constrained in the model. The toner hopper,laser unit, all bolts and fasteners have been excluded from the productdesign. These components will be fixed across all product configura-tions. In the figure, Reu, Rmf and Sol refers to whether the specificcomponent variant is reused, remanufactured and sold, respectively.

Table 1 shows the configuration (specific variants selected for eachcomponent) and performance measures for the baseline toner cartridgeand two other toner cartridges: the design with minimum total lifecyclecost and that with the minimum environmental impact, obtained fromthe Pareto optimal solutions.

The energy use, water use and GWP data for all the componentvariants, obtained through LCA modeling (using GABI software), wasprovided by the company. Lifecycle cost data (modified to conceal

competitive information) was also provided by the company. Estimatesfor all other criteria (e.g. demand cycle, percentages of post-use stra-tegies), for which data was not available, were determined in con-sultation with relevant engineers/managers from the company.

4. Results and discussion

In this section, the results of the industrial case study based on theproposed methodology is presented. The Monte Carlo simulation modelwas formulated and coded with the Matlab software. The model wasrun 100 K times and took 359 s. Fig. 4 shows the product return ratesgenerated based on the normal distribution, where the mean andstandard deviation are 0.5 and 0.1, respectively. The return rate variesbetween 0.075 and 0.93, but as can be observed from Fig. 4, the pro-duct return rate seems to vary between 0.4 and 0.6 in the majority ofsituations.

Fig. 5 shows the variation of economic and environmental perfor-mance measures for the minimum cost cartridge design obtained fol-lowing the simulation model. In the charts, the x-axis shows the var-iation in the respective performance measure; the y-axis shows thefrequency of obtaining a particular value for the performance measure.

The average total lifecycle cost for the minimum cost cartridgedesign is $ 9.37M (range $ 8.80M to 11.44M; a variation of -6%to+ 22% from average). Because there are no additional costs of col-lecting items when the product return rate is below or equal to 0.5, themajority of total lifecycle cost scenarios obtained from the simulationfall into the $ 8.80M to $ 9.5M range. The average GWP for theminimum cost cartridge design is 30.67M kg CO2eq (range 22.81M to39.53M kg CO2eq; a variation of -26% to+29% from average). Theaverage water use for the minimum cost cartridge design is 49.84M m3

(range 37.87M to 62.60M m3 ; a variation of -24% to+ 26% fromaverage). Similarly, the average energy use for the minimum cost car-tridge design is 646.7 T J (range 558.8 T J to 764 T J; a variation of-14% to+ 18% from average). Thus, it is evident that there is potentialfor much larger variations in the GWP and water use for the minimumcost cartridge design, than the potential variation for total lifecycle costand energy use, as the product return rate varies. While the step-wiseincreasing cost of higher product return rates gives rise to the somewhatskewed distribution of total lifecycle cost, other performance measures(GWP, energy and water use) are only dependent on the quantity ofproduct returns leading to more symmetrical distributions.

Fig. 6 shows the variation of economic and environmental perfor-mance measures for the minimum environmental impact cartridge de-sign.

The average total lifecycle cost for the environmental impact designis $ 9.93M (range $ 9.61M to 11.53M; a variation of -3% to+16%from average). The majority of total lifecycle cost scenarios obtained

Fig. 3. Variable and fixed components of laser toner cartridges and their var-iants [Note: PC refers to photoconductor].

Table 1Comparison of component variants and performance of toner cartridges.

Toner cartridge design Baseline Min cost Min env. impact

Toner Housing PC-ABS PC-ABS HIPSDeveloper Roll Foam-Metal Urethane-Metal Foam-MetalDoctor Blade Flex Spring SpringToner Paddle Mixed Mixed MixedToner Bushings Metal Plastic PlasticAuger POM POM SteelWaste Toner Housing PC-ABS PC-ABS HIPSPC Drum Diameter 28mm 20mm 20mmPC Drum Bushings Metal Plastic PlasticCharge Roll Contact Contact ContactPerformance measures Baseline Min cost Min env. impactTotal Lifecycle Cost ($) M 10.22 9.28 9.84GWP (CO2eq) M 31.11 30.67 28.28Total Water Use (m^3) M 61.67 49.84 47.56Total Energy Use (MJ) M 650.37 646.52 634.65

R. Aydin et al. Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

5

from the simulation fall into the $ 9.61M to $ 10M range. The averageGWP for the minimum cost cartridge design is 28.28M kg CO2eq (range21.50M to 35.71M kg CO2eq; a variation of −24% to+26% fromaverage). The average water use for the minimum cost cartridge designis 47.56M m3 (range 36.44M to 59.17M m3 ; a variation of −23%

to+ 24% from average). Similarly, the average energy use for theminimum cost cartridge design is 634.8 TJ (range 566.5 TJ to 727.4 TJ;a variation of -11% to+15% from average). Thus, it is evident thatthere is potential for much larger variations in the GWP and water usefor the minimum environmental impact cartridge design, than the

Fig. 4. Product return rate distribution.

Fig. 5. Variation of performance measures for minimum cost cartridge design.

R. Aydin et al. Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

6

potential variation for total lifecycle cost and energy use, as the productreturn rate varies. From Figs. 5 and 6, it can be observed that thevariation of performance measures for the minimum environmentalimpact cartridge design is less than that for the minimum cost cartridge

design.Fig. 7 shows the economic and environmental performance varia-

tion for different cartridge designs as the product return rate changes.These performance results also reinforce the observations shown in

Fig. 6. Variation of performance measures for minimum environmental impact cartridge design.

Fig. 7. Comparison of alternate cartridge performance measure variation with return rate.

R. Aydin et al. Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

7

Table 1 for the relative performance of the minimum cost and minimumenvironmental impact toner cartridge designs, in comparison to thebaseline design. Both the minimum cost and minimum environmentalimpact designs perform better than the baseline toner cartridge designwith respect to all measures. As can be observed, the total life cycle costfor the cartridge designs decreases with an increase in the product re-turn rate until it is around 0.75. The minimum total cost can beachieved when the return rate is between 0.7 and 0.8. However, thetotal cost begins to increase when the product return rate is around 0.8.Thus, it is evident from the total lifecycle cost variation that there existsan optimal value (or range) for the product return rate that will helpminimize the cost of any toner cartridge configuration design.

As can also be observed from the charts in Fig. 7, the energy use,water use and GWP all decrease as the product return rate increases.However, given there is an optimal value (range) for the product returnrate that enables achieving the lowest cost, the case company wouldbenefit most by implementing strategies to maintain the product returnrate within that optimal range.

As shown in the total lifecycle cost chart, there will be an increase inthe cost as the product return rate increases. This is partly due to havingto spend more to collect the EoL products (step-wise cost structurepresented earlier). Another factor that also leads to a lower marginalbenefit from the increase in product return rate is likely due to over-collection. That is, the situations where the product return quantity isgreater than the product demand. In such cases, though expenses areincurred in collecting the EoL products, the company will not be able touse the components from those products to off-set costs. This is anotherreason for observing less marginal cost savings as the product returnrate increases. Other impacts of over-collection can also be seen in thetotal energy use and GWP variations. As the product return rate in-creases, the total energy use decreases due to savings offered by re-using/remanufacturing components from those products. However, asproduct return rate increases and results in over-collection, no furthersavings in energy consumption are possible. Thus, energy consumptionreaches a minimum value as product return rate increases and levels offas it results in over-collection of items which are sold off. The GWPcomputation shows a slightly different pattern of variation as the pro-duct return rate increases. This is because GWP savings from the over-collected items that are sold to other parties is off-set against the totalGWP consumption for the case company’s product. Hence the con-tinuous decline of GWP, though at lower rate, as the product return rateincreases.

5. Conclusions

In this study, we examined the effect of uncertainty and variation inproduct return rate on the economic and environmental performancesresulting from product configuration designs using a Monte Carlo si-mulation-based methodology. The proposed methodology is im-plemented on an industrial case study to demonstrate the applicabilityand effectiveness of the proposed approach. The results show that theuncertainty of product return rate would cause wide variations ineconomic and environmental performance measures for two cartridgedesigns with minimum cost and minimum environmental impact, re-spectively.

However, this study has some assumptions and limitations. First,demands for products were known throughout the entire lifecycle. Theuncertainty in the demand for remanufactured products was not ex-amined. A dynamic demand model could be integrated into the pro-posed model while considering the uncertainty in the demand. Second,the quality of product returns which involves high uncertainty was notinvestigated in this study. The proposed methodology can be extendedconsidering the uncertainty in the quality of product returns that mayhave an impact on both economic and environmental performances ofproducts. Third, the return rate is considered constant throughout theentire demand cycle. The future work could involve modeling varying

return rates in different phases of the demand cycle. It is also possible tofurther assess the robustness of alternate product designs by scenarioanalysis, for example, considering different cost structures to imple-ment product return strategies.

Acknowledgement

This work was supported by the Digital Manufacturing and DesignInnovation Institute (DMDII) [grant number 15-05-08].

References

[1] Badurdeen F, Jawahir IS. Strategies for value creation through sustainable manu-facturing. Procedia Manuf 2017;8:20–7.

[2] Jawahir IS, Dillon OW, Rouch KE, Joshi KJ, Venkatachalam A, Jaafar IH. Total life-cycle considerations in product design for sustainability: a framework for compre-hensive evaluation Barcelona, Spain. Proceedings of the 10th InternationalResearch/Expert Conference 2006:1–10.

[3] Chang D, Lee CKM, Chen CH. Review of life cycle assessment towards sustainableproduct development. J Clean Prod 2014;83:48–60.

[4] Hatcher GD, Ijomah WL, Windmill JFC. Integrating design for remanufacture intothe design process: the operational factors. J Clean Prod 2013;39:200–8.

[5] Aydin R, Kwong CK, Ji P. Coordination of the closed-loop supply chain for productline design with consideration of remanufactured products. J Clean Prod2016;114:286–98.

[6] Go TF, Wahab DA, Hishamuddin H. Multiple generation life-cycles for productsustainability: the way forward. J Clean Prod 2015;95:16–29.

[7] Kalverkamp M, Pehlken A, Wuest T. Cascade use and the management of productlifecycles. Sustainability 2017;9(9).

[8] Aydin R, Kwong CK, Geda MW, Okudan Kremer GE. Determining the optimalquantity and quality levels of used product returns for remanufacturing under multi-period and uncertain quality of returns. Int J Adv Manuf Technol 2017:1–14.

[9] Hatcher GD, Ijomah WL, Windmill JFC. Design for remanufacture: a literature re-view and future research needs. J Clean Prod 2011;19(17–18):2004–14.

[10] Mangun D, Thurston DL. Incorporating component reuse, remanufacture, and re-cycle into product portfolio design. IEEE Trans Eng Manage 2002;49(4):479–90.

[11] Jun H-B, Cusin M, Kiritsis D, Xirouchakis P. A multi-objective evolutionary algo-rithm for EOL product recovery optimization: turbocharger case study. Int J ProdRes 2007;45(18–19):4573–94.

[12] Kwak M, Kim HM. Assessing product family design from an end-of-life perspective.Eng Optim 2011;43(3):233–55.

[13] Kwak M, Kim HM. Design for life-cycle profit with simultaneous consideration ofinitial manufacturing and end-of-life remanufacturing. Eng Optim2015;47(1):18–35.

[14] Govindan K, Soleimani H, Kannan D. Reverse logistics and closed-loop supplychain: a comprehensive review to explore the future. Eur J Operat Res2015;240(3):603–26.

[15] Akcali E, Cetinkaya S. Quantitative models for inventory and production planningin closed-loop supply chains. Int J Prod Res 2011;49(8):2373–407.

[16] Zeballos LJ, Gomes MI, Barbosa-Povoa AP, Novais AQ. Addressing the uncertainquality and quantity of returns in closed-loop supply chains. Comput Chem Eng2012;47:237–47.

[17] Denizel M, Ferguson M, Souza GGC. Multiperiod remanufacturing planning withuncertain quality of inputs. IEEE Trans Eng Manage 2010;57(3):394–404.

[18] Kim T, Goyal SK. Determination of the optimal production policy and product re-covery policy: the impacts of sales margin of recovered product. Int J Prod Res2011;49(9):2535–50.

[19] Ma J, Kim HM. Predictive model selection for forecasting product returns. J MechDes 2016;138(5).

[20] Mashhadi AR, Esmaeilian B, Behdad S. Uncertainty management in re-manufacturing decisions: a consideration of uncertainties in market demand,quantity, and quality of returns. J Risk Uncertainty Eng Syst B 2015;1(2).

[21] Niknejad A, Petrovic D. Optimisation of integrated reverse logistics network withdifferent product recovery routes. Eur J Operat Res 2014;238:143–54.

[22] Jena SK, Sarmah SP, Sarin SC. Joint-advertising for collection of returned productsin a closed-loop supply chain under uncertain environment. Comput Ind Eng2017;113:305–22.

[23] Chen C-K, Weng T-C, Lo C-C. Optimal replenishment quantity for new products andreturn rate of used products for a retailer. Appl Math Model 2016;40(23):9754–66.

[24] Min H, Jeung Ko H, Seong Ko C. A genetic algorithm approach to developing themulti-echelon reverse logistics network for product returns. Omega2006;34(1):56–69.

[25] Shi J, Zhang G, Sha J. Optimal production planning for a multi-product closed loopsystem with uncertain demand and return. Comput Oper Res 2011;38(3):641–50.

[26] Ghoreishi N, Jakiela MJ, Nekouzadeh A. A cost model for optimizing the take backphase of used product recovery. J Remanuf 2011;1(1).

[27] Lo S-C, Ma H, Lo S-L. Quantifying and reducing uncertainty in life cycle assessmentusing the Bayesian Monte Carlo method. Sci Total Environ 2005;340:23–33.

[28] Hung M-L, Ma H. Quantifying system uncertainty of life cycle assessment based onMonte Carlo simulation. Int J Life Cycle Assess 2009;14:19–27.

[29] Vinodh S, Rathod G. Application of life cycle assessment and Monte Carlo simula-tion for enabling sustainable product design. J Eng Des Technol

R. Aydin et al. Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

8

2014;12(3):307–15.[30] Diaz R, Marsillac E. Evaluating strategic remanufacturing supply chain decisions.

Int J Prod Res 2017;55(9):2522–39.[31] Ewertowska A, Pozo C, Gavaldà J, Jiménez L, Guillén-Gosálbez G. Combined use of

life cycle assessment, data envelopment analysis and Monte Carlo simulation forquantifying environmental efficiencies under uncertainty. J Clean Prod2017;166:771–83.

[32] Badurdeen F, Srivastava S, Saunders C. Predictive modeling for digitally-enabled,

multi- criteria decision making for innovative product design & analysis for totallifecycle sustainability. DMDII-15-05-08 Quarterly Technical Review Report 4.2017.

[33] Lulietti R, Rosa P, Terzi S, Taisch M. Life cycle assessment tool in product devel-opment: environmental requirements in decision making process. Procedia CIRP2016;40:202–8.

[34] Toktay LB, Wein LM, Zenios SA. Inventory management of remanufacturable pro-ducts. Manage Sci 2000;46(11):1412–26.

R. Aydin et al. Journal of Manufacturing Systems xxx (xxxx) xxx–xxx

9