Development and implementation of a green logistics-oriented framework for batch process industries: two case studies

Transcription

1 Logist. Res. (2016) 9:9 DOI /s ORIGINAL PAPER Deveopment and impementation of a green ogistics-oriented framework for batch process industries: two case studies Nagham M. E-Berishy 1 Bernd Schoz-Reiter 1 Received: 17 December 2015 / Accepted: 3 June 2016 / Pubished onine: 15 June 2016 The Author(s) This artice is pubished with open access at Springerink.com Abstract In recent years, green ogistics has received increasing interest due to the rise in greenhouse gases emission from transport operations. It is a promising approach to managing suppy chain decisions to reduce environmenta damages. Many modes have been deveoped in the iterature to study the environmenta aspects of routing probems. However, these modes have not been impemented in practice, particuary in batch processing industries. In this paper, we deveoped a two-stage stochastic mode which integrates the tactica and operationa decision eves to reduce the economic and environmenta impacts of transport activities. This mode generates optimum vehice routes and their deivery veocities with the objective of minimizing the tota distribution cost and the reated emissions. We impemented the proposed mode in two case studies drawn from the fast-moving consumer goods industry. This resuted in a cost reduction of up to 13 % of the tota reated economic and environmenta costs compared to the actua situation at the test sites. These resuts aso show the potentia of green ogistics-based modes to improve the current modeing capabiities for batch distribution panning. This artice is part of a focus coection on Dynamics in Logistics: Digita Technoogies and Reated Management Methods. & Nagham M. E-Berishy eb@biba.uni-bremen.de Bernd Schoz-Reiter bsr@biba.uni-bremen.de 1 Internationa Graduate Schoo (IGS) for Dynamics in Logistics, University of Bremen, Bremen, Germany Keywords Green ogistics Batch process industry Suppy chain distribution Carbon emissions Environmenta objective 1 Introduction The ever-growing popuation wordwide and their consumption of basic need goods (such as fast-moving consumer goods, beverages, ceaning products) pose new chaenges for research in distribution panning. These goods are commony produced in batches due to their ow unit voume. In order to meet the requirements for high consumption rate, optimum distribution pans for batch products have to be generated across different ocations. An effective and efficient management of batch fow eads to huge savings in absoute terms for both the companies and the consumers [10]. Additionay, the growth in road traffic and the use of road freight transport raise the eve of emitted greenhouse gases (GHG) in the environment. As one of today s top market concerns, a aspects of an organization s ogistics must be green [15]. This shift towards greening the suppy chain through an environmentay friendy ogistics network design is not ony to abide by governmenta reguations but aso to meet customers expectations and socia responsibiities [22]. Distribution panning can support decision making by generating the optimum routes, minimizing the tota transportation costs as we as saving energy and, consequenty, GHG emissions. The optimum deivery routes for each vehice are generated at the tactica eve (route panning) and the optimum deivery veocities at the operationa eve (schedue panning). Therefore, the mode described in this paper is formuated as two-stage stochastic mode. The mode integrates the decisions made

2 9 Page 2 of 10 Logist. Res. (2016) 9:9 at the two decision eves to reduce the environmenta impact of transport activities. Aso, GHG emissions under uncertainties are incorporated as a function of randomy seected deivery veocities at different distribution routes within the suppy network. The vehice veocity has the most significant effects on fue consumption [20]. Thus, optima speed coud ead to greater reduction of CO 2 emissions and fue consumption for a given routing pan [6]. Athough there are a number ogistics modes in the iterature, there is a ack of research into rea-ife appications underpinned by case studies [23]. The current study vaidates a newy deveoped green ogistics mode via two rea-word case studies of mutinationa organizations drawn from the batch process industry. This paper is organized as foows: Sect. 2 reviews the extant iterature on impementation of green ogistics modes in batch distribution. This section aso describes the case studies and the background to the study. Section 3 defines the methodoogy and the deveopment of optimum distribution routes, optimum deivery veocities and reated CO 2 emissions. Computationa resuts and discussions are presented in Sect. 4. Finay, the concusions and main findings are highighted in Sect Greening batch distribution: iterature review There has been a marked increase in the consumption of goods produced using batch processes. This has resuted in a commensurate rise in production and reated distribution operations [21]. Consequenty, these reated distribution operations have contributed immensey to the tota goba transport emissions. According to Bueher and Pucher [2], a fifth of wordwide GHG emissions are due to transport activities. In fact, the transportation sector wi be the second argest emitter accounting for 24 % of the tota EU emissions in Therefore, there is a renewed sense of urgency to investigate the environmenta parameters during transport operations and deveop appropriate distribution panning modes [12]. Scheduing most of the distribution pans is based on the vehice routing probem (VRP) [17]. The VRP is a cassica combinatoria optimization probem that has become a key component of distribution and ogistics management. Through VRPs, the reduction in tota vehice s traveed distance, by itsef, provides environmenta benefits. This is due to the reduction in fue consumption, the consequent drop in poutants and resutant shorter ead times [3, 5, 7]. In recent research contributions, there have been a number of modes suggested to minimize the emissions of GHG associated with vehice routing probems based on vehice veocities. The veocity of vehices is the major determinant of environmenta emissions for a specific vehice oad [6]. The usage of speed to minimize emissions in VRP modes can be cassified into two main categories: the first is the set of modes where vehice speeds are inputs to the mode and the second consists of modes where vehice speed vaues are optimized and used as decision variabes. Furthermore, Kumar et a. [13] pointed out the necessity of impementing integrated optimization modes that combine operationa and environmenta costs eements in rea case studies. Moreover, impementing these modes in rea environment pays a vita roe in exporing their vaidity in achieving environmentay sustainabe ogistics practices [9]. As an appication of green ogistics concept in the batch process industry, Ubeda et a. [23] examined aspects of green transportation at the operationa eve where vehices speeds are inputs and proposed some green changes within the VRP based on a case study at one of the eading distribution companies in Spain. They found that an average saving of 25.5 % in CO 2 emissions was possibe due to reduction in used routes compared to the prevaiing situation in that company at the time. Moreover, Pishvaee et a. [19] proposed a bi-objective fuzzy mathematica programming mode for green ogistics design under uncertain conditions. Their mode integrated the transportation and production strategic decisions, and its appicabiity was examined in an industria case study. As their decisions were at the strategic eve, veocity speeds were reported as inputs as we. In contrast, Maden et a. [14], using vehice speed as decision variabes, minimized the tota vehice emissions per unit distance required for distribution operations. A CO 2 saving of around 7 % was recorded using a tabu search agorithm. Simiary, Qian and Egese [20] minimized the fue emissions in a road network where speed of each vehice aong each road in its path is treated as decision variabes. The integrated mode presented in this paper is nove given that veocity vaues were used both as inputs and decision variabes in an optimization mode. A two-stage stochastic mode is designed to generate the goba optimum soution for both tactica and operationa decision eves. In the first stage, the vehice veocities are known ony within certain bounds and considered as inputs to the mode. The optimum vaues for these veocities are uncertain at this stage. Once the optimum routes are known, the deivery veocities are used as decision variabes in the second stage. The optimum deivery veocities which correspond to the minimum emission vaues are then generated. Integrating emission vaues within the modeing of distribution pans is essentia in order to achieve a sustainabe baance between economic and environmenta objectives [16].

3 Logist. Res. (2016) 9:9 Page 3 of 10 9 As the impementation of the deveoped mode in case studies is as crucia as deveoping these approaches and modes, data were coected from two rea-word case studies of mutinationa batch process companies. In each case study, data used for the numerica anaysis beong to one fu month of activities. The data showed a reguar trend. In the next sections, we wi focus on generating the optima vehice routing that woud enabe reduction of the environmenta impact by optimizing the distribution panning. Thereafter, we wi propose a practica framework which considers both tactica and operationa decisions. We wi aso discuss the impementation of this framework using the two case studies. 3 Methodoogy and case studies In this section, the methodoogy used to hande batch distributions is described. This section composes two subsections: defining the environmenta vaues and deveoping the mathematica mode. Afterwards, the mode is impemented in two case studies from batch process industry in order to verify the appicabiity of the mode in practice. The characteristics of the test probems and some impementation detais are described. The optima deivery routes sequencing, the corresponding numbers of required deivery trucks and the optimum deivery speeds required to achieve the environmenta targets were then determined. 3.1 Environmenta vaues Athough there is no standard way to measure the vaues of fue consumption and GHG emissions for a mobie source, a number of formuas coud be appied to easiy cacuate their approximate vaues. Generay, there are two ways to cacuate CO 2 emissions: fue-based or distance-based methodoogy [23]. Whie the fue-based method is more reiabe, the distance-based method is more popuar due to the type of the data required. The distance-based method of estimating CO 2 emissions wi be used in this artice due to readiy avaiabe data. Two recorded vaues are required to cacuate emissions based on this method: the distance traveed (D k ) and fue consumption (EM k ) by vehices [23]. Additionay, this research used vaues of veocities instead of distances. Deivery veocities are defined as vectors of uncertain vaues on each route (between any two nodes in the network) in each period. This vector can be mathematicay formuated as foows: = (MinV, MaxV) where MinV and MaxV correspond to the minimum and maximum vehice veocity. It is expected that each vehice is driven with veocity (V tk ) and resuts in amount of emissions (E tk ). A range of MinE and MaxE corresponds to a range of deivery vehice veocities MaxV tk (x) and MinV tk (x) where x is a given vaue of uncertain parameters. If these vaues are not practicay appicabe, penaties for overage emissions F k (paid to the governments) and that for underage emissions G k (associated with deivery deays for customers) are imposed. The idea behind this definition of penaties is to obtain meaningfu combinations of companies and customers concerns. From the companies perspective, the goa is to deiver the batches to different customers as efficienty as possibe (concerning economic cost and overage emissions which are paid to the governments). From the customers point of view, the main concern is to reiaby receive the deiveries on time. Therefore, the objective function has emissions imits and stochastic trave veocities, eading to stochastic arriva times. 3.2 The mathematica formuation The proposed study presents a research methodoogy with a mathematica core based on both the economic and environmenta data as we as the capacity constraints. Six-step procedure was appied to reaize the proposed soution methodoogy as shown in Fig. 1. The first step is to buid a mode which captures the properties of the production process and its corresponding fows of materias. The second step is to extract a the required data in order to generate feasibe production pans at minimum costs. Afterwards, the decision-maker at the pant uses this data as we as further knowedge or expectations about the current and future situation on the shop foor to generate a set of assumptions. These inputs are caed scenarios. Then, an initia soution is generated for each given scenario. The fifth step is to anayse the production schedue and interactive modifications of the deveoped mode based on the experience and knowedge of the decision-maker and the feasibiity of the production pans. The sixth and ast step in this procedure is to approve the generated soution. This approva is based on the decision-maker evauation for a avaiabe aternatives. In reaity, these steps shoud be foowed by execution and updating. In order to fufi the demand of customers in L ocations, there are severa products P which are produced and grouped into batches during T time periods. The distribution probem is modeed on a fu graph G = (V, L), where V = {1, 2,,n} as the set of vertices, and L = {(,k):, k[l, = k} as the set of arcs. The formuated probem exhibits the foowing characteristics: (a) known homogeneous feet size (each vehice has a capacity indicated by CAP), (b) singe depot (denoted as V o ), (c) deterministic demand (each customer has a certain demand A tp which shoud be deivered by their due dates DD p ), (d) oriented

4 9 Page 4 of 10 Logist. Res. (2016) 9:9 1. MODEL BUILDING A Two-stage Stochastic Programming Formuation Mathematica Mode 2. REQUIRED DATA ETRACTING Ecnomic and Enviromenta Data 3. SCENARIOS GENERATION Impementation Case Study Approach 4. INITIAL SOLUTION GENERATION Exact Agorithm no 5. SCHEDULES ANALYSIS AND MODIFICATIONS 6. SCENARIOS APPROVAL SCHEDULES EECUTING AND UPDATING Fig. 1 Research methodoogy network (n nodes represent geographicay customer ocations L separated by distance D k ), (e) the transportation cost from the depot to these ocations can vary (a unit transportation cost DC k is a function of the vehice traveed distance for deivering the customer orders) and (f) objective: minimizing tota distribution costs. Given this uncertainty, the two-stage stochastic programming mode is deveoped because the mode deas with two eves of decisions. Both stages present the tactica and operationa decisions. The uncertainties are mosty found in the operationa stage since most operationa parameters are not fuy known when the tactica decisions have to be made. The scenario-based approach is used in this paper to mode the probem under uncertainty. The compact form of the stochastic mode can be stated as foows [1, 18]: min z ¼ c T x þ E n min qðxþ T yðxþ ð1þ yes s:t: Ax ¼ b ð2þ TðxÞx þ WyðxÞ ¼ hðxþ ð3þ x 0; yðxþ 0 ð4þ The objective function contains a deterministic term c T x and the expectation of the second-stage objective qðxþ T yðxþ taken over a reaizations of the random event x. The first-stage decisions are represented by the vector x and matrices c, b and A. In the second stage, a number of random events x 2 ( denotes the set of a possibe scenarios and x a particuar scenario) may be reaized. For each x, the vaue yðxþ is the soution of a inear programme. For a given reaization x, the second-stage probem data qðxþ; hðxþ; and TðxÞ become known. Each component of q; h; and T is thus a possibe random variabe. In the deveoped stochastic optimization mode, the deivery veocities and their corresponding greenhouse gas emissions are considered as uncertain parameters. The second-stage decision variabes are E tk ðxþ the tota carbon emissions vaue in route between ocations and k in time period t. Tota produced carbon emissions over (OE tk ðxþ) and under (UE tk ðxþ) are the permissibe imits in route between ocations and k in time period t; E tk ðxþ are defined. Additionay, we define binary decision variabes Z tk ( = k), equa to 1 if and ony if the route f ocation to ocation k wi be used in time period t and 0 otherwise. Given the previous probem definition, the resuting formuation is given in the foowing steps by (5) (20). min DC k D k Z tk þ E x ½Q; ð H; Z; xþ ð5þ t2t p2p Z tk ¼ 1 8 k 2 Lnf1; g; t 2 T ð6þ 2L Z tk ¼ 1 k2l p2p A tkp p2p 8 2 Lnf1; kg; 8 t 2 T ð7þ A tp CAP þ CAP ðz tk þ Z tk Þ 8 t 2 T; 8 k 2 L A tkp CAP ð8þ p2p CAP! Q tkp Z t1k p2p 8 t 2 T; 8 k 2 L p2p A tp p2p Q tp p2p i2lnfg 1 Q tip Z ti 0 8 t 2 T; 8 2 L ð9þ ð10þ A tp CAP 8 2 L; t 2 T ð11þ p2p k2lnf1g Z t1k 0:99 2Lnf1g Q tp CAP 8 t 2 T; 8 p 2 P ð12þ

5 Logist. Res. (2016) 9:9 Page 5 of 10 9 Z tk 2 f0; 1g 8 2 L ð13þ A tp L; 8 t 2 T; 8 p 2 P ð14þ where Q (,H,Z,x) is the optima vaue of the foowing second-stage probem: Min Q; ð H; Z; xþ ¼ ½ðEM k B Þ V tk ðxþþf k OE tk ðxþþg k UE tk ðxþ 2L t2t k2lnfg ð15þ Min V tk ðxþ Z tk V tk ðxþ Max V tk ðxþ Z tk 8 t 2 T; 8 2 L; 8 k 2 Lnfg; 8 x 2 ð16þ 2L EM k V tk ðxþ ¼ E tk ðxþ 8 t 2 T; 8 2 L; 8 k 2 Lnfg; 8 x 2 ð17þ OE tk ðxþ E tk ðxþ Max E Z tk 8 t 2 T; 8 2 L; 8 k 2 Lnfg; 8 x 2 ð18þ UE tk ðxþ Min E Z tk E tk ðxþ 8 t 2 T; 8 2 L; 8 k 2 Lnfg; 8 x 2 ð19þ OE tk ðxþ; UE tk ðxþ; V tk ðxþ 8 t 2 T; 8 2 L; 8 k 2 Lnfg; 8 x 2 ð20þ Equation (5) is the objective function to be minimized incuding the tota distribution costs of distribution across the tactica stage, in addition to the expectation of the second-stage (recourse vaue) probem across the operationa stage. Constraints (6) and (7) ensure that each ocation [except the depot] is visited exacty once in each time period. Furthermore, constraints (8), (9) and (10) generate the optima route sequencing. Moreover, constraint (11) contros the deivering capacity, whie constraint (12) ensures that enough vehices are sent out of the depot. Constraint (13) sets the binary restrictions for the first-stage binary variabes, and constraint (14) isthe non-negativity constraint for a first-stage continuous variabes. The second-stage objective function (15) minimizes the tota costs associated speed and their reated emissions, penaties for overage and underage emissions. Constraint (16) ensures that the chosen veocity at each route in each time period at each scenario is between the reaized minimum and maximum veocities on that route. Constraint (17) cacuates the tota emissions eve of each chosen route. Constraints (18) and (19) cacuate the overage or underage emissions, respectivey, based on the comparison between the actua emissions that woud resut due to the chosen veocity and the standard emissions imits. Constraint (20) is the non-negativity constraint for the secondstage variabes. 3.3 Case studies In order to check the appicabiity of the deveoped mode in practice as identified in Sect. 3.2, a quaitative approach has been used based on a case study anaysis. The mathematica mode is tested in various scenarios of rea-word case studies from batch process industry. The case studies seection approach has been motivated based on the information-oriented process [8]. In this approach, the information content shoud be withdrawn into two phases: preiminary set of companies seection and participation wiingness in research. In the first phase, onine database containing information about mutinationa companies was used. The seection of the initia set of companies was based on the foowing specific criteria: (a) producing in batches, (b) providing different in the range of batches, (c) foster a vocation to socia responsibiity and environmenta-friendy management, (d) invoved in previous research projects in order to faciitate the information fow and (e) their data foow a reguar trend, except during the hoidays. Thus, sampe of those data can be generaized to represent the batch process industry. A ist of nine companies was seected as possibe cases due to these criteria. In the second phase, an enquiry was conducted to secure the acceptance of companies to participate in this research and to check the above isted criteria. Out of five companies agreed to be invoved in the case study investigation, ony two case studies from batch process industry are suitabe for this research. The first case study was conducted in one of the word s eading companies in the fied of foods and beverages processing industry with empoys more than 1500 empoyees. This company is a singe huge pant with seven departments. Currenty, the core business of the company incudes processing and distribution of wide range of dairy and frozen products packaged under more than 40 brand names. Across 62 countries around the word, these products are ordered in batches. In 2005, the company was determined to be one step ahead on socia responsibiity and environmenta-friendy management issues. It started by using new materias during the production processes, importing eco-efficiency methods and increasing empoyee invovement. Recenty, the management has focused on deveoping soutions which consider the environmenta impact of their distribution activities particuary in the fied of reducing emission-reated issues. The second case study was conducted in a mutinationa consumer goods company and manufacturer of product ranges incuding persona care, househod ceaning, aundry detergents, prescription drugs and disposabe nappies.

6 9 Page 6 of 10 Logist. Res. (2016) 9:9 In 2014, it recorded $83.1 biion in saes as profit from more than 80 brands across more than 70 countries a over the word. Since 1999, the company has provided products of superior quaity and vaue that improve the ives of the word s consumers. Their sustainabiity objective is to create industry-eading vaue with consumer-preferred brands and products whie conserving resources, protecting the environment and improving the socia conditions. By 2010, the company has reduced the tota CO 2 emissions by about 14 % per unit of production. Currenty, they focus to make progress by optimizing the distribution routes and invoving economic, environmenta and socia sustainabiity objectives during the panning and distribution processes. In both case studies, demand and suppy foow a reguar trend. Based on the customers demands, the number of required deivery vehices is cacuated. The current distribution strategy is to ship a the requirements of the target customer using homogeneous trucks, singe-size vehices and singe-deivery shipment. A batches are packaged in a standard oad pack before distribution. The unit oad stacked at each paet is 7 kg. Therefore, the amount of each product carried in a paet varies from one product type to another. This amount differs depending on the batch type, dimensions and size. Each vehice is a twenty-foot equivaent unit. The numbers of paets that can fit on a vehice are 11. For the used heavy duty trucks which emit four times more than passenger cars on average [11], GHG emissions vaues are kg CO 2 per vehicemie, g CH 4 per vehice-mie and g N 2 O per vehice-mie at the average veocity (50 mph) (Cimate Leaders GHG Inventory [4]). These vaues demonstrate that CO 2 represents more than 99 % by mass of a the gaseous components of exhaust. Based on these vaues, the emission factor (B ) is estimated for each case study. 4 Impementation, resuts and discussions In this section, we wi discuss resuts obtained from the impementation of the mathematica mode in the case studies. Furthermore, the quaity of the stochastic programming soution wi be compared to those obtained using a deterministic approach. Then, the decision of the deivery veocities and reated CO 2 emissions as environmenta parameter wi be numericay anaysed. In the first case study, the mode was impemented in the iquid department. The seection of this department was based on its good reputation with about 30 % of the market share. Within that department, three types of iquids are produced: ong-ife processed mik, favoured mik and ong-ife juices. Each of these iquids is packaged under different brand names with different voumes. The second case study was carried out in the beauty, hair and persona care iquids section. The deveoped mode was impemented in the shampoo department (of the iquids section) which brought the company about 8 % of its income. A the required characteristics which are reated to demand, routes, depot, distribution nodes and vehices were coected. These characteristics generate the inputs to the mathematica mode as shown in Fig. 2 and Tabe 1. Figure 2 shows that the geographica fied (to scae) of each case study represents the distance between a pairs of nodes incuding the depot (manufacturing faciity). A the possibe connections between nodes have been taken into consideration whie generating the distribution network. Any node can be served directy from the depot or after another deivery points according to the optimum routes and vehice capacity. Moreover, speed imits of these routes were noted and used to cacuate the emission-reated vaues. The mode is soved using the standard sequentia inear programming. As the mode cass is an integer inear programming (ILP), the goba optimum soution is stated using branch and bound agorithm (i.e. the sover type is branch and bound). The mode is inear in both the first and second stages. A computations were run using the LINGO 10.0 optimization package on a machine with CORE TM i5 2.6 GHz processor and 4 GB RAM under Windows 7 Enterprise. For the distribution decisions, distribution pans were generated. These pans were then presented in the form of the optima deivery routes. These routes shoud be driven to distribute the production quantities to the different ocations according to customers needs. The vaues of the binary decision variabe Z tk yieded these distribution routes. The resuts of the second stage were aso optimized. The vaues of the optimum veocity and its reated emission quantities were determined. These vaues are yieded ony when the route is seected when the vehices routing decision variabe vaue Z tk equas one. The main resuts are summarized in Tabe 2. After obtaining the optima soution, the anaysis of the outcomes is performed to show the importance of modeing the presented framework in this way. This anaysis invoves: (a) evauating the usefuness of the stochastic modeing approach instead of a deterministic one and (b) sensitivity anaysis of the veocity because it is the most infuentia mode parameter around the optima vaues. Commony, due to the computationa difficuty of soving the stochastic modes, the usefuness of the stochastic modeing approach is aways questionabe. To check the suitabiity of using stochastic programming, the soutions of the stochastic programming mode to that of a

7 Logist. Res. (2016) 9:9 Page 7 of nd case study 1 st case study L1 L4 L3 L6 L4 L1 Depot L6 Depot L2 L7 L2 L5 L3 L5 Fig. 2 Geographica fied of case studies to scae Tabe 1 Case study characteristics Character First case study Second case study Weeky demand (batches) 14,505 16,6215 Time periods 2 3 Number of batches famiies 7 10 Nodes (incuding depot) 8 7 deterministic optimization probem are quantified. For stochastic modes, when mutipe runs of the modes are averaged out, it sti cannot be concuded that the output wi aso be the same. But after ooking at the definition of the modes and how the resuts are produced, it is possibe to get a better understanding of how modes wi behave under different inputs and different scenarios [24]. In order to obtain the deterministic vaues of the deveoped mode, the stochastic vaues of the deivery veocities and their reated emission vaues are repaced by their average vaues. Therefore, the environmenta objective function and constraints wi be reformuated as foows: min DC k D k Z tk þ t2t p2p 2L V 0tk þ F k OE tk þ G k UE tk k2lnfg ½ðEM k B Þ t2t ð21þ MinV tk Z tk V 0tk MaxV tk Z tk 8 t 2 T; 8 2 L; 8 k 2 Lnfg 2L ð22þ EM k V 0tk ¼ E tk 8 t 2 T; 8 2 L; 8 k 2 Lnfg OE tk E tk MaxE Z tk 8 t 2 T; 8 2 L; 8 k 2 Lnfg UE tk MinE Z tk E tk 8 t 2 T; 8 2 L; 8 k 2 Lnfg OE tk ; UE tk 0 8 t 2 T; 8 2 L; 8 k 2 Lnfg ð23þ ð24þ ð25þ ð26þ Furthermore, two vaues shoud be cacuated: the expected recourse probem (RP) soution vaue and the expected resut of using the expected vaue (EEV). RP is the generated vaue when a mode is soved using the rea vaues. EEV corresponds to one particuar scenario in Tabe 2 Main computationa resuts Character First case study Second case study Tota variabes (integers) 10,702 (177) 18,413 (229) Constraints 13,042 22,408 Number of iterations ti optimaity 945,891 15,065 Objective function vaue 288, ,304.3 Tota number of generated routes 5 16

8 9 Page 8 of 10 Logist. Res. (2016) 9:9 Tabe 3 VSS resuts for different tested scenarios Number of scenarios RP vaue EEV vaue VSS vaue (VSS/RP) % First case study , ,797 37, , ,230 31, , ,840 26, , ,590 18, , ,341 14, Second case study , ,680 48, , ,780 39, , ,655 35, , ,368 24, , ,360 21, which the corresponding expected vaue (EV) repaced the observation of the random vaues. The difference between the RP and EEV is defined as the vaue of the stochastic soution (VSS). The VSS is the concept that precisey measures the quaity of a decision [1]. To dispay these vaues, datasets of 10, 20, 30, 40 and 50 scenarios were tested. Tabe 3 shows the vaues of running these numbers of scenarios and their corresponding running time. For each dataset, the RP, EEV and VSS vaues are cacuated. Moreover, the VSS as a percentage of the overa RP soution is reported. Since a VSS have positive vaues, deaing with uncertainty reay matters. Therefore, optima soutions are sensitive to the vaue of the random eements. The difference between the optima soution (generated from the stochastic modeing) and the expected soution (generated form the deterministic modeing) is between 13 and 4 percentages for datasets of 10, 20, 30, 40 and 50 scenarios. The VSS vaues verify the significance and benefits of using the stochastic modeing approach. In the deterministic mode, the average deivery veocity (km/h) was 75 and the average emission factor (kg CO 2 /km/h) was 1.2 in both case studies. Moreover, the deivery veocity of the base in the second stage in the deveoped stochastic mode is examined. The generated emissions during distribution depend mainy on the deivery veocity as mentioned earier. The optimum transporting veocity vaues were generated in the case studies. When the required batches are deivered faster than the optimum, more emissions are generated which is directy proportiona to emission costs. Tabe 4 shows the effect of changing the deivery veocities in the distribution phases on the tota cost. Resuts show that the change of the veocity negativey or positivey resuts in the increase in the vaue of the tota cost. However, decreasing the veocity parameter increases the costs at a higher rate more than the effect of increasing this factor. The optima soution guarantees that the output schedue generated with the east cost vaue. The concusion is that the optimum soution is more sensitive to decreasing the deivery veocity time than increasing it. Tabe 4 Change of objective cost vaue in response to change of veocity Veocity changing (%) The tota cost surges when the veocities fa due to rise in the ateness penaty. On the other hand, when the veocities increase, the tota cost escaates due to heightened eve of the emission penaty. In sum, the resuts of the mode anaysis show that it can generate the optima soution for the batch process industry. This is verified and successfuy impemented in two rea-ife industria case studies that prove the mode s capabiity as a decision support system which addresses both routing and scheduing distribution decisions. Using the previous scenarios, the fexibiity of the proposed integrated green mode is proved whie taking into consideration most of the specia features in the batch process industry. 5 Concusions % of the environmenta/tota cost Sensitivity coefficient in the environmenta cost Sensitivity coefficient in the tota cost The main goa of this study is to show the potentia of the introduction of a green ogistics mode in batch process industry. The proposed mode can be considered a decision

9 Logist. Res. (2016) 9:9 Page 9 of 10 9 support too for distribution panning. It heps the decisionmaker in deivering decisions on both tactica and operationa eves. The advantage of using the deivery veocities vaues and the reated emissions has been proved. On the one hand, the importance of optimizing transport panning was shown in cost and emission savings. Based on the resuts, a batch process faciity improves its efficiency objectives from economic and ecoogica perspectives. On the other hand, the optima transporting veocity ranges shoud be stricty observed during ogistics physica activities. Due to the uncertain dynamic environment, controing deivery veocities is not that easy in practice. For these reasons, extra costs must be incurred in training and education of human resources who are invoved in the distribution processes in order to increase their eve of green awareness and knowedge. This coud be studied in depth for further research. Furthermore, the proposed mode supports academic green ogistics modes and rea-word suppy chain decision making in batch process industry. Deveoping and impementing the proposed green ogistics mode provide a practica too which inks being green and being economicay successfu. Within the gobaized market, greening the batch suppy chain through an environmentay friendy ogistics network design guarantees competitive advantages and meets the customers expectations and socia responsibiities requirements. In addition, the mode can aso be appied to other companies in order to assess the accuracy of the concusions reached in this paper. Further research may aso ead to finding ceaner routes, through the deveopment of exact or approximate agorithms. Additionay, adding other sources of uncertainties within the distribution phases wi positivey enrich the credibiity of this mode. In this dynamic environment, uncertainties are represented by the gap between the panned and actua system status. This gap shows a new chaenge for the management of compex and dynamics suppy chain processes. Finay, defining a useabe way to obtain rea measures for CO 2 emissions wi highy improve the accuracy of the resuts and therefore the decision of the driver during deivery. There are many prospects for further research on this work. One is invoving the green aspects in the production stage, rather than just in the distribution phase. A second addition coud be using different emission estimation ways and comparing them atogether. A third extension is adding other sources of uncertainty, such as the production costs and capacities. A fourth and fina direction coud be cacuating the effect of considering the socia criteria such as improving abour conditions and human rights. Acknowedgments The authors appreciate the support of Ms. Yosra Abdekhaek, Product Suppy Speciaist, Mondeēz Internationa (Kraft Foods), and Mr. Esam Saeed Mekky, Suppy Network Operations Manager, Procter & Gambe, for providing us the case study data. We woud aso ike to thank Dr. Ay Megahed, IBM Research Amaden Research Center for his vauabe technica support, hep and advice. Specia thanks to Professor Yahaya Yusuf, University of Centra Lancashir, for his academic support. This research was financiay supported by the Deutscher Akademischer Austausch Dienst (DAAD) and the Egyptian Ministry of Higher Education under Grant GERLS The authors woud ike to thank Internationa Graduate Schoo for Dynamics in Logistics (IGS) in University of Bremen and the anonymous reviewers. Funding This study was funded by the Deutscher Akademischer Austausch Dienst (DAAD) and the Egyptian Ministry of Higher Education under Grant GERLS Compiance with ethica standards Confict of interest The authors decare that they have no confict of interest. Open Access This artice is distributed under the terms of the Creative Commons Attribution 4.0 Internationa License ( tivecommons.org/icenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the origina author(s) and the source, provide a ink to the Creative Commons icense, and indicate if changes were made. References 1. Birge JR, Louveaux F (2011) Introduction to stochastic programming, 2nd edn. Berin, Springer 2. Bueher R, Pucher J (2011) Sustainabe transport in Freiburg: essons from Germany s environmenta capita. Int J Sustain Transp 5: Cari T, God H (2008) Vehice routing probem. ISBN , InTech 4. Cimate Leaders GHG Inventory Protoco (2008) Optiona emission from commuting, trave, and product transport. Office of air and radiation. United State Environmenta Protection Agency. Accessed 03 Aug Dekker R, Boemhof J, Maidis I (2012) Operations research for green ogistics an overview of aspects, issues, contributions and chaenges. Eur J Oper Res 219(3): Demir E, Bektaş T, Laporte G (2014) Invited review a review of recent research on green road freight transportation. Eur J Oper Res 237(3): Egese R, Bektaş T (2014) Vehice routing: probems, methods, and appications; green vehice routing. Society of industria and appied mathematics: MOS-SIAM series on optimization, 2nd edn, Fyvbjerg B (2001) Making socia science matter. Cambridge University Press, Cambridge 9. Gao S, Qi L, Lei L (2015) Integrated batch production and distribution scheduing with imited vehice capacity. Int J Prod Econ 160: Ha NG, Potts CN (2005) The coordination of scheduing and batch deiveries. Annu Oper Res 135(1 4): Kean AJ, Harey R, Kenda G (2003) Effects of vehice speed and engine oad on motor vehice emissions. Environ Sci Techno 37(17):

10 9 Page 10 of 10 Logist. Res. (2016) 9:9 12. Kopfer HW, Schönberger J, Kopfer H (2014) Reducing greenhouse gas emissions of a heterogeneous vehice feet. Fex Serv Manuf J 26(1 2): Kumar S, Teichman S, Timpernage T (2012) A green suppy chain is a requirement for profitabiity. Int J Prod Res 50(5): Maden W, Egese RW, Back D (2010) Vehice routing and scheduing with time-varying data: a case study. J Oper Res Soc 61: Min H, Kim I (2012) Green suppy chain research: past, present, and future. Logist Res 1(4): Paaniveu P (2010) Green ogistics. White paper. TATA Consutancy Services Limited (TCS), pp Park YB (2005) An integrated approach for production and distribution panning in suppy chain management. Int J Prod Res 43(6): Pishvaee MS, Joai F, Razmi J (2009) Technica paper: a stochastic optimization mode for integrated forward/reverse ogistics network design. J Manuf Syst 28(4): Pishvaee MS, Torabi SA, Razmi J (2012) Credibiity based fuzzy mathematica programming mode for green ogistics design under uncertainty. Comput Ind Eng 62: Qian J, Egese R (2016) Fue emissions optimization in vehice routing probems with time-varying speeds. Eur J Oper Res 248(3): Srinivasu R (2014) Fast moving consumer goods retai market, growth prospect, market overview and food infation in Indian market an overview. Int J Innov Res Sci Eng Techno 3(1): Srivastava SK (2007) Green suppy chain management: a state of the art iterature review. Int J Manag Rev 9(1): Ubeda S, Arceus FJ, Fauin J (2011) Green ogistics at Eroski: a case study. Int J Prod Econ 131: Van Dam KH, Adhitya A, Srinivasan R, Lukszo Z (2009) Critica evauation of paradigms for modeing integrated suppy chains. Comput Chem Eng 33(10):