Investigating the environmental and economic impact of loading conditions and repositioning strategies for pallet pooling providers

Journal of Cleaner Production 172 (2018) 155e168

Contents lists available at ScienceDirect

Journal of Cleaner Production journal homepage: www.elsevier.com/locate/jclepro

Investigating the environmental and economic impact of loading conditions and repositioning strategies for pallet pooling providers Fabiana Tornese a, Jennifer A. Pazour b, *, Brian K. Thorn c, Debjit Roy d, Andres L. Carrano e a

University of Salento, Department of Innovation Engineering, Campus Ecotekne, Via per Monteroni, 73100 Lecce, Italy Rensselaer Polytechnic Institute, Department of Industrial and Systems Engineering, 110 8th Street, Troy, NY 12180, USA c Rochester Institute of Technology, Department of Industrial and Systems Engineering, 81 Lomb Memorial Drive, Rochester, NY 14623, USA d Indian Institute of Management Ahmedabad, Department of Production and Quantitative Methods, Vastrapur, Ahmedabad, India e Georgia Southern University, Department of Manufacturing Engineering, PO Box 7995, Statesboro, GA 30460-7995, USA b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 17 January 2017 Received in revised form 27 June 2017 Accepted 6 October 2017 Available online 12 October 2017

Pallets are fundamental assets critical to worldwide supply chain logistics. This research develops models for closed-loop pallet pooling providers to understand the environmental and economic impact of customer characteristics and design options. First, an analytical model is developed to quantify the effects of repair facility location and pallet service conditions on a pallet pooling system's economic and environmental performance. Next, a simulation model is developed to investigate two common operational policies, crossdocking and take-back, and to quantify the impact of pallet handling and loading conditions and customer network structures on several key performance indicators. Results indicate that pallet handling and loading conditions are the most important factors determining the cost and carbon equivalent emission of a pallet pooling operation. Better pallet handling and appropriate loading increase the percentage of pallets that can be repositioned with little or no repair. This increases the radius within which a closed-loop pallet pooling system is feasible. Under random handling/loading conditions and distances, a crossdocking approach satisfies demand with 28% fewer pallets than a take-back policy. This is due to a quicker reissue time under a crossdocking approach. However, associated costs and emissions of the two policies are nearly identical due to the increased transportation costs associated with crossdocking. The models and insights proposed in this work can help support decision making by pallet pooling providers to determine operational regions and customer selection, among other network design trade-offs. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Pallet pooling Repositioning Carbon emissions Simulation Remanufacturing Closed-loop logistics

1. Introduction Pallets are the most common platform for unit load formation, which enable seamless and efficient transportation of goods in supply chains. The importance of pallets and the extent to which supply chains rely on them are often underestimated: approximately 80% of the United States (US) trade is carried on pallets (Raballand and Aldaz-Carroll, 2005) and more than 2 billion pallets are in circulation in the US (Buehlmann et al., 2009). The European Union boasts more than 280 million pallets in circulation every year (Raballand and Aldaz-Carroll, 2005). The wood pallet and container industry is vast, complex and geographically dispersed. In the US,

* Corresponding author. E-mail address: [email protected] (J.A. Pazour). https://doi.org/10.1016/j.jclepro.2017.10.054 0959-6526/© 2017 Elsevier Ltd. All rights reserved.

this industry, comprised of more than 2600 establishments, accounted for $7 billion dollars in estimated receipts in 2012 (NAICS, 2012). In spite of this, pallet logistics have not been extensively studied and the available scientific literature is limited. Given that 99% of the establishments engaged in pallet logistics are small businesses (The United States Census Bureau, 2016; Millwood Inc. 2015), industry data is generally not available. Because of this demographic, the bandwidth and resources to embark in these studies and models within the industry is almost non-existent. Closed-loop supply chains have been defined in literature as “a special type of supply chains that consider the return flow of used materials in addition to the downstream flow of products” (Glock, 2017). Specifically, in pallet management, closed-loop systems allow the collection of used pallets at the end-point of the supply chain, for reuse, repair or recycling, as opposed to open-loop schemes, where the pallet remains at the final customer (Elia and

156

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

Gnoni, 2015). A growing closed-loop model for pallet management is pallet pooling, where a service company (the pallet pooling provider) owns the pallets and manages their flows (Roy et al., 2016). Customers of the pallet pooling provider subscribe to access a pool of pallets to transport their goods. Under a pallet pooling scenario, pallets are generally collected after use at a downstream location in the supply chain (e.g. at the regional distribution center (DC) or retailer) by the pooling company or a network of regional pallet recyclers working in coordination with the pooling provider (Mazeika Bilbao et al., 2010). These collected pallets are processed and assimilated into the pool of pallet assets owned by the pooling company and further repositioned wherever needed (likely at an upstream location in the supply chain). The conditions of the pallets for immediate reuse may be inspected on-site at the docks during retrieval or after the pallets are transported to the recycler's repair depot. If the assessment of the pallet conditions is performed onsite, those pallets deemed in good condition for reuse can be immediately repositioned at a different facility while only pallets needing repair are transported to the repair facility for remanufacturing; a practice referred to as “crossdocking”. The alternate practice, referred to as “take back”, is to collect and all pallets from a customer's end point regardless of their condition and transport them to a repair facility for inspection, sortation, repair (if needed) and repositioning back into service. Pallet pooling providers offer supply, management and tracking of pallet assets and have emerged as an alternative to companies who prefer outsourcing pallet management tasks to a third party. Pallet pooling providers operate a closed-loop system requiring decisions about how to best handle the reverse logistics required to backhaul pallets and pre-position them, as well as manage pallet repair and disposal activities. Previous research efforts have addressed decision-making aimed at improving pallet management performance, but consideration of a pallet pooling provider remains unexplored. The aim of this work is to investigate design and operational decisions from a pooling provider perspective, specifically repair depot locations, pallet loading and handling conditions, and pallet repositioning policies. Adopting both economic and environmental sustainability perspectives, the key performance indicators (KPIs) are economic costs and carbon equivalent (CO2-eq) emissions associated with transportation and repair in a pallet pooling system. An analytical model focuses on used pallet collection, repair and repositioning, while a simulation model considers a broader supply chain perspective. In particular, take-back and crossdocking repositioning strategies are analyzed considering the impact of operational factors, such as pallet handling, loading and repositioning distances, being stochastic. This paper is organized as follows. Section 2 contrasts our work with existing studies in pallet management literature, motivating the need to investigate pallet handling/loading conditions and repositioning strategies. Section 3 describes our research methodology and data sources. In Sections 4 and 5, the analytical and simulation models are presented and used to generate new insights related to closed-loop pallet pooling performance. Conclusions are presented in Section 6. 2. Literature review and scope of the work 2.1. Background on reverse logistics and closed-loops supply chains Reverse logistics and closed-loop supply chain network design decisions have been studied from environmental, legal, social, and economic perspectives (see Govindan et al., 2015 for a comprehensive review). In particular, several studies focus on optimal location of remanufacturing and inspection facilities. For example, Alshamsi and Diabat (2015) elucidates the optimal selection of

sites, the capacities of inspection centers and remanufacturing facilities, and selects between in-house vs. outsourced transportation. Srivastava (2008) develops a conceptual model for simultaneous locationeallocation of facilities for a cost effective reverse logistics network. Others investigate the impact of product take-back recovery rate on the logistic network design decisions. For example, Fleischmann et al. (2001) investigate the impact of different product recovery rates on forward logistics network design on copier manufacturing and paper recycling. They show that product recovery can in many cases be implemented without many changes in existing forward production-distribution networks using a facility location model. Other logistics network design methods were also developed in the context of empty container repositioning in shipping liner networks and vehicle repositioning for car rental companies (For example, see Shintani et al., 2007 and Roy et al., 2014). While these studies attempt to minimize the overall logistics cost, they do not incorporate asset repair nor consider environmental impacts. 2.2. Literature on pallet management Scientific literature focused on pallet management strategies and pallet supply chains is not vast, although interest has been growing over the last decade. Table 1 summarizes pallet management literature identifying four main research areas: (i) economic/ environmental evaluation of pallet management strategies, (ii) environmental analysis of pallet operations, (iii) closed-loop pallet supply chain modeling and (iv) traceability in pallet management. These areas are contrasted with respect to the modeling approach, stakeholder perspective, KPI, decision level, problem addressed, method employed, and whether repositioning strategies or handling conditions are captured in the research. Several authors focus on the economic and/or environmental analysis of different pallet management strategies/operations. Most of the works in this category have a specific environmental perspective. Gasol et al. (2008) perform a Life Cycle Assessment (LCA) to compare the environmental performance of different reuse intensity policies, pointing out the role that maintenance plays in reducing the overall impacts. Bengtsson and Logie (2015) perform an LCA to assess one-way and pooled pallet alternatives and demonstrate that pooled softwood pallets outperform all other alternatives. Carrano et al. (2015) explore different pallet management strategies, and provide an optimization model for minimizing emissions under several handling, loading, and EoL scenarios. Tornese et al. (2016) examine the pallet remanufacturing phase in great detail, performing a carbon footprint analysis to support decision making in this phase. This work highlights the importance of distance and handling/loading conditions on the environmental impacts of remanufacturing operations. Two works focus only on the economic dimension of pallet management: Roy et al. (2016) develop cost relationship models to compare open and closedloop pallet management schemes from a user perspective, while Ray et al. (2006) compare the financial outcomes of rental and purchased pallets through simulation modeling, showing that renting can be more expensive. Finally, one work attempts to consider both perspectives in the evaluation of pallet management strategies: Mazeika Bilbao et al. (2011) model the environmental impacts of pallet management operations by developing a linear minimum cost network flow model to support decision making. Research efforts focusing on the environmental analysis of pallet operations include: Bhattacharjya and Kleine-Moellhoff (2013) present an overview of sustainability issues in a pallet lifecycle, identifying the key stakeholder challenges. Carrano et al. (2014) analyze the carbon footprint of pallet operations for each phase of a pallet lifecycle, from raw materials to end-of-life (EoL),

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

157

Table 1 Characteristics of main literature about pallet management. Area Article

Models

(i) (i)

Perspective KPI

Level

Problem

Method

Repositioning Handling conditions

Closed-loop General Open & closed-loop General

Environmental Tactical Environmental Strategic

Reuse intensity Open vs closed-loop

LCA LCA

No No

No No

(i)

Gasol et al., 2008 Bengtsson and Logie, 2015 Carrano et al., 2015

Open & closed-loop General

Environmental Strategic

Yes

Tornese et al., 2016

Closed-loop

Environmental Tactical

Yes

Yes

(i) (i)

Ray et al., 2006 Roy et al., 2016

Open & closed loop Customer Open & closed-loop Customer

No No

No No

(i)

Mazeika Bilbao et al., Closed-loop 2011

Linear programming Linear programming Simulation Stochastic model Linear programming

No

(i)

No

No

(ii)

e

No

No

(ii)

Bhattacharjya and Kleine-Moellhoff, 2013 Carrano et al., 2014

Pallet management strategy selection Environmental impact of pallet remanufacturing Open vs closed-loop Costs of pallet management strategies Pallet management operations environmental impact Pallet management environmental issues

Carbon Footprint Carbon Footprint LCA

No

Yes

(ii)

Pallet management environmental issues Pallet design alternatives

No

No

No

No

Simulation

No

No

Simulation

No

No

Activity No Based Costing e No Analytical and Yes simulation

No

General

Economic Economic

Strategic Strategic

Customer

Environmental Strategic and economic

e

General

Environmental e

Closed-loop

General

Environmental e

Ng et al., 2014

e

General

Environmental Strategic

~ ona García-Duran et al., 2016 (iii) Bottani et al., 2015

e

General

Environmental Strategic

Closed-loop

Customer

(iii) Elia and Gnoni, 2015

Closed-loop

Customer

Economic and strategic Economic and strategic Economic

(ii)

(iv)

Gnoni and Rollo, 2010 Closed-loop

(iv) (i)

Gnimpieba et al., 2015 Closed-loop Our work Closed-loop

Logistic provider e Pallet pooling provider

e Economic and environmental

Environmental impact of pallet manufacturing Tactical Cost optimization in RTI management Tactical Pallet interchange scenarios in closed-loop supply chain Strategic Efficiency of RFID for pallet tracking Strategic Pallet tracking Strategic, tactical Closed-loop network design & and operational Customer Selection

considering different handling, loading, and EoL scenarios. Ng et al. (2014) use carbon footprint to compare the use of virgin softwood and recycled wood pallets, showing the environmental benefits of ~ ona et al. (2016) analyze the the recycled product. García-Duran supply chain for sawn wood in pallet production and identifies electricity for heat treatment and the use of steel for nails as the largest contributors to the environmental impact. Tornese et al. (2016) evaluates carbon footprint of the remanufacturing phase in a closed-loop scheme. Two previous efforts contribute to closed-loop pallet supply chain modeling in pallet management. Bottani et al. (2015) propose a multi-objective optimization model for a closed-loop supply chain consisting of a pallet provider, a manufacturer and seven retailers. The model explores economic and strategic performance under three scenarios with different operating conditions. Elia and Gnoni (2015) outline the critical factors and KPIs in a closed-loop pallet management system, and develop a simulation model with different pallet interchange scenarios. Finally, implementation of traceability tools in pallet management have been addressed by Gnoni and Rollo (2010), who investigated the use of RFID for plastic and wood pallet tracking, and by Gnimpieba et al. (2015), who propose a collaborative platform based on RFID and Internet-ofThings technologies to manage communication, tracking and data sharing in pallet management.

No Yes

documented in archival literature (Gasol et al., 2008; Bengtsson and Logie, 2015; Carrano et al., 2015; Roy et al., 2016) but remain largely unexplored. Moreover, as evident from Table 1, the impact of operational factors, such as the pallet handling/loading conditions or the repositioning strategy adopted, although recognized as significant, have not been systematically studied. In fact, only one work (Tornese et al., 2016) considers different repositioning strategies, but it focuses exclusively on the evaluation of carbon footprint of the remanufacturing phase in a closed-loop scheme. A few studies consider handling conditions (Carrano et al., 2014, 2015; Tornese et al., 2016), but focus their impact on the environmental performance (without considering most economic aspects). None consider a pallet pooling provider perspective. Studies are needed to (1) integrate environmental and economic sustainability measures, (2) explore uncertainties in closed-loop operations, and (3) integrate strategic, tactical, and operational decisions. Starting from these considerations, the research questions addressed in this work are: RQ1. For a given set of pallet pick-up and repositioning locations, what is the maximum average distance at which a repair facility can be located such that a closed-loop pallet system is both economically and environmentally sustainable? RQ2. What is the impact of customer handling/loading conditions, network distances, and alternative repositioning strategies on economic and environmental KPIs of a pallet pooling system?

2.3. Research questions and contributions of this work 3. Methodology While research to improve pallet management exists, the incorporation of modern pallet pooling management approaches has yet to be thoroughly studied. In particular, the advantages with pallet pooling systems when compared to open-loop schemes, are

The research questions are addressed with a two-pronged methodology (Fig. 1). The key measures are costs and CO2-eq emissions of collecting, remanufacturing, and repositioning pallets.

158

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

Fig. 1. Research objectives and two-pronged methodology.

conditions and distances commonly found in pallet supply chains (Glock, 2017). Data on pallet production and management were acquired through direct observations and time studies performed at 12 facilities involved in the various aspects of pallet manufacturing, repair, and disposal operations, including lumber mills, pallet manufacturers, pallet recyclers, and pallet-pooling providers (both third-party-owned and third party-managed pooling companies) located in the US. During these visits, a complete documentation of their standard practices was conducted. This study only considered data on 48 by 40 inch, reusable block pallets made of solid wood, which represent 90 to 95 percent of the pallets worldwide (Buehlmann et al., 2009; Mead, 2010). The scope of the study includes all economic costs and carbon equivalent (CO2-eq) emissions associated with transportation and repair in a pallet pooling system (Fig. 2). Data on pallet service life (Table 2), damage rate and repair were obtained from the Pallet Design System (PDS®) software as well as from Tornese et al., (2016). The sources elucidate the longevity of the pallets based on the handling and loading conditions. Three conditions are considered: (i) good handling and treatment, lightduty loads (1000 lbs), (ii) average handling and treatment, medium-duty loads (2000 lbs), and (iii) rough handling and treatment, heavy-duty loads (3000 lbs), which we refer to as good, average and rough handling/loading, respectively. Table 2 summarizes the state of a pallet after being subjected to the service conditions in a facility. The carbon emission coefficients (Table 3) and cost data (Table 4) were obtained from literature or estimated through direct observations. Carbon emission coefficients are based on a national weighted average conversion of the US electricity grid of 0.648 kgCO2-eq/kWh (US EIA, 2011). The carbon equivalent

Fig. 2. Closed loop model and system boundaries.

An analytical model is built to analyze RQ1, which compares KPIs in a closed-loop system to those of manufacturing new pallets in an open-loop system. RQ2 is explored through an agent-based discrete event simulation model, which represents a generic set of customers (product supply chains) and a pallet pooling provider (with a pallet manufacturing and a pallet repair facility). While the analytical model explores the limits of a closed-loop solution, the simulation model evaluates take-back and crossdocking repositioning strategies by incorporating uncertainty in handling/loading

Table 2 State probabilities of a pallet after a use cycle (Tornese et al., 2016). Probability of pallet state immediately after use Handling and loading conditions in current use cycle

Reusable without repair

In need of repair (or not reusable)

Good Average Rough

50% 30% 10%

50% 70% 90%

Table 3 Emission factors and coefficient for pallet repair, transportation and manufacturing. Phase

Emission factor

Unit

Value

Source

Transportation (32t EURO5 truck) Transportation (16t EURO5 truck) Repair Manufacturing

Ctr32 Ctr16 Crep Cman

kgCO2/(kg*km) kgCO2/(kg*km) kgCO2/pallet kgCO2/pallet

0.107*10
3 0.168*10
3 0.351 3.870

(Swiss Centre for Life Cycle Inventories, 2009) (Swiss Centre for Life Cycle Inventories, 2009) Derived from (Tornese et al., 2016) (Carrano et al., 2014)

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

159

Table 4 Unitary costs for transportation, repair and manufacturing. Phase

Cost factor

Unit

Value

Source

Transportation Repair Manufacturing

ktr krep kman

$/(full truckload * km) $/pallet $/pallet

1.05 1 3

(Ford Torrey and Murray, 2014) (“Getting Paid What You're Worth” 2015), direct observations (Stephens Baker, 2011), direct observations

emissions for pallet manufacturing include materials, assembly and heat treatment while the carbon equivalent emissions for pallet repair include materials and operations.

Table 3 shows the emissions resulting from transportation, repair and manufacturing of a 48  40 wooden block pallet. The transportation emission per unit distance traveled are modeled with the GaBi software modeling principles (“Documentation for Truck Transport Process” 2016).

4. Analytical model


CtrðloadÞ ¼ CtrðemptyÞ þ CtrðfullÞ
CtrðemptyÞ

The analytical model compares a closed-loop pooling system and an open-loop single use pallet system. For the pallet pooling system, economic costs and CO2-eq emissions are captured for pallet transportation (from a product supply chain end point to the repair facility), repair activities and repositioning (from the repair facility to a product supply chain start point). All other lifecycle phases (materials, use, and EoL) are not included in the analysis. The following factors are incorporated into the model:  N ¼ total number of pallets collected at a supply chain end point. We assume the service provider collects a full load of pallets from the end point (or ships a full load of repaired pallets to the start point). We consider an average capacity of 480 pallets (32t truck) and 180 pallets (16t truck) as derived from direct observations.  The physical condition of each of the pallets collected from the supply chain's end point can be: (i) in good condition, thus immediately reusable, (ii) in need of some repair, or (iii) too damaged to be repaired, thus dismantled and disposed. We denote the percent of pallets collected that are in good condition as g, the percentage that needs to be repaired as r, and the percentage that is too badly damaged as b. All pallets collected must be in one of these three conditions, thus, r þ g þ b ¼ 1.  The customers' supply chain networks are captured by considering start and end points. One advantage of pallet pooling is that the end and start points do not have to belong to the same customer's supply chain. Specifically, in the model, d1 denotes the average distance [km] from an end point in the supply chain to the repair facility; d2 represents the average distance [km] from the repair facility to a start point in the supply chain. Also, D represents the average distance [km] from the pallet manufacturing facility to a start point in the supply chain.

cargo loadcapacity

(1)

where, Ctr(empty) denotes the emission per km traveled of an empty truck, obtained by multiplying the Ctr factor by the tare truck weight [kgCO2/km]; Ctr(full) is the emissions per km traveled of a truck loaded at its maximum capacity, obtained by multiplying the Ctr factor by the laden (truck þ load) weight [kgCO2/km]; w denotes the average weight of one pallet, assumed at 31.58 kg (Tornese et al., 2016); load capacity represents the truck load capacity [kg]; cargo is the total weight transported [kg]

cargo ¼ Nw

(2)

For the purposes of this work, a 32t EURO5 diesel truck has a load capacity of 24.7 t and vehicle weight of 7.3 t. Further, a 16t truck has a load capacity of 11.4 t and vehicle weight of 4.6 t (“Documentation for Truck Transport Process” 2016). For transportation to both a repair facility and pallet repositioning locations, the truck has one empty and one loaded run. Thus, CO2-eq emissions of transportation (TE) for the closed loop is:

  TE ¼ ðd1 þ d2 Þ CtrðemptyÞ þ CtrðloadÞ

(3)

The CO2-eq emissions related to collection and repair in a closed loop include the repair of the fraction r of the N pallets collected that can be repaired. In an open-loop, CO2-eq emissions are caused by the production of a number of pallets equal to those that could be collected and recovered in a closed-loop and transported to a supply chain start point. The corresponding emission per km Ctr(newPallets) is calculated through equation (1), considering the load of N(gþr) pallets shipped. Thus, the inequality defining the maximum distance making a pallet pooling system with repair more environmentally efficient than an open system is:

The analytical model assumes information about the network topology and truck type capacities being known and given.

  ðd1 þ d2 Þ CtrðemptyÞ þ CtrðloadÞ þ NðrÞCrep    Nðr þ gÞCman þ D CtrðemptyÞ þ CtrðnewPalletsÞ

4.1. Emissions function

d1 þ d2 

(4)

  Nðr þ gÞCman þ D CtrðemptyÞ þ CtrðnewPalletsÞ
NðrÞCrep   CtrðemptyÞ þ CtrðloadÞ

The following carbon equivalent emissions are compared for an open and closed-loop policy: (i) The emissions in a closed-loop pooling system, derived from transporting N pallets from an end point of a customer's supply chain to a repair facility, from repairing those that can be recovered, and from repositioning them to a start point of a customer's supply chain; (ii) The emissions in an open-loop system, derived from manufacturing new replacement pallets and shipping them to the customer.



(5)

4.2. Cost function For the pallet pooling provider, asset recovering is economically feasible only if the total cost of collection, repair and repositioning is lower than the manufacturing and transportation costs of new pallets. Most logistics providers charge a fixed fee per km traveled. This somewhat ignores the variability of costs associated with the

160

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

Table 5 Maximum total repositioning distance (in km) as a function of CO2-eq emissions (16t and a 32t truck). (d1 þ d2) max 16t truck

32t truck

r

g ¼ 0.5

g ¼ 0.3

g ¼ 0.1

g ¼ 0.5

g ¼ 0.3

g ¼ 0.1

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

341.55 376.43 411.31 446.19 481.06 515.94

266.74 301.62 336.50 371.38 406.25 441.13 476.01 510.89

191.93 226.81 261.69 296.56 331.44 366.32 401.20 436.08 470.96 505.84

478.02 543.80 609.59 675.37 741.15 806.93

335.88 401.66 467.44 533.22 599.00 664.79 730.57 796.35

193.73 259.51 325.29 391.07 456.86 522.64 588.42 654.20 719.99 785.77

Table 6 Maximum total repositioning distance (in km) as a function of cost, (16t and 32t truck). (d1 þ d2) max 16t truck

32t truck

r

g ¼ 0.5

g ¼ 0.3

g ¼ 0.1

g ¼ 0.5

g ¼ 0.3

g ¼ 0.1

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

378.57 395.71 412.85 430.00 447.14 464.28

327.14 344.29 361.43 378.57 395.71 412.86 430.00 447.14

275.71 292.86 310.00 327.14 344.29 361.43 378.57 395.71 412.86 430.00

592.86 638.57 684.29 730.00 775.71 821.43

455.71 501.43 547.14 592.86 638.57 684.29 730.00 775.71

318.57 364.29 410.00 455.71 501.43 547.14 592.86 638.57 684.29 730.00

load, even though it influences fuel consumption (Ford Torrey and Murray, 2014). We also assume the cost per km does not vary significantly based on truck types. Table 4 displays the cost values used for the analysis. For a given mix of good, repairable and disposable pallets, the inequality defining the maximum distance (d1þd2) for a pallet pooling system is:

2ðd1 þ d2 Þktr þ NðrÞkrep  Nðr þ gÞkman þ 2Dktr

(6)

The left-side accounts for the cost of transportation from a customer's supply chain end point to the repair facility, and from there to a customer's supply chain start point, plus the cost for repair in a closed loop model. The right-side accounts for the cost of producing and supplying the new pallets necessary to replace the abandoned ones in an open-loop model. Thus:

ðd1 þ d2 Þ 

Nðr þ gÞkman þ 2Dktr
NðrÞkrep 2ktr

(7)

4.3. Results The analytical models are used to determine the maximum total repositioning distance for a closed loop model to be economic and environmental viable compared to an open loop. We assume D ¼ 250 km. For both truck types, Tables 5 and 6 report the maximum average repositioning distance, respectively for the environmental and economic measures. Fig. 3 graphically displays results considering the values of g (percentage of good pallets collected) associated with the three handling/loading conditions. Impact of handling/loading conditions: As seen across Fig. 3, larger fractions of good pallet cores (g) or repairable cores (r) always result in larger maximum distances within which a pallet pooling system is viable. As the fraction of repairable pallets (r) increases, for a fixed percentage of good pallets (g), the fraction of pallets that are either lost or discarded (b) and consequently would need to be replaced with newly manufactured pallets is further reduced. This incentivizes adoption of pickup locations farther away. From an environmental perspective, this behavior is mainly due to the emissions associated with the repair operations of a recovered pallet being only a fraction (less than 4%) of the emission associated with the manufacturing of a new one required in the open loop system. From an economic standpoint, the incentive for farther pickup operations (thus the adoption of a pooling system) comes from the cost of repairing a pallet being a third of the cost of manufacturing a new one. In both cases, the maximum distance is

Fig. 3. Maximum total repositioning distance as a function of CO2-eq emissions and costs, for a 16t and a 32t truck.

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

161

Fig. 4. Equidistant contour lines for breakeven between repairing and sourcing a new pallet with a 16ton truck, in which the x-axis is the fraction of good pallets (g), and the y-axis is the faction of repairable pallets (r).

determined when the transportation emissions and costs surpass these differences between sourcing a newly manufactured pallet and repairing a used one (with transportation). The proportion of pallet cores in good condition (g) is largely dependent on the handling and loading treatment during service (as observed through the PDS® software). Pallet recyclers can benefit from a thorough understanding of the handling and loading circumstances of their potential customers and its impact on their bottom line. For example, when using a 32ton truck, an increase in the fraction of repairable pallets (r) by 10% would result in an increase of 65 km (environmental) and 45 km (economic) in the end-to-start breakeven distance. In turn, increasing the fraction of good pallets (g) by 10% allows for an even larger increase in the distance: 71 km (environmental) and 68 km (economic). This can allow expansion of the customer base for pallet pooling providers to areas previously considered too far. Rather more proactively, pallet pooling providers exploring such customers might engage in management actions (such as forklift operation training, etc.) leading to such increase in both the repairable and good pallet fractions. When focused on cost measures, the equidistant contour graph

in Fig. 4a for a 16 ton truck reveals that smaller benefits result from increasing the fraction of repairable pallets (r) than from a similar increase on the fraction of good pallets (g). This, however, is not true of the environmental function (Fig. 4b). The benefits from increasing either fraction g or r by identical amounts is approximately the same. These contour plots constitute guidelines for pooling companies seeking to improve their environmental footprint by maintaining the operations within the radius of maximum distance (d1þd2) to their recycler within the carbon favorable zone or to improve their business profitability. For example, a pooling company considering expansion of its network into an industrial park located just over 160 km away, would find it environmentally advantageous to engage with companies whose business practices result in fractions that are g > 0.2 and r > 0.4 (i.e. shadowed quadrant in Fig. 4b). A similar analysis can be drawn for the cost function. Impact of transportation mode: As seen across the plots in Fig. 3, for fixed fractions of repair and good pallets, employing larger capacity trucks results in larger maximum distances (d1þd2) where a pooling system is economically and environmentally

Fig. 5. Flowchart of the simulation model for crossdocking and take-back scenarios.

162

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

Table 7 Pallet SLA data for different handling and loading conditions. Handling/loading conditions

Good/light

Average/medium

Rough/heavy

Expected Service Life in cycles (SLh) Average number of cycles with repair during pallet lifetime (CRh) Repair ratio (CRh/SLh) Degradation factor (dfh)

30 14 0.47 1

15 11 0.73 2

9 8 0.89 3

The main difference occurs at the retailer's facility, where in a crossdocking scenario inspection occurs and only pallets needing repair are transported to the repair facility. In take-back, inspection occurs at the repair facility. The simulation model is built using the commercial software Anylogic® (release 7.2) with the following assumptions:

feasible. This can be appreciated by the steeper slopes obtained when employing the larger capacity 32 ton truck and explained by economies of scale. As full loads of pallets are assumed in every trip, the fixed emissions and costs associated with the movement of an empty truck are allocated across a larger batch than when employing a smaller truck. When calculating the slopes for the environmental measure, the linear relationships for the 32 ton trucks generate a slope approximately twice that of the 16 ton truck. Similarly, the slope for the cost measure for the 32 ton truck is larger than the 16 ton truck by a ratio of 2.6 to 1. Considering that the 32 ton truck presents smaller empty truck emissions, the relationship for this transportation choice will be a steeper line.

 To represent a generic set of customers experiencing the same expected demand but located at different distances from the pallet pooling provider, the repositioning distances (d1 and d2) and the crossdocking distance (d3) are randomized with uniform distributions. The repositioning distances (d1, d2, and d3) are consistent with the results obtained from the analytical model to ensure feasibility of the closed-loop model.  The pallet demand and reorder levels for the product manufacturer and the distribution centers are simulated based on data gathered from a case study on a grocery supply chain (Mazeika Bilbao, 2011). Demand is assumed to follow a normal distribution, with mean and standard deviation of demand for the product manufacturer mp ¼ 2263, sp ¼ 212.09, while for the distribution centers mdc ¼ 622, sp ¼ 37.72. The reorder level is calculated assuming a cycle service level of 95% and a fixed order quantity given by the truck capacity (480 pallets). Every time the reorder level is reached at the product manufacturer or distribution center facilities, new pallets are ordered from the pallet manufacturer. The lead time for pallet orders is assumed to be two days.  Each customer's location in the supply chain is characterized by a pre-defined set of pallet handling and loading conditions as defined in section 4.1.  The Virginia Tech FasTrack protocol (FasTrack, 2008) defines a pallet handling cycle as a sequence of 16 tasks. Based on this definition, the simulation model assumes one cycle encompasses all the activities performed from the time a pallet arrives to a location in the customer's supply chain and ends when it reaches the subsequent location.  To model the degradation function and the rate of repair for pallets after each cycle, data from Tornese et al. (2016) derived from the PDS® software are used. For each condition, the Service Life Analysis (SLA) of the PDS® provides the expected breakdown profile and service life (Table 7). The repair ratio between the number of cycles after which a pallet needs repair during its lifetime and its expected service life (CRh/SLh) is calculated for each handling condition h. This value is used as a proxy for the

5. Simulation model An agent-based discrete event simulation model is developed to determine the impact two common operational policies, take-back and crossdocking, and customer characteristics have on economic and environmental KPIs from a pallet pooling provider perspective. Initially, we developed a simplified closed-loop queuing network of the pallet system with a retailer, a distributor, a repair center, and travel nodes among the three entities. The retailer, distributor, and repair centers are modeled using first-come firstserve (FCFS) single-server nodes with exponential time duration. Further, the travel nodes are modeled as delay centers with a general service time distribution. Closed-form expressions for performance measures, such as the inventory turns, are difficult to obtain for even simplified networks with tractable service time distributions. Instead, the network is numerically analyzed with three pallet batches and different pallet mix conditions. We vary the repair probability from 0 to 1 in increments of 0.05 under both cross-docking and take-back policy and capture throughput and inventory turn differences. Our results indicate that the inventory turns of the pallets increase by an average of 15% with crossdocking in comparison to the take-back scenario. However, this model is not analytically tractable and does not consider the effect of pallet handling/loading conditions, inventory control mechanisms at the retailer and distributor end, or variations in demand levels. Therefore, we develop a detailed agent-based discrete-event simulation model that considers customers' supply chains on three echelons (product manufacturer - PM, distribution center - DC, retailer), and a set of pallet manufacturing and repair facilities, as shown in Fig. 2. Each pallet is treated as an agent, allowing palletspecific data to be collected. Fig. 5 illustrates a flow chart of the process simulated for the crossdocking and take-back scenarios.

Table 8 Cost and emission coefficients used in the simulation model. Coefficient

Notation

Unit

Value

Source

Cost of lost pallets Holding operational cost Holding capital cost Repair emissions after good handling/loading Repair emissions after average handling/loading Repair emissions after rough handling/loading

kl ko kc Crep,1 Crep,2 Crep,3

$/pallet lost $/pallet position-year $/pallet-year kgCO2-eq/pallet kgCO2-eq/pallet kgCO2-eq/pallet

7.000 0.500 0.750 0.184 0.335 0.535

Direct observations Derived from (Roy et al., 2016) Derived from (Roy et al., 2016) Derived from (Tornese et al., 2016) Derived from (Tornese et al., 2016) Derived from (Tornese et al., 2016)

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

163

Table 9 Factors in the simulation model. Factors

Levels

Handling and loading conditions at the facilities

Good Average Rough

Repositioning policy

Take-back Crossdocking

Network type

Long end-to-start distance (low customer density, average crossdocking and take-back distance z 225 km) Short end-to-start distance (high customer density, average crossdocking and take-back distance z 40 km)

probability that a pallet will need repair after each cycle under the specified conditions.  Each new pallet is provided a maximum service life, expressed in number of cycles, which is modeled as a triangular distribution with mode 30.  A degrading function is developed to model the deterioration of a pallet after each use cycle. For each cycle iþ1, the number of cycles left in the pallet service life is:

were compounded at 25% annually (Roy et al., 2016). Table 8 contains notation and values used in the emissions and cost functions. Sets:  H, set of handling/loading conditions: h¼1 (good), h ¼ 2 (average), h ¼ 3 (rough).  I, set of truck types: i ¼ 1 (16t truck); i ¼ 2 (32t truck).  J, set of trips, indexed on j ¼ 1, …, m. Simulation outputs:

cyclesleftiþ1 ¼ cycleslefti
dfh

(8)

 The degradation factor dfh is calculated as the integer closest to the ratio of the maximum service life realizable under good handling/loading conditions (i.e., 30), and the expected service life under conditions h (SLh).

dfh ¼ intðSL1 =SLh Þ

(9)

 In each cycle, 5% of the pallets get lost, which is consistent with Carrano et al. (2015) (6%) and Roy et al. (2016) (5e15%). The cost of lost pallets is accounted for in the total costs borne by the pallet pooling provider.  A 32t EURO5 diesel truck is used to ship new and repaired pallets to the customer (d2), to collect damaged pallets from the start point and the distribution center, and for shipping goods along the supply chains. A smaller truck (16t EURO5) is employed for the collection of used pallets from the end point, both for crossdocking (d3) and for shipping to the repair facility (d1). Full loads are considered for all trips.  The handling/loading conditions influence the state of a pallet and severity of the damage, thus the type of repair activity and materials needed. This is accounted for in the repair CO2-eq emissions (Tornese et al., 2016).  The KPIs are: total cost for the pooling company; total carbon equivalent emissions; and total number of pallets in the pool. In addition, the mean time to reissue, an indicator which measures the average time for a pallet to travel along the supply chain and come back to the start point, the pallet cycles and the number of repairs per pallet are monitored. These pallet-related indicators are collected at the end of a pallet lifetime. The simulation model captures additional costs and emissions beyond those captured by the analytical model. Specifically, costs of lost pallets, operational holding costs (including the storage and operational costs of average inventory held at the repair facility) and holding costs for capital investments. Operational costs are calculated using warehouse storage costs of $6/sq-ft/year and operational utilities of approximately $2.5/sq-ft/year. Capital costs

 Nman ¼ total number of pallets manufactured.  Nl ¼ total number of pallets lost.  Nh ¼ total number of pallets repaired after a cycle with handling/ loading conditions h.  In ¼ average inventory of the network per year [pallets/year].  Ir ¼ average inventory at the repair facility per year [pallets/ year].  dj ¼ distance covered in trip j [km]. Also, Ctr,tot,i ¼ Ctr(empty),i þ Ctr(load),i and denotes the emission coefficients for transportation (as defined in Section 4.2) for the truck type i [kgCO2-eq/km]. The total costs (K) and carbon equivalent emissions (E) per year are calculated as.

E ¼ Cman Nman þ

X

Crem;h Nh þ

h2H

X

Ctr;tot;i dj

(10)

j2J; i2I

K ¼ kman Nman þ kl Nl þ kc In þ ko Ir þ krem

X h2H

Nh þ ktr

X

dj

j2J

(11)

5.1. Simulation experiment The aims are (1) to understand if repositioning strategies significantly alter KPIs for the pallet pooling provider, and (2) to quantify how a customer's characteristics (network design and handling/loading condition) influence pallet pooling provider's performance. The simulation model systematically explores the three factors in Table 9. The network type is explored via the total repositioning distance. In Fig. 6 the difference between the take back and the crossdock scenarios is outlined. With take-back, the total repositioning distance is the distance from the repair facility to the start of the customer's supply chain. With crossdocking, the total repositioning distance is the distance from the end of a supply chain (a retailer) back to the beginning of another supply chain (a manufacturer). The simulation experiments include two steps. Step 1 represents a baseline case, in which a pallet pooling provider has

164

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

Fig. 6. Schematic of repositioning scenarios.

customers with different repositioning distances and different handling/loading conditions in each facility. Network type and handling/loading conditions are randomized. Each pallet can experience good, average or rough handling/loading in each cycle with equal probability. The repositioning distance follows a uniform distribution that includes both short and long distances (i.e., d1 varies between 60 and 80 km; d2 and d3 vary between 30 and 250 km). In step 2, the simulation model evaluates how the response variables react when the extreme cases are considered. This results in 23 cases with the following factors and levels: handling/loading (all good or all bad conditions), network type (long or short end-to-start distance), and scenarios (take-back and crossdocking). Specifically, long end-to-start distances are modeled through a uniform distribution ranging from 200 to 250 km, while short ones go from 30 to 50 km, and a generic set of customers is still represented. For both steps, 15 replications per instance are run using a simulation time of one year, and a confidence level of 95%. Results from the simulation model described in Section 5.2 capture

steady-state performance. Mean and half-width performance are reported. Also reported is their expected difference, calculated as the expected takeback performance minus the expected crossdocking performance, divided by the expected crossdocking performance. Performance differences are considered significant if the mean performance is not within the other's half-width.

5.2. Results For take-back and crossdocking, the baseline results, captured via step 1, are summarized in Table 10. A crossdocking scenario reduces the time a pallet spends in the system because good pallets are immediately sent back to a start point. In fact, crossdocking presents a 5% faster time to reissue. This differs from a take-back scenario, in which all pallets (even those in good condition) are sent to the repair facility before being re-injected to a start point. This reduced time is the fundamental advantage of a crossdocking over take-back policies. For a given service level, a crossdocking

Table 10 Baseline results from simulation model (step 1). Take-back

Crossdocking

Mean

Half-width

Mean

Half-width

Cross vs take

Significant

68637.33 25,235.06 242.31 185,760.00 365418.20 48144.00 241811.54 172052.11 0.00 1107300.55

2368.56 1922.00 10.83 9488.31 9724.21 1395.27 5677.21 5443.67 0.00 35514.74

69,445.60 18,047.60 243.93 169536.00 370176.67 43728.00 245797.25 171180.42 12246.31 1100401.78

2296.91 1389.68 11.22 7657.53 9859.17 898.12 5838.62 4996.06 1016.09 33578.81

þ1%
28% þ1%
9% þ1%
9% þ2%
1% e
1%

no yes no yes no yes no no yes no

Manufacturing Repair Transport Transport repair Transport repositioning Transport crossdocking Total emissions

239630.40 140517.36 50125.51 241000.69 171191.87 0.00 842465.83

12239.92 3750.02 1439.44 5648.06 5416.46 0.00 28036.75

218701.44 142194.75 45761.82 244967.73 170324.51 13042.33 834992.57

9878.22 3815.11 947.66 5808.79 4971.09 1082.14 26147.20


9% þ1%
9% þ2%
1% e
1%

yes no yes no no yes no

N pallets in system N pallets disposed Orders new pallets PM Orders new pallets DC N cycles per pallet N repairs per pallet Mean time to reissue

33646.73 18467.93 36.53 37.93 14.60 10.47 9396.71

2562.67 283.48 4.19 14.91 0.01 0.03 203.96

24063.47 22527.73 31.47 36.80 14.84 10.13 8903.73

1852.90 389.28 4.62 14.49 0.01 0.02 197.04


28% þ22%
14%
3% þ2%
3%
5%

yes yes yes no yes yes yes

Costs [$] Lost pallets Holding capital Holding operational Manufacturing Repair Transport to start point Transport to repair Transport repositioning Transport crossdocking Total costs Emissions [kg CO2-eq]

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

165

Table 11 Effect of handling/loading conditions in crossdocking scenario, compared to the results of step 1 (baseline). Crossdocking Good

Half-width

vs baseline

Rough

Half-width

vs baseline

77159.13 22280.17 168.31 124416.00 187316.33 34160.00 115416.05 86536.77 23679.39 671132.15

2014.38 1631.59 6.66 5116.92 1163.20 1363.10 782.35 1448.25 2465.86 4844.49

þ11% þ23%
31%
27%
49%
22%
53%
49% þ93%
39%

33444.13 217345.50 68453.28 893184 448096.70 229061.30 458559.50 209413.90 1688.21 2559246.71

284.97 4331.93 1303.61 16473.16 108.47 3696.13 2535.01 2632.54 169.08 29630.22


52% þ1104% þ27963% þ427% þ21% þ424% þ87% þ22%
86% þ133%

Manufacturing Repair Transport Transport repair Transport repositioning Transport crossdocking Total emissions

160495.97 34465.54 36278.09 115239.01 86104.09 25218.54 457801.23

6600.74 214.11 1301.59 777.73 1441.02 2626.15 5571.96


27%
76%
21%
53%
49% þ93%
45%

1152207.36 229041.03 229340.8 456667.31 208366.87 1797.95 2287812.26

21250.38 58.02 3682.95 2521.20 2619.38 180.06 28553.43

þ427% þ69% þ401% þ86% þ22%
86% þ174%

N pallets in system N pallets disposed Orders new pallets PM Orders new pallets DC N cycles per pallet N repairs per pallet Mean time to reissue

29706.87 742.4 34.2 0 29.57 16.56 21415.22

2175.44 217.92 1.77 0 0.06 0.53 1424.53

þ23%
97% þ9%
100% þ99% þ64% þ141%

289794 3156.27 191.13 220 9.98 7.38 100432.55

5775.90 319.45 0.67 11.33 0.02 0.05 1096.69

þ1104%
86% þ507% þ498%
33%
27% þ1028%

Costs [$] Lost pallets Holding capital Holding operational Manufacturing Repair Transport to start point Transport to repair Transport repositioning Transport crossdocking Total costs Emissions [kg CO2-eq]

scenario can be achieved with lower total number of pallets in the system. Under a crossdocking scenario, the same customer demand is satisfied with a 28% reduction in the number of pallets in the pool. This results in significantly lower holding and manufacturing costs and emissions for crossdocking. In crossdocking, the orders for new pallets to the manufacturer was also reduced by 14%, because the demand is partially satisfied by the used pallets recovered through crossdocking. Fewer shipments of new pallets to the start point results in a 9% reduction of transportation costs and emissions. However, these advantages must be traded-off with the increased transportation costs and emissions required in crossdocking good pallets to be reinjected into the supply chain. A 16t truck is used for transportation to repair (d1) and cross-docking (d3), while a 32t truck is employed in the repositioning phase (d2). Therefore, even though a full load is considered for all trips, the smaller capacity of the truck increases the number of trips required to transport the same amount of pallets, causing inefficiency due to economies of scale. In the baseline case, the reduction in holding and manufacturing costs and emissions from crossdocking were met with an equivalent increase in the transportation costs. Thus, the total costs and carbon equivalent emissions are not significantly different between the two scenarios. In addition, crossdocking is more complex to implement and requires additional inspection resources, not captured in the model. Therefore, for the baseline scenario, a takeback scenario is recommended. The scenarios also do not significantly influence the costs nor emissions due to lost pallets, operational holding, repair, or the transportation to repair and repositioning. A crossdocking policy requires fewer new pallets than takeback, due to increased pallet turns. Thus, whether a take-back or crossdocking policy is better depends on the trade-off between repair costs and pallet manufacturing costs. In this analysis, a new pallet manufacturing cost of $3 per pallet was considered. However, if a pooling provider chooses not to produce its own pallets and buys them from a pallet manufacturer instead, the cost for a new

pallet can be significantly higher. For higher purchase costs, crossdocking becomes more economical than take-back. As an example, considering a purchase cost of 15$/pallet, crossdocking would result in 4% cost savings compared to take-back. In Step 2, the impact of handling/loading conditions and repositioning distance on the system performance are evaluated and compared to the baseline results of step 1 (for both crossdocking and take-back). Results for handling/loading conditions are displayed in Tables 11 and 12, while those related to repositioning distance are in the appendix (Tables A.1 and A.2). For both scenarios, all performance factors comparing good and rough handling were significant. Impact of Handling/loading conditions: under both scenarios the model shows handling/loading conditions significantly affect all costs and carbon equivalent emission components. In the extreme case of good handling/loading in all of the customer's facilities, costs and emissions are respectively 39% (37%) and 45% (41%) lower in crossdocking (take-back) than those from the baseline. This reduction is due to good handling/loading conditions resulting in a lower number of pallets being repaired. Additionally, good handling/loading practices decrease the costs and emissions for transportation to the repair facility and repositioning by about 50% in both scenarios. Moreover, the number of new pallets produced decreases, with a reduction of 27% in both manufacturing costs and emissions for crossdocking and 8% for take-back. This reduction is due to the increased pallet longevity that allows for multiple reuses of the same pallet with minimal repair (about þ100% in both scenarios). For this reason, the number of pallets in the system at the end of the simulation time is significantly higher (þ23% and þ38%), as fewer pallets are disposed in the time considered. Finally, the mean time to reissue increases by 141% in crossdocking and 97% in take-back. This is due to a lower pallet repair rate: with good handling/loading, a pallet is more likely to survive several cycles before enduring damage that requires repair and repositioning to the start point.

166

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

Table 12 Effect of handling/loading conditions in take-back scenario, compared to the results of step 1 (baseline). Take-back Good

Half-width

vs baseline

Rough

Half-width

vs baseline

74244.80 34830.87 161.95 171360.00 174859.10 47024.67 110140.70 82971.23 e 695593.30

1613.60 261.22 4.45 1049.89 827.25 321.25 510.92 1333.80 e 3211.06

þ8% þ38%
33%
8%
52%
2%
54%
52% e
37%

33216.87 217576.23 68202.30 893952.00 448069.27 229232.00 458325.13 209838.61 e 2558412.41

331.33 4340.79 1258.93 16500.76 103.11 3718.77 2445.87 2238.17 e 29106.08


52% þ762% þ28 047% þ381% þ23% þ376% þ90% þ22% e þ131%

Manufacturing Repair Transport Transport repair Transport repositioning Transport crossdocking Total emissions

221054.40 32174.08 49714.5 109989.30 82556.37 e 494793.10

1354.35 152.22 315.41 509.31 1327.13 e 2279.92


8%
77%
2%
54%
52% e
41%

1153198.08 239717.07 229209.72 456431.47 208789.42 e 2287345.77

21285.98 55.17 3705.43 2432.97 2226.98 e 28035.23

þ381% þ71% þ357% þ89% þ22% e þ172%

N pallets in system N pallets disposed Orders new pallets PM Orders new pallets DC N cycles per pallet N repairs per pallet Mean time to reissue

46441.13 72.47 50.27 0.47 29.47 19.96 18499.68

348.29 5.69 0.38 0.73 0.09 0.13 60.63

þ38%
100% þ38%
99% þ102% þ91% þ97%

290101.60 3137.13 191.33 220.13 9.98 7.41 100298.26

5787.72 329.76 0.77 11.32 0.02 0.05 1053.27

þ762%
83% þ424% þ480%
32%
29% þ967%

Costs [$] Lost pallets Holding capital Holding operational Manufacturing Repair Transport to start point Transport to repair Transport repositioning Transport crossdocking Total costs Emissions [kg CO2-eq]

In the extreme case of rough handling/loading in every facility, both costs and emissions increase drastically when compared to the baseline. This is true regardless of whether a take-back or crossdocking scenario is implemented. Overall, rough handling/ loading practices cause higher costs (þ133%, þ131%) and increased emissions (þ174%, þ172%) for crossdocking and take-back, respectively. More importantly, it requires from about 700% to 1000% more pallets in the pool to obtain the same service level. Impact of repositioning distance: while this factor directly affects transportation emissions and costs, the magnitude of the impact is less than that from handling/loading conditions. Transportation costs and emissions decrease by approximately 70% for short distances in both scenarios, with an overall decrease of 12% in total costs and about 15% in emissions with respect to the baseline. Long distances can increase transportation costs and emissions by about 60%, with a total increase from about 10% to 13% in costs and emissions. 6. Conclusions Analytical and simulation models are developed to investigate pallet pooling systems from both an environmental and economic perspective. The analytical models are used to consider maximum total repositioning distance for a closed loop model to be economically and environmentally viable compared to an open loop. These models can generate two types of insights for a pallet pooling provider: (1) given a network of customers, determine the maximum average repositioning distance of a repair facility from the end and start points of the customers' supply chain (facility location problem); (2) given a repair facility, evaluate the viability of serving a specific customer, considering the distances of its end and start points (customer choice problem). A simulation model is used to evaluate the impact of repositioning strategies and to verify

how a customer's characteristics (network design and handling/ loading condition) influence the pooling company performance. While crossdocking policies were able to satisfy customer demand with a 28% reduction in the number of pallets in the pool compared to the baseline take-back policy, the overall costs and emission differences were not significant. The fundamental advantage of crossdocking over take-back policies is the decrease in non-value added time associated with good pallets being transported to repair facilities and then being sent out for use. However, such savings were not found to be enough to offset the additional transportation costs required with crossdocking due to smaller trucks transporting the good pallets from the crossdock locations to a start point. Thus, take-back policies are recommended for baseline scenarios. Crossdocking reduces the holding and manufacturing costs/emissions of new pallets, and can be beneficial if a higher manufacturing (or purchase) cost for a new pallet is considered. Results from both models indicate that customers' handling/ loading conditions have the greatest effect on both a pallet pooling provider's costs and carbon equivalent emissions. Thus, a thorough understanding of the handling and loading circumstances of their potential customers should be a key focus during customer selection and in pallet pooling logistics network design. The scenarios analyzed for pallet pool networks in this paper are static in nature. Developing dynamic repositioning strategies for the pallet batches based on the current state of the system is a topic for future research.

Acknowledgments This project was partially funded by Material Industry of America (MHIA) Research Grant #11100576.

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

167

Appendix

Table A.1 Effect of repositioning distance in crossdocking scenario, compared to the results of step 1's baseline. Crossdocking Short distance

Half-width

vs baseline

Long distance

Half-width

vs baseline

69144.6 18041.68 250.00 169440.00 370363.00 43749.33 245859.63 49098.21 3454.36 969400.82

2292.03 1295.97 8.82 7254.37 9514.91 861.77 5647.85 1333.37 224.98 28357.74

0% 0% þ2% 0% 0% 0% 0%
71%
72%
12%

69105.40 18169.93 244.46 169920.00 370003.53 43717.33 245628.19 276128.77 19624.05 1212541.65

2255.40 1354.49 12.90 7480.70 9561.39 908.49 5662.18 7188.24 1376.91 35671.98

0% þ1% 0% 0% 0% 0% 0% þ61% þ60% þ10%

Manufacturing Repair Transport Transport repair Transport repositioning Transport crossdocking Total emissions

218577.60 142295.72 45783.22 245031.55 48852.73 3678.89 704219.72

9358.13 3676.72 909.86 5620.11 1326.70 239.61 21081.84

0% 0% 0% 0%
71%
72%
16%

219196.80 142096.80 45751.07 244798.13 274748.12 20899.63 947490.55

9650.10 3706.51 955.57 5633.90 7152.29 1466.40 28465.08

0% 0% 0% 0% þ61% þ60% þ13%

N pallets in system N pallets disposed Orders new pallets PM Orders new pallets DC N cycles per pallet N repairs per pallet Mean time to reissue

24055.53 22528.67 31.33 37.00 14.84 10.12 8903.94

1727.96 391.74 4.56 14.11 0.01 0.01 201.70

0% 0% 0% þ1% 0% 0% 0%

24226.53 22541.27 31.47 37.07 14.83 10.13 8916.04

1805.99 686.72 4.49 14.15 0.03 0.01 189.18

þ1% 0% 0% þ1% 0% 0% 0%

Costs [$] Lost pallets Holding capital Holding operational Manufacturing Repair Transport to start point Transport to repair Transport repositioning Transport crossdocking Total costs Emissions [kg CO2-eq]

Table A.2 Effect of repositioning distance in take-back scenario, compared to the results of step 1's baseline. Take-back Short distance

Half-width

vs baseline

Long distance

Half-width

vs baseline

68476.80 25352.98 238.95 186144 365404 48101.33 241838.20 48938.89 e 984495.18

2300.74 1842.33 9.19 9150.22 9386.32 1292.49 5497.19 1308.59 e 30663.99

0% 0%
1% 0% 0% 0% 0%
72% e
11%

68551.47 25156.17 236.2 185568.00 365979.13 48080.00 242183.77 275116.76 e 1210871.51

2315.99 1753.71 9.23 8845.86 9152.56 1314.23 5320.79 6853.14 e 35473.84

0% 0%
3% 0% 0% 0% 0% þ60% e þ9%

Manufacturing Repair Transport Transport repair Transport repositioning Transport crossdocking Total emissions

240125.76 140528.43 50082.83 241029.13 48693.55 e 720459.69

11803.78 3625.10 1335.65 5469.03 1301.45 e 23461.04

0% 0% 0% 0%
72% e
14%

239382.72 140714.17 50062.57 241373.23 273741.17 e 945273.87

11411.15 3524.69 1357.65 5293.04 6818.87 e 28355.56

0% 0% 0% 0% þ60% e þ12%

N pallets in system N pallets disposed Orders new pallets PM Orders new pallets DC N cycles per pallet N repairs per pallet Mean time to reissue

33803.93 18461.67 36.53 38.20 14.61 10.48 9407.35

2456.43 285.56 4.09 14.47 0.02 0.02 194.72

0% 0% 0% þ1% 0% 0% 0%

33541.53 18519.40 36.33 38.2 14.61 10.48 9404.61

2338.29 291.00 4.21 14.50 0.01 0.02 189.75

0% 0%
1% þ1% 0% 0% 0%

Costs [$] Lost pallets Holding capital Holding operational Manufacturing Repair Transport to start point Transport to repair Transport repositioning Transport crossdocking Total costs Emissions [kg CO2-eq]

168

F. Tornese et al. / Journal of Cleaner Production 172 (2018) 155e168

References Alshamsi, A., Diabat, A., 2015. A reverse logistics network design. J. Manuf. Syst. 37 (3), 589e598. Bengtsson, J., Logie, J., 2015. Life cycle assessment of one-way and pooled pallet alternatives. Proc. CIRP 29, 414e419. https://doi.org/10.1016/ j.procir.2015.02.045. Bhattacharjya, J., Kleine-Moellhoff, P., 2013. Environmental concerns in the design and management of pallets. In: Camarinha-Matos, Luis M., Scherer, Raimar J. (Eds.), Collaborative Systems for Reindustrialization, vol. 408. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 569e576. http://link.springer.com/10.1007/ 978-3-642-40543-3_60. Bottani, E., Montanari, R., Rinaldi, M., Vignali, G., 2015. Modeling and multiobjective optimization of closed loop supply chains: a case study. Comput. Ind. Eng. 87, 328e342. https://doi.org/10.1016/j.cie.2015.05.009. Buehlmann, U., Bumgardner, M., Fluharty, T., 2009. Ban on landfilling of wooden pallets in North Carolina: an assessment of recycling and industry capacity. J. Clean. Prod. 17 (2), 271e275. https://doi.org/10.1016/j.jclepro.2008.06.002. Carrano, A.L., Pazour, J.A., Roy, D., Thorn, B.K., 2015. Selection of pallet management strategies based on carbon emissions impact. Int. J. Prod. Econ. 164, 258e270. https://doi.org/10.1016/j.ijpe.2014.09.037. Carrano, A.L., Thorn, B.K., Woltag, H., 2014. Characterizing the carbon footprint of wood pallet logistics. For. Prod. J. 64 (7e8), 232e241. https://doi.org/10.13073/ FPJ-D-14-00011. Documentation for Truck Transport Process, 2016. Fraunhofer Institute for Building Physics. University of Stuttgart. In: http://www.gabi-software.com/fileadmin/ GaBi_Databases/GaBi_Database_Upgrade_2016_Documents_for_website_ upload/Documentation_GaBi_Transport_Processes_Duty_Vehicles_2016.pdf. Elia, V., Gnoni, M.G., 2015. Designing an effective closed loop system for pallet management. Int. J. Prod. Econ. 170, 730e740. https://doi.org/10.1016/ j.ijpe.2015.05.030. FasTrack, 2008. FasTrack Protocol: VT Unit-load Material Handling FasTrack. http:// unitload.vt.edu/facilities/william-sardo-pallet-lab/FasTrack-Info.pdf. Fleischmann, M., Beullens, P., Bloemhof-Ruwaard, J.M., Van Wassenhove, L.N., 2001. The impact of product recovery on logistics network design. Prod. Oper. Manag. 10, 156e173. Ford Torrey, W., Murray, D., 2014. An Analysis of the Operational Costs of Trucking: a 2014 Update. American Transportation Research Institute. ~ ona, L., Farreny, R., Navarro, P., Boschmonart-Rives, J., 2016. Life cycle García-Duran assessment of a coniferous wood supply chain for pallet production in catalonia. Spain. J. Clean. Prod. 137, 178e188. https://doi.org/10.1016/ j.jclepro.2016.07.032. Gasol, C.M., Farreny, R., Gabarrell, X., Rieradevall, J., 2008. Life cycle assessment comparison among different reuse intensities for industrial wooden containers. Int. J. Life Cycle Assess. 13 (5), 421e431. https://doi.org/10.1007/s11367-0080005-0. Getting Paid What You’re Worth. 2015. Pallet Enterprise. Accessed October 4. http:// www.palletenterprise.com/articledatabase/view.asp?articleID¼2050&Page¼3. Glock, C.H., 2017. Decision support models for managing returnable transport items in supply chains: a systematic literature review. Int. J. Prod. Econ. 183, 561e569. https://doi.org/10.1016/j.ijpe.2016.02.015. Gnimpieba, Z.D.R., Nait-Sidi-Moh, A., Durand, D., Fortin, J., 2015. Using Internet of things technologies for a collaborative supply chain: application to tracking of

pallets and containers. Procedia Comput. Sci. 56, 550e557. https://doi.org/ 10.1016/j.procs.2015.07.251. Gnoni, M.G., Rollo, A., 2010. A scenario analysis for evaluating RFID investments in pallet management. Int. J. RF Tech. 2 (1), 1e21. https://doi.org/10.3233/RFT2010-001. Govindan, K., Soleimani, H., Kannan, D., 2015. Reverse logistics and closed-loop supply chain: a comprehensive review to explore the future. Eur. Jour. Oper. Res. 240 (3), 603e626. Mazeika Bilbao, A., Carrano, A.L., Thorn, B.K., 2010. Environmental impact analysis of pallets management. In: IIE Annual Conference. Proceedings. Institute of Industrial Engineers-Publisher, Cancun, Mexico. Mazeika Bilbao, A., 2011. Environmental Impact Analysis of Alternative Pallet Management Systems. MS thesis. Rochester Institute of Technology. Mazeika Bilbao, A., Carrano, A.L., Hewitt, M., Thorn, B.K., 2011. On the environmental impacts of pallet management operations. In: Alfred, L., Guiffrida (Eds.), Manag. Res. Rev. 34 (11), 1222e1236. https://doi.org/10.1108/ 01409171111178765. Mead, J.M.B., 2010. Reliable Pallets: Carrying the World on Their Backs. ISO Focusþ, pp. 18e19. February. Millwood Inc, 2015. Technical Communication with Ralph Rupert, Manager of Unit Load. NAICS, 2012. 321920-Wood Container and Pallet Manufacturing. Ng, R., Shi, C.W.P., Tan, H.X., Song, B., 2014. Avoided impact quantification from recycling of wood waste in Singapore: an assessment of pallet made from technical wood versus virgin softwood. J. Clean. Prod. 65, 447e457. https:// doi.org/10.1016/j.jclepro.2013.07.053. Raballand, G., Aldaz-Carroll, E., 2005. How do differing standards increase trade Costs? The case of pallets. SSRN Electron. J. https://doi.org/10.2139/ssrn.665064. Ray, C.D., Michael, J.H., Scholnick, B.N., 2006. Supply-chain system costs of alternative grocery industry pallet systems. For. Prod. J. 56 (10), 52e57. Roy, D., Carrano, A.L., Pazour, J.A., Gupta, A., 2016. Cost-effective pallet management strategies. Transp. Res. E Log. 93, 358e371. https://doi.org/10.1016/ j.tre.2016.06.005. Roy, D., Pazour, J.A., de Koster, R., 2014. A novel approach for designing rental vehicle repositioning strategies. IIE Trans. 46 (9), 948e967. Shintani, K., Imai, A., Nishimura, E., Papadimitriou, S., 2007. The container shipping network design problem with empty container repositioning. Transp. Res. ELog. 43 (1), 39e59. Srivastava, S.K., 2008. Network design for reverse logistics. Omega 36 (4), 535e548. Stephens Baker, DeAnna, 2011. Cost Accounting Shows Profitability, Opportunities. Pallet Enterprise. June 9. http://www.palletenterprise.com/articledatabase/ view.asp?articleID¼3477. Swiss Centre for Life Cycle Inventories, 2009. Final Report EcoInvent Data v2.1. Swiss Centre for Life Cycle Inventories, Zurich. The United States Census Bureau, 2016. Industry Statistics Portal. Accessed November 7. http://www.census.gov/econ/isp/sampler.php? naicscode¼321920&naicslevel¼6. Tornese, F., Carrano, A.L., Thorn, B.K., Pazour, J.A., Roy, D., 2016. Carbon footprint analysis of pallet remanufacturing. J. Clean. Prod. 126, 630e642. https://doi.org/ 10.1016/j.jclepro.2016.03.009. US EIA, 2011. Annual Energy Review. US Energy Information Administration. www. eia.gov.