An Integrated Approach for the Optimization of the Sustainable performance: a Wood Supply Chain

An Integrated Approach for the Optimization of the Sustainable performance: a Wood Supply Chain

7th IFAC Conference on Manufacturing Modelling, Management, and Control International Federation of Automatic Control June 19-21, 2013. Saint Petersbu...

413KB Sizes 9 Downloads 135 Views

7th IFAC Conference on Manufacturing Modelling, Management, and Control International Federation of Automatic Control June 19-21, 2013. Saint Petersburg, Russia

An Integrated Approach for the Optimization of the Sustainable performance: a Wood Supply Chain T. Boukherroub*, A. Ruiz** A. Guinet*, J. Fondrevelle* *Laboratoire DISP (Décision et Information pour les Systèmes de Production) INSA-Lyon, campus Lyon Tech Bât Léonard de Vinci, 21 avenue Jean Capelle, 69621 Villeurbanne cedex – France (Tel: 33-472-437-536; e-mail: [email protected], [email protected], [email protected]) **Centre interuniversitaire de recherche sur les réseaux d’entreprise, la logistique et le transport (CIRRELT) Université Laval, 2325, rue de la Terrasse, Québec (Québec) G1V 0A6 – Canada (e-mail: [email protected]) Abstract: This paper proposes an integrated approach that embeds the economic, environmental and social performances in the planning activities of the supply chain (SC). The approach is applied on a realistic case study inspired by the lumber industry where divergent manufacturing processes and various cutting patterns are involved. We first identify the sustainability objectives to be measured. Then, we link these objectives to the SC decision planning. Next, we define performance measures to assess the achievement of each objective. This triptych-based approach is transposed to a multi-objective mathematical programming (MOP) that serves as a performance optimizing and assessing tool. The MOP models a tactical planning problem where the SC is represented as a network of activities. The problem is resolved using the weighted sum method. Keywords: wood industry, divergent processes, cutting patterns, supply chain planning, sustainability performance, activity network, optimization, multi-objective mathematical programming, weighted sum. 1.

INTRODUCTION

(Boukherroub et al. 2012b). We found that only 30% of them took into account the environmental dimension and less than 10% considered the social dimension. Obviously, there is a lack in the literature regarding integrated approaches considering the three dimensions of sustainability performance when planning the SC. Moreover, D'Amours et al., (2008) concluded in their literature review that wood SC planning problems involving divergent processes are poorly studied by researchers.

In recent years, efficiency has been identified as the only way for the survival of Canadian forest companies. However, although the efficiency of operations is at the heart of the success of any supply chain (SC), a broader approach integrating environmental and social aspects could provide other avenues of research in the forest product sector, leading to sustainable success. The recognition of the role of the forest in both economic, ecological and social dimensions (wealth creation for the forest industry and local authorities, carbon sequestration, outdoor activity space etc.) on the one hand, and the consequences of the slowdown on the forest industry that occurred in Canada in recent years (plant closures, job losses, weakening of rural communities, etc.) on the other hand, led society, customers, investors, government, and other stakeholders to require a sustainable management of forest product SCs.

Taking these two facts into account, this paper focuses on the sustainable performance optimization of a lumber production SC. Our objective is twofold: (1) integrate all the three dimensions of sustainability performance when planning the SC, and (2) study a SC wood industry presenting divergent manufacturing processes. To this end, we first identify the sustainability objectives to be measured. Then, we link these objectives to the SC decision planning. Next, we define performance measures to assess the achievement of each objective. This triptych-based approach is then transposed to a multi-objective mathematical programming (MOP) that serves as a performance optimizing and assessing tool. The MOP models a tactical planning problem where the SC is represented as a network of activities. The remainder of the paper is organized as follows: the lumber industry context is introduced in the next section. In section 3 we present our integrated approach. Section 4 presents our experimentation and reports preliminary results. Finally, we draw some conclusion and give insights on our future work in section 5.

Recent studies suggest that a strategy based on corporate social responsibility (CSR) is a differentiating factor (Li and Toppinen, 2011; Boukherroub et al. 2012a) and its implementation would improve the company relationships with its stakeholders while improving its profitability (Carroll and Shabana, 2010). Yet, studies quantifying the impact of such an approach and analyzing the relationship between environmental protection, social welfare and economic viability in an integrated way are quite rare in the literature. Indeed, we conducted a systematic literature review in the field which led us to examine a total of 62 papers 978-3-902823-35-9/2013 © IFAC

186

10.3182/20130619-3-RU-3018.00205

2013 IFAC MIM June 19-21, 2013. Saint Petersburg, Russia

2.

THE LUMBER INDUSTRY CONTEXT

The lumber industry is a key business sector in Canada generating thousands of direct employments. 80% of the wood harvested in Canada is used to produce lumber. Sawmills play an important role in the value chain of forest products since they are often holders of the CAAF, the license that allows to source wood from the public forests (which represents more than 80% of the forests in Canada). Pulp and paper mills and many other second and third transformation factories (pellet-mills, panel-mills, cogeneration plants, etc.) are mainly supplied from sawmills. The lumber SC starts from the forest. Trees are harvested (i.e. they are cut and branches are removed) and then bucked into logs of specific dimensions. Bucking activity could also be done in lumber yards. Harvesting is often done in two ways: by direct labor (in-house) or under outsourcing contract (forest contractors). Logs once transported to the sawmill, are sawn and planks of different dimensions are obtained (green lumber). These are then dried and the resulting finished products (lumber) are shipped to the customers (located in Canada and USA). During bucking and sawing, wood residues are generated. These are chipped and the resulting flakes are sold mainly to pulp and pellet mills. The commonly modes of transportation used to transport the various products from one business unit to another in the SC are trucks, trains and ships. Bucking and sawing are two key manufacturing activities in sawmills since from one product, a multitude of output products are obtained. Such processes are divergent, a characteristic that lumber industry shares with meat or metal sheet industries for example. In addition, several cutting patterns (recipes) could be applied to obtain different mix of output products. Choosing a cutting pattern is crucial as it specifies the dimensions and quantity of output products on the one hand, and how the input product is consumed on the other hand. Indeed, customer demand should be satisfied while optimizing the raw material utilization. Different technologies could be used for bucking and sawing, each of them presenting a set of applicable cutting patterns. 3. AN INTEGRATED APPROACH FOR THE OPTIMIZATION OF A LUMBER SUPPLY CHAIN Our approach encompasses a triptych “objectives-correlationmeasures” (see Fig. 1-I): which aims at (1) identifying the objectives of sustainability to be integrated to the SC planning, (2) correlating these objectives with the SC planning decisions, and (3) setting the performance indicators to measure the achievement of each objective. The triptychbased approach is independent of the SC planning level, and different modeling tools such as mathematical programming or simulation can be implemented into our approach to evaluate the performance of the SC. Different methods can be used to model the production system (material balance or activity-based approaches) as well. In this paper, we represent the production system as a network of activities (Fig.1-II) (Lakhal et al., 2001). We also use the multiobjective mathematical programming (MOP) for optimizing and evaluating the SC sustainability performance (Fig.1-III). 187

Fig. 1. The proposed integrated approach 3.1. The triptych “objectives-correlation-measures” • Sustainability objective identification: To identify relevant objectives, we relied on the scientific literature and international standards such as SCOR model, OECD guidelines, GRI, ISO 26000, etc. Inspired by the work of Baumann (2011), we selected 12 objectives to cover the three dimensions of sustainability (Table 1): 5 objectives related to the economic dimension, 4 to the environmental dimension and 3 objectives related to the social dimension. • Objective correlation with the SC planning decisions: Once the objectives were set, their links to the planning decisions of the SC were analyzed. At the strategic level, for example, the location of a production site could impact the production cost, the delivery time, the greenhouse gas (GHG) emissions, and even the employment level in the region. At the tactical level, closing a business unit during one or more periods could affect the level of employment and responsiveness. At the operational level, the choice of a scheduling plan could impact cost, delivery, flexibility, and stability of employment. • Measurement of the achievement of the objectives: To assess the achievement of each objective (i.e. performance evaluation), performance indicators must be defined. Sustainability performance indicators can be found in the scientific literature. Table 1. Sustainability objectives (adapted from Baumann, 2011) Dimension Objective 1. Improve financial performance Economy 2. Improve responsiveness (Eco) 3. Improve flexibility 4. Improve reliability 5. Improve quality 1. Reduce the use of resources Environment 2. Reduce GHG emission (Env) 3. Reduce pollution 4. Manage hazardous materials 1. Preserve health and security of Society (Soc) employees 2. Create job and wealth 3. Ensure good conditions of work

2013 IFAC MIM June 19-21, 2013. Saint Petersburg, Russia

- Wood harvesting can be done by direct labor (see section 2) or subcontracted during one or more periods if the internal capacity is insufficient or subcontracting costs less. Nevertheless, part of harvesting could be done “in-house” even the activity is outsourced. - Forest contractors are the usual partners of the company. Therefore, fixed and binary variables related to supplier selection are not required. - Transport is done by third-party logistics providers (3PL). The modes of transportation as well as the potential carriers are known and selected before the beginning of the planning horizon. Fixed costs of transport and binary variables related to carrier selection are thus omitted.

3.2. SC modeling as an activity network Since the nature and amounts of input and output products depend on the technology type, the material requirements cannot be determined without prior knowledge of the manufacturing process being used. This method refers to activity-based approach (Hammami et al., 2009). In the lumber industry, as mentioned before, different technologies can be used for bucking and sawing. In addition, two distinct technologies may consume different quantities of energy or labor, and generate different amounts of CO2 and other emissions. We associate to each technology a single activity. By doing so, we define a bijective relationship as in (Hammami et al., 2009). We therefore consider as many distinct activities as available technologies and associate to each activity a set of multiple cutting patterns.

• Sets of indexes : set of all products.      set of raw materials (i.e. trees). : set of processed products.        : set of intermediate products.  : set of finished products (lumber and flakes).   : set of outsourced products (logs and stems). : set of forests procuring trees    (as the forests are owned by the state, an acquisition cost is required to harvest trees from the forest, modeled here as a supplier). : set of potential forest contractors procuring stems and logs    : set of sawmills and harvesting areas. : set of customers. : set of transportation modes.           .  : set of transportation modes from forests to harvesting areas (although harvesting is done in the forest, we consider in our formulation a harvesting area as a site sourcing raw materials from the forest).  : set of transportation modes from forest contractors to sawmills. set of transportation modes between sawmills. : set of transportation modes from sawmills to customers. : set of processing activities.  set of activities implemented in sawmill or harvesting area . : set of recipes (cutting patterns for bucking and sawing).: set of recipes associated to activity a. : set of activities using recipe .   : set of products on which recipe could be applied  : set of products      used to obtain the products   .: set of recipes applicable to the product      .  : set of recipes used to obtain the product    R: set of residency regions of employees. : set of residency regions where the harvesting area or sawmill  could hire its employees. !: set of sawmills and harvesting areas hiring in the residency region!. "  #$% &% ' % (: set of the horizon panning periods.

3.3 Multi-objective mathematical programming MOP is an extension of traditional mathematical programming adapted to the case where multiple objective functions need to be optimized simultaneously. MOP resolution leads to a multitude of solutions. Among them, a restricted set contains Pareto optimums; solutions that cannot be improved in one objective without deteriorating their performance in at least one of the rest. The solution to be selected by the decision maker (DM) will reflect the compromise he/she wishes to make. We transpose our triptych-based approach to MOP as follows: 1. We select three sustainability objectives from Table 1.: (Eco.1) cost reduction for the economic dimension, (Env.2) GHG emission reduction for the environmental dimension, and (Soc.2) local job creation for the social dimension. 2. We link each objective to the MOP decision variables (SC planning decisions). The correlation could be materialized by different factors such as unit costs, GHG emission factors, etc. associated to the decision variables. 3. We set three objective functions as performance indicators to measure each objective of sustainability. An important advantage of using MOP is its ability to “optimize” (Pareto optimums) the performance from all the three dimensions of sustainability besides measuring its value. The next paragraphs present the MOP we developed for the lumber industry SC planning problem. We focused on the tactical level. The problem is to optimize the use of SC resources (materials, machine, labor, storage capacity, etc.) in order to meet the customer demand at the lowest cost while reducing GHG emissions and promoting local employment. We consider multiple forests and potential forest contractors, multiple sawmills, and various customer zones. Different transportation modes are also considered.

• Decision variables )*+,-  volume1 of product  to be consumed by activity  using recipe  in the harvesting area or sawmill  at period . .*-  inventory of the product p to be consumed at sawmill i at the end of period  / * -  volume of the product (tree)  procured from forest 0by harvesting area  at period .

• Hypotheses of the mathematical formulation - The SC operations are managed by a single entity according to “make-to-stock” policy (MTS). - The planning horizon is divided into multiple periods. One period corresponds to a season of the year (a quarter) as lumber demand in Canada follows a seasonal behaviour.

1

The quantities are expressed in terms of volume (m3 for trees, stems, logs, and rejects, board foot for green lumber and lumber, and metric ton for flakes).

188

2013 IFAC MIM June 19-21, 2013. Saint Petersburg, Russia

2 "?*4: unit cost by km. of transporting intermediate products  between harvesting area or sawmill  and sawmill  using transportation mode 3 at period . 2 : unit cost by km. of transporting lumber or flakes "?*6from sawmill  to customer 7 using transportation mode 3 at period . 8- : unit labor cost at period . A,* : GHG emission factor of activityper unit of volume of the product  processed (CO2 equivalent kg). B2* : GHG emission factor of transportation mode 3 per unit of weigh of the product transported and per kilometer travelled (CO2 equivalent kg /kg.km). C: GHG emission factor per employee per kilometer travelled (CO2 equivalent kg /km). 2  distance between forest 0 and harvesting area  when D  using transportation mode 3. 2  distance between forest contractor  and sawmill  when D  using transportation mode 3. D2 E  distance between sawmill  and sawmill  when using transportation mode 3. 2 D6  distance between sawmill  and customer 7 when using transportation mode 3. D:  distance between the region ! and harvesting area or sawmill . F*  volume of one unit of the product  (in m3). G* : weigh of one unit of the product  (in kg)

/ * -  volume of the product (stem and log)  sourced from forest contractor  by sawmill  at period . 1 2* - : volume of the product (tree)  carried from forest 0 to harvesting area  using transportation mode 3 at period . 1 2* - : volume of stems and logs  carried from forest contractor  to sawmill  using transportation mode 3 at period . 2 : volume of product  carried from harvesting area or 1*4sawmill  to sawmill  4  5  using transportation mode 3 at period . 2 1*6: volume of product  carried from sawmill  to customer 7 using transportation mode 3 at period . 89:-  number of employees living in region ! and working at the harvesting area or sawmill  at period (approximated by a continues variable). • Parameters ;*6-  demand of customer 7 for product  (lumber or flakes) at period . ,- : capacity of activity  (in hours) at harvesting area or sawmill  at period . .- : storage capacity of sawmill at period . <  minimum inventory of product  at sawmill  at the end .*of period  .* : initial inventory of product  at sawmill . (in

* -  capacity of forest 0 for tree  at period volume).

* -  capacity of forest contractor  for product  at period (in volume).   capacity of activity  (in hours) required to transform =*+, one unit volume of the product  using recipe . 8, : number of man-hours required for 1 machine-hour of activity . >+,*4* : volume of product processed by activity  using recipe  from 1 unit of volume of the product 4  89:  employment capacity in region !at period . 89:-  initial number of employees living in region ! and working at the harvesting area or sawmill  8?- : employee capacity during the period t. @ 2 : transport capacity of the mode 3 in volume (m3).  2 : transport capacity of the mode 3 in weigh (kg). *,- : unit cost of processing the product  by activity  at the harvesting area or sawmill  at period . .*- : unit handling cost of the product  at the sawmill  at period . * - : unit cost of sourcing trees  from forest 0at period . * - : unit cost of sourcing stems or logs  from forest contractor  at period . 2 "?* : unit cost of carrying trees  (per kilometer travelled) from forest 0 to harvesting area  using transportation mode 3 at period . 2 "?* : unit cost of carrying stems or logs  (per kilometer travelled) from harvesting area 0 to sawmill  using transportation mode 3 at period .

• Objective functions The economic objective is “optimized” and measured using  H :

H  I-JIPI,O  I+NK,, I*K+ LM+ *,-  )*+,-  Q8- I:RP 89:- Q I*K .*-  .*-  Q I T I*RS * - IP / * - Q I TN I*USKVW * - IP / * - Q 2 2 2 I*RS "?* I2XTP I P IP D   1 * Q 2 2 2 I2XTNT I TN IT D  I*USKVW "?* -  1 * Q 2 2 2 I2XPP IP I E P%4Y D4 I*SKVW "?*4-  1*4- Q 2 2 2 I2XPN IP I6N D6 I*SKW "?*6 1*6Z

(1)

The environmental objective is “optimized” and measured using  [ :

[   I-JIPI,O  I+NK,, I*K+ LM+ A,*  )*+,- Q C I:RP D: 89:-  Q 2 2 I2XTP  I T IP D  I*RS B2*  G*  1 * Q 2 2 I*USKVW B2*  G*  1 * I2XTNP  I TN IP D 

2 2 I*SKVW B2*  G*  1* Q I2XPP  IP I E P%Y4 D4 E- Q 2 2 I2XPN  IP I6N D6 I*SKW B2*  G*  1*6- Z

(2)

The social objective is “optimized” and measured using  \ :

\    I- IP I:RP D: 89:(3) • Constraints  I+NK,, I*K+ LM+ =*+,  )*+,-  ]  ,-  ^  % ^  % ^  "

189

(4)

2013 IFAC MIM June 19-21, 2013. Saint Petersburg, Russia

^0  %^  %^3   %^  " I* F*  1 2* -  ] @ 2 ^  %^  %^3   %^  " I* G*  12*6-  ]  2 ^  %^7  %^3  %^  " I* G*  12*4-  ]  2 ^  %^ 4  %  5  4 % ^3  % ^  " I* G*  1 2* -  ]  2 ^0  % ^  %^3   %^  " I* G*  1 2* - ]  2 ^  %^  %^3   %^  " )*+,- ,.*- % / * - , 2 2 2 2 / * - % 1 * % 1 * % 1*4% 1*6% 89:- h _

)*+,-  _^  ^  `% ^  %^    %^  " (5) I*K F*  .*- ]  .^  % ^  " (6) < ] .*^  %^  %^  " (7) .*I:RP 89:-   aI,O  8, I+NK,, I*K+ LM+ =*+,  )*+,- bc8?-  ^  %^  " (8) I:R`RP 89:-  _ ^  % ^  " (9) IPR: 89:- ]  89: ^!  %^  " (10) IP / * - ]  * - ^  % ^0  % ^  " (11) IP / * - ] * - ^   % ^  % ^  " (12) 2 I2XPN IP 1*6 ;*6- ^7  % ^   % ^  " (13) I+RNKef-* I,O dONK+ I*ESKLM * >+,*E*  )*E+,H Q 2   I2XPN I6N 1*6H Q .*H  .* (14) ^   % ^   I+NKef-* I,O dONK+ I*E SKLM * >+,*4*  )*E+,- Q 2 Q .*-  .*-gH  I2XPN I6N 1*6^   %^  % ^  "`$ (15) 2 I2XPP I E P%4Y 1*4H Q Q I+NKef-* I,O dONK+ I*ESKLM * >+,*E*  )*E+,H Q  I+NK* I,O dONK+ )*+,H  Q .* 2 I2XPP I4Y 1*4H Q.*H (16) ^    `  %^   2 I2XPP I E P%4Y 1*4Q Q I+NKef-* I,O dONK+ I*ESKLM * >+,*E*  )*E+,- Q .*-gH  I+NK* I,O dONK+ )*+,-  Q 2 Q.*I2XPP I4Y 1*4^    `  %^  %^  "`$ (17) 2 I2XPP I E P% E Y 1* E H Q I TN / * H Q I+NKef-* I,O dONK+ I*E SKLM * >+,*E*  )*E+,H  Q   I+NK* I,O dONK+ )*+,H Q .* 2 I2XPP I4Y 1*4H  Q .*H  ^   %^   (18) 2 I2XPP I E P% E Y 1* E - Q I TN / * - Q I+NKef-* I,O dONK+ I*E SKLM * >+,*E*  )*E+,-  Q .*-gH  I+NK* I,O dONK+ )*+,- Q 2 I2XPP I E Y 1* E - Q .*-  ^   %^  %^  "`$ (19) I T / * H Q  .* =I+NK* I,O dONK+ )*+,H Q (20) .*H ^  %^   I T / * - Q  .*-gH =I+NK* I,O dONK+ )*+,- Q .*- ^  %^  %^  "`$ (21) / * -   I2XTP 1 2* ^  %^0   %^  % ^  " (22) / * -   I2XTNP 1 2* -  (23) ^   %^   %^  % ^  " I* F*  12*6-  ] @ 2 ^  %^7  %^3  %^  " (24) I* F*  12* E - ] @ 2  ^  %^ 4  %  5  4 % ^3  %^  " (25) 2 2 I* F*  1 * -  ] @ 

(26) (27) (28) (29) (30) (31) (32)

The set of constraints (4) states that production capacity is limited. Constraints (5) indicate that products can be consumed by an activity only at sites where it is implemented. Constraints (6) specify that the storage capacity is limited, and constraints (7) state that there must be a minimum stock at the end of each period. Equations (8) define the relationship between the amount of consumed products at a site and the number of employees in that site. Constraints (9) specify that production facilities cannot employ people in some regions (too far from the sites). Constraints (10) specify that the number of employees per region is limited. Constraint sets (11) and (12) reflect the fact that the capacity of the forests and forest contractors (respectively) is limited. Constraints (13) specify that the demand for all finished products should be satisfied. The sets of equations (14) to (23) ensure the flow product conservation. Since the manufacturing processes are divergent, we focused on the consumption of input products to describe the material flow and its conservation through the SC. Sets of constraints (24) to (27) limit the volume transport capacity of all modes while the sets of constraints (28) to (31) limit the capacity in terms of weight. Finally, constraints (32) ensure that all decision variables are continuous and positive. 4.

NUMERICAL RESULTS

To illustrate our approach, we used a realistic case inspired by “virtu@l-lumber” developed by Vila et al. (2006) jointly with the three biggest players in the Canadian lumber industry, two Canadian research centers on forestry (FOR@C and Forintek) and the Ministry of Natural Resources and Fauna of Quebec. The case covers 6 periods, each period lasting 3 months. The SC includes 3 forests, 2 potential forest contractors, 3 sawmills, and 5 customers (1 paper mill for flakes, and 4 industrials for lumber). We considered 2 transportation modes; train and truck, 1 bucking technology presenting 3 cutting patterns, 2 sawing technologies with respectively 5 and 4 cutting patterns, 2 drying technologies, 1 harvesting technology, and 1 shipping technology. We solved the MOP using the weighted sum method. As we associated a GHG emission factor to employees, the objective function ( \ ) is omitted in our experimentation. This objective will be examined in details in our future work. We associated two weights iH % i[  to the economic and 190

2013 IFAC MIM June 19-21, 2013. Saint Petersburg, Russia

environmental objectives such that iH Q  i[ = 1, and then aggregated the two in one single objective. We ranged the weight values between 0 and 1 (step of 0.1) such that to each pair (iH % i[ ) corresponds a “compromise” solution. We ran our model using CPLEX 12.2 on an Intel-Pentium M PC with 1.73 GHz processor and 2 GB RAM. Less than 30 seconds were required to solve the MOP. The preliminary results are shown in Fig.2.

H : Cost (107 $)

objective identification, the correlation of SC planning decisions with the objectives, and performance indicator definition. It presents the advantage of being independent from the SC planning level and the performance evaluation tool. The MOP allows the SC performance optimization besides measuring it. We finally demonstrated the solvability of the MOP using the weighted sum method. One limitation of our work is tied to the social performance analysis. In our future work, we will conduct more experiments and emphasise more on the social dimension, which is a very important aspect. On the other hand, we will analyse more in detail sustainability objective correlation with SC decisions (decision variables) in order to improve the operationalization of sustainability when dealing with SC planning problems while ensuring coherency. We will also generalize our approach to other industries (e.g. convergent manufacturing processes). ACKNOWLEDGEMENTS Funding for this project was provided by la Région RhôneAlpes and the consortium of research FOR@C.

[ : GHG emission (107 CO2 equivalent kg)

Fig. 2. Representation of Pareto optimums REFERENCES Each solution corresponds to both a monetary and a socioenvironmental performance levels. It is thus possible for the DM to choose the solution that reflects the best the compromise he/she wishes to operate. In Fig. 2, two particular solutions, corresponding to two extreme situations, can be remarked. The first corresponds to the situation where only the economic objective is considered (iH  $% i[  _) while the second one considers only the environmental (and social) objective (iH  _% i[  $). Apart from the 2 extreme solutions, we obtained environmental performance levels close to the optimum without paying more than 5.92% of the minimum possible cost. When we set iH to 0.9 and i[ to 0.1, the environmental performance is significantly improved while the economic one is slightly deteriorated (0.54% from the optimum). This solution can therefore be interesting for the DM as a slight increase in cost makes a significant improvement in environmental performance. Similarly, the weights (0.1, 0.9) deteriorate slightly the optimum value of the environmental performance (0.02%) while the economic one is significantly improved (the deterioration becomes 5.92% instead of 12.03%). All other weights lead to tradeoffs more or less “well balanced”. The solution to be chosen by the DM will depend on his/her level of involvement in corporate social responsibility (CSR) strategy. 5.

Baumann E., (2011). Modèles d’évaluation des performances économique, environnementale et sociale dans les chaînes logistiques. Thèse de doctorat, Institut National des Sciences Appliquées de Lyon, France. Boukherroub, T., Guinet, A., and Fondrevelle, J. (2012a). Méthode d’aide à la décision multicritères pour l’internalisation/externalisation “durable”. 9th International Conference on Modeling, Optimization, and Simulation, June 6-8, Bordeaux (France). Boukherroub, T., Fondrevelle, J., Guinet, A., and Ruiz, A. (2012b). Multi-criteria decision making for the supply chain design: A review with emphasis on sustainable supply chains, 4th International Conference on Information Systems, Logistics and Supply Chain, August 26-29, Quebec (Canada). Carroll, A.B., and Shabana, K.M. (2010). The business case for corporate social responsibility: A review of concepts, research and practice. International Journal of Management Reviews, 12(1), p. 85-105. D’Amours, S., Rönnqvist, M., and Weintraub, A. (2008). Using Operational Research for Supply Chain Planning in the Forest Products Industry. INFOR: Information Systems and Operational Research, 46(4), p. 265-281 Hammami, R., Frein, Y., and Hadj-Alouane, A.B. (2009). A strategic-tactical model for the supply chain design in the delocalization context: Mathematical formulation and a case study. International Journal of Production economics, 122, p. 351-365. Lakhal, S., Martel, A., Kettani, O. and Oral, M. (2001). On the optimization of supply chain networking decisions. European Journal of Operational Research, 129(2), p. 259-270. Li N., and Toppinen, A., (2011). Corporate responsibility and sustainable competitive advantage in forest-based industry: Complementary or conflicting goals? Forest Policy and Economics, 13, p. 113-123. Vila, D., Martel, A., and Beauregard, R. (2006). Designing logistics networks in divergent industries: A methodology and its application to the lumber industry. International Journal of Production Economics, 102(21), p. 358-378.

CONCLUSION AND PERSPECTIVES

This paper addressed the optimization of the sustainability performance of a lumber product SC. Two challenging features were tackled: (1) divergent manufacturing process modelling and (2) sustainability performance integration to the SC planning. We proposed an integrated approach that models the SC as a network of activities and transposes a triptych-based approach for sustainability performance evaluation to a multi-objective mathematical programming (MOP). Our triptych-based approach includes sustainability 191