RATIO MATHEMATICA 26 (2014), 39–64 ISSN:1592-7415 Fuzzy bi-objective optimization model for multi-echelon distribution network Kanika Gandhi, P. C. Jha Bhavan’s Usha & Lakshmi Mittal Institute of Management, New Delhi, India Department of Operational Research, University of Delhi, Delhi, India gandhi.kanika@gmail.com,jhapc@yahoo.com Abstract It is important for modern businesses to search the ways for contin- uous improvement in performance of their supply chains. The effective coordination and integrated decision making across the supply chain enhances the performance among its various partners in a multi stage network. The partners considered in this paper are product suppliers, processing points (PP), distribution centres (DC) and retail outlets (RO). The network addresses an uncertain environment threatened by different sources in order to captivate the real world conditions. The uncertain demand of deteriorating products and its dependent costs develop uncertainties in the environment. On the other hand, suppliers and processing points have restricted capacities for the re- tail outlets’ order amount happened in each period. A bi-objective non-linear fuzzy mathematical model is developed in which the uncer- tainties are represented by the fuzzy set theory. The proposed model shows cost minimization and best supplier selection coordination un- der the conditions of capacity constraints, uncertain parameters and product’s deteriorating nature. The fish and fish products give good examples for the proposed model. To solve, the model is converted into crisp form and solved with the help of fuzzy goal programming. Key words: multi stage, supplier selection, processing point, fuzzy goal programming, supply chain, Bi-objective. 39 Gandhi, Jha 1 Introduction With the growing importance of supply chain management (SCM) in en- terprise development and in the operation of socio-economic systems, cost management has become a strategic business issue in recent years. It in- volves not only the financial flows but also the associated material flows and information flows among supply chain partners. Moreover, it plays an in- dispensable role in bringing profits and competitive advantage to firms, and consequently receives increasing attention from both supply chain managers and academics. Activities in supply chain system consist of transforming natural resources, raw materials and components into finished product and their final delivery to the end customers. Most of these economic activities form an integral part of the value chain. From this view point, cost manage- ment in supply chains is not limited to individual enterprises, but extends to all the purchasing, warehousing, production and distribution activities along the chain. Its goal is to provide a management tool and method to design the integrated chain, to promote its development and to reduce the total cost of supply chain system. However, a lot more complexity is involved in effectively integrating all the supply chain activities in a cost efficient man- ner owing to shorter life cycle of products and increased competition among suppliers who are offering different opportunities to the retailer. The un- certain demand of deteriorating products and their dependent costs creates uncertainty in the environment and consequently results in an indecisive and unsure environment for the decision makers. Choosing high level of procured quantity and inventory to avoid shortages will definitely lead to an immense increase in the cost of purchase and inventory holding. In this regard, op- erations management practices and mathematical models provide a sound framework for effective and integrative decision making across supply chain. For minimizing the cost and improving the overall performance, major func- tions considered are economic ordered quantity decisions, supplier selection decisions, inventory & capacity decisions and transportation policies in multi periods and for multi products. While economic ordered quantity decisions aim to minimize the cost of procurement, inventory and transportation, the intent of supplier selection and transportation policy selection decisions is to maximize inbound logistics performance by attaining a high degree of quality and delivery performance. Due to the inherent interdependency among these decisions, a firm cannot optimize them separately. Hence the main purpose of this paper is to develop a model addressing above issues i.e. to character- ize the optimal decisions that each partner in supply chain should adopt to motivate the chain partners to coordinate so that everyone benefits from the improved performance of the system. 40 Fuzzy bi-objective optimization model Though procurement functions need to consider cost minimization objec- tive, yet in doing so one cannot compromise on quality and delivery related criteria. Nowadays, quality and delivery related objectives are being given higher priority than cost criterion during procurement decisions. Suppliers’ performance on quality and delivery criteria has a significant influence on the ordered quantity and the total transportation costs. Taking into account the above observations, in this study we develop a fuzzy bi-objective non-linear programming model for an integrated economic ordered quantity, supplier selection and transportation policy problem. We investigate a problem in which multi products are procured from multiple suppliers in multiple pe- riods considering limitations on capacity at supplier point and processing point for deteriorating products. We also incorporate cost of inventory at distribution centres & retail outlets and transportation cost and policy con- cepts in one stage to another. Imprecise demand and other uncertain known parameters make the environment of model uncertain and fuzzy. To summa- rize the above discussions, the present work shows (1) a fuzzy bi-objective multi stage non-linear optimization model that includes computation of cost of procurement, processing, holding and transportation as first objective and the other objective shows the process to choose best supplier on the basis of delivery and quality; (2) the coordination among multi stages, i.e. (i) pro- curement stage; (ii) processing stage constituted of (a) Receiving & Scanning, (b) Sorting & Packaging & (c) Scanning & Dispatching; (iii) distribution cen- tres and (iv) retail outlets; (3) transportation policies and minimum cost per weight from processing stage to distribution centres and transportation cost per unit from distribution centre to retail outlet; (4) fuzzy set theory to coor- dinate uncertain parameters; (5) coordination in procurement, demand and inventory so the zero shortage is ensured. 2 Literature Review There are vast researches working on supplier selection problems with different approaches. One of the most important decisions related to pro- curement operations is supplier evaluation and selection. There are several factors involved such as price offered by the supplier, lead time, the quality of items, the capacity of supplier and the geographical location of supplier while making supplier evaluation and selection decisions (Ho et al., 2010). Ho et al. (2010), the three most important criteria considered while selecting suppliers are product quality, delivery lead time and price. Hassini (2008) studies a lot sizing and supplier selection problem when supplier capacity reservation dependent on lead time. Ravindran, Bilsel, Wadhwa, and Yang (2010) study 41 Gandhi, Jha supplier selection and order allocation considering incremental price breaks. Liao and Rittscher (2007) propose a multi objective programming model for supplier selection, procurement lot sizing and carrier selection decisions. Razmi and Maghool (2010) propose a fuzzy bi-objective model for multiple items, multiple period, supplier selection and purchasing problem under ca- pacity constraint and budget limitation. Zhang and Zhang (2011) formulate a mixed integer programming model for selecting suppliers and allocating the ordering quantity properly among the selected suppliers to minimize the selection, purchase and inventory costs. Jolai, Yazdian, Shahanaghi, and Khojasteh (2011) proposed a two-phase approach for supplier selection and order allocation problem under fuzzy environment for multiple products from multiple suppliers in multiple periods. Pal, Sana, and Chaudhuri (2012) ad- dressed a multi-echelon suppler chain with two suppliers in which the main supplier may face supply disruption and the secondary supplier is reliable but more expensive, and the manufacturer may produce defective items. Kilic (2013) discussed an integrated approach including fuzzy Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) and a mixed integer linear programming model is developed to select the best supplier in a multi- item/multi-supplier environment. Few of the studies have addressed problems having multi objectives and with fuzziness. Madronero, Peidro, and Vassant (2010) used S-curve mem- bership functions for Fuzzy aspiration levels for objective functions, maxi- mum capacity of the vendors as RHS, budget amount allocated to vendors as RHS with Fuzzy programming by using modified Werner’s fuzzy or op- erator. Wu, Zhang, Wu and Olson (2010) used Trapezoidal membership functions for Fuzzy model parameters as objective function coefficients and right hand side (RHS) constants with Sequential quadratic programming. Arikan (2011) used Triangular and Right triangular membership functions for Fuzzy aspiration levels for objective functions and demand level as RHS with Lai and Hwang’s augmented max–min model. Concerning with multi- choice goals, decision-making behaviour and limit of resources, Lee, Kang, and Chang (2009) develop a fuzzy multiple goal programming model to help downstream companies to select thin film transistor liquid display suppli- ers for cooperation. They used triangular membership functions for fuzzy aspiration levels for objective functions. Further, a multi-objective model for supplier selection in multi-service outsourcing is developed by Feng, Fan, and Li (2011). A multi objective mathematical model has been discussed by Seifbarghy and Esfandiari (2013), which includes minimizing the transaction costs of purchasing from suppliers as well as other objectives as minimizing the purchasing cost, rejected units, and late delivered units, and maximizing the evaluation scores of the selected suppliers. The problem is converted into 42 Fuzzy bi-objective optimization model single objective using weighting method and solved using meta-heuristics. Aghai, Mollaverdi and Saddagh (2014), outlined a fuzzy multi-objective pro- gramming model to propose supplier selection taking quantitative, qualita- tive, and risk factors into consideration. Also quantity discount has been considered to determine the best suppliers and to place the optimal order quantities among them. From the literature, it is evident that most studies have not paid much attention to uncertainty in supplier’s information and many problematic cri- teria in the conditions of multi product, transportation modes and multiple sourcing. The main purpose of this paper has been outlined as (1) to pro- pose a fuzzy bi-objective mathematical model to choose the supplier with best performance on the basis of quality & delivery percentages and to keep the cost optimum while procurement, processing of products and transportation, the ideal number of inventory items so that shortages does not take place, and optimum quantity from suppliers subject to the constraints pertaining to demand, suppliers capacity, processing capacity and inspection, (2) the objectives are conflicting in nature as minimization of cost and performance maximization of the supplier. Because of uncertain parameters the envi- ronment of the problem becomes fuzzy, for which, fuzzy goal programming method has been used to solve the mathematical model of cost minimization and suppliers selection with maximum performance. 3 Problem Definition To manage different entities to minimize their cost and simultaneously measuring the suppliers’ performances in the environment of uncertainty, the current paper presents a fuzzy bi-objective mixed integer non-liner model. The first objective of the proposed model minimizes the cost of integration of procurement and distribution. This comprises of multi source (suppliers), two processing points, multi distribution centres & multi retail outlets and incorporating transportation costs and policies. The second objective focuses on performance and selection of suppliers on the bases of on-time delivery percentage and acceptance percentage of the ordered quantity. The first stage of first objective explains procurement cost as per optimum procured quantity from the active suppliers, processing cost per unit in three levels at processing point. At this point receiving, scanning, sorting and packing of goods takes time, hence holding cost is included in the processing cost. The second stage shows the fuzzy cost of holding at distribution centres and cost of transportation of goods from processing points to distribution centres which is completed through two modes of transportation as full truck 43 Gandhi, Jha load (TL) mode and truck load (TL) & less than truck load (LTL) mode. In truck load transportation mode, the cost is fixed of one truck up to a given capacity. In this mode, the company may use less than the capacity available but cost per truck will not be reduced. However, sometimes the weighted quantity may not be large enough to corroborate the cost associated with a TL mode. In such situation, a LTL mode may be used. LTL is defined as a shipment of weighted quantity which does not fill a truck. In such a case, transportation cost is taken on the bases of per unit weight. The third stage includes inspection, fuzzy holding cost at retail outlet and transportation cost per unit in the account from distribution centres to retail outlet. The second objective is to find best suppliers with the combination of fuzzy on- time delivery percentage and fuzzy acceptance percentage of the ordered quantity. The model integrates inventory, procurement and transportation mecha- nism to minimize all costs discussed above and also chooses the best supplier. In the model, all the co-ordinations among supply chain partners are being managed under one buyer who is taking care of processing points, distribution centres and retail outlets but not sources (suppliers) directly. The total cost of the model becomes fuzzy due to fuzzy holding cost and demand. On the other hand, performance level is also fuzzy as percentage of on-time delivery and acceptances are fuzzy. Hence, the model discussed above is fuzzy bi- objective mixed integer non-linear model. In the solution process, the fuzzy model is converted into crisp and further fuzzy goal programming approach is employed where each objective could be assigned a different weight. 4 Proposed Model Formulation The model is based on following assumptions: • Finite planning horizon • Demand at retail outlet is uncertain and no shortages are allowed • Initial inventory at the beginning of planning horizon is zero • Inventory at retail outlet deteriorates at constant rate • Inspection cost of received goods at retail out is fixed • No transportation cost is discussed as it is considered as part of pur- chasing cost • Holding cost is part of processing cost at processing point 44 Fuzzy bi-objective optimization model 4.1 Sets Set Cardinality Index Product P i Supplier J j Processing Point Z z Distribution Centre M m Retail outlet O o Time period T t 4.2 Parameters ∼ C : Fuzzy total cost C0 & C ∗ 0 : Aspiration & Tolerance level of fuzzy total cost ∼ PR : Fuzzy performance of supplier PR0 & PR ∗ 0 : Aspiration & Tolerance level of fuzzy performance of sup- plier ∼ HD imt & HDimt : Fuzzy & Defuzzified holding cost per unit of product i for tth period at mth distribution centre ϕijzt : Unit purchase cost for i th product in tth period from supplier j for zth processing point A: Cost per weight of transportation in LTL policy Kzmt : Fixed freight cost for each truck load in period tfrom processing point z to distribution centre m TCimot : Transportation cost for unit in period tfrom distribution centre m to retail outlet o ∼ HR iot & HRiot : Fuzzy & defuzzified holding cost per unit of product i for tth period at retail outlet o λiot : Inspection cost per unit of product i in period t at retail outlet o ∼ D iot & Diot : Fuzzy & defuzzified demand at retail outlet o for product i in period t IN izt: Initial Inventory processing point z in beginning of planning hori- zon for product i η : Deterioration percentage of ith product at retail outlet wi: Per unit weight of product i ω : Weight transported in each full truck 45 Gandhi, Jha ∼ DT ijzt & DT ijzt : Fuzzy & defuzzified percentage of on-time delivery time for product i in period t for supplier j for processing point z ∼ AC ijzt & ACijzt : Fuzzy & defuzzified percentage of acceptance for product i in period t for supplier j for processing point z δijz: Capacity at supplier j for product ifor z th processing point αizrt : Capacity of Receiving & Scanning level (r) at z th processing point for product i in period t Cizrt : Cost of Receiving & Scanning level (r) at z th processing point for product i in period t βizst : Capacity of Sorting & Packing level (s) at z th processing point for product i in period t Cizst : Cost of Sorting & Packing (s) at z th processing point for product i in period t γizdt : Capacity of Scanning & Dispatching level (d) at z th processing point for product i in period t Cizdt : Cost of Scanning & Dispatching (d) at z th processing point for product i in period t 4.3 Decision Variable Xijzt : Optimum ordered quantity of product i ordered in period tfrom supplier j transported to processing point z Vijt: If ordered quantity is procured by active supplier j for product i in period tthen the variable takes value 1 otherwise zero uzmt: Usage of modes, either TL & LTL mode (value is 1) or only TL mode (value is 0) 4.4 Operating Variables Yizt : Procured quantity reached at Receiving & Scanning level of zth processing point from all the active suppliers Aizt : Goods moved to Sorting & Packaging from Receiving & Scanning level at zth processing point Eimt : Goods reaching at m th distribution centre from all processing points jzmt : Total number of truck loads in period t from processing point z to distribution centre m Qzmt : Weighted quantity in excess of truckload capacity Giot : Total quantity reached at retail outlet o from all distribution centres Iizt : Inventory at processing point in period t for product i Iimt : Inventory at distribution centre in period t for product i 46 Fuzzy bi-objective optimization model Iiot : Inventory at retail outlet in period t for product i Bizmt : Quantity of product i shipped from z th processing point to mth distribution centre in period t Fimot : Quantity of product i shipped from m th distribution centre to oth retail outlet in period t Lzmt : Weighted quantity transported from z th processing point to mth distribution centre in period t 4.5 Fuzzy Optimization Model Formulation Fuzzy dependent environment with respect to uncertain independent vari- ables cannot be quantified by Crisp mathematical programming approaches. Fuzzy optimization approach permits adequate solutions of real problems in the presence of vague information by defining the mechanisms to quantify uncertainties directly. Therefore, we formulate fuzzy optimization model for vague aspiration levels on cost, demand, on-time delivery percentage and acceptance percentage the decision maker may decide the aspiration and tol- erance levels on the basis of past experience and knowledge. 4.5.1 Formulation of objectives Initially a bi-objective fuzzy model is formulated which discusses about fuzzy total cost and performance of the suppliers. The first objective of the model minimizes the total cost, consisting of procurement cost of goods from supplier, processing cost, holding cost at distribution centres, transportation cost from processing point to distribution centres and further to retail out- lets, holding cost at retail outlets and finally inspection cost of the reached quantity at retail outlets. Minimize, ∼ C = T∑ t=1 Z∑ z=1 J∑ j=1 P∑ i=1 ϕijztXijztVijzt + T∑ t=1 Z∑ z=1 P∑ i=1 [ R∑ r=1 CizrtYizt + ( S∑ s=1 Cizst + D∑ d=1 Cizdt ) Aizt ] + T∑ t=1 M∑ m=1 P∑ i=1 ∼ HD imt Eimt 47 Gandhi, Jha + T∑ t=1 M∑ m=1 Z∑ z=1 [(AQzmt + jzmtKzmt) uzmt + (jzmt + 1) Kzmt (1 −uzmt)] + T∑ t=1 O∑ o=1 M∑ m=1 P∑ i=1 TCimotFimot + T∑ t=1 O∑ o=1 P∑ i=1 ∼ HR iot Iiot + T∑ t=1 O∑ o=1 P∑ i=1 λiotGiot The second objective discusses the performance of suppliers and maximizes the performance percentage of supplier as per on-delivery time percentage and acceptance percentage of ordered quantity. Maximize ∼ PR = T∑ t=1 Z∑ z=1 J∑ j=1 P∑ i=1 ( ∼ DT ijzt + ∼ AC ijzt ) Vijzt 4.5.2 Constraint Formulation All the suppliers must have enough capacity to fulfil the orders. The following equation ensures that the active supplier shall have enough capacity to complete the orders from processing point. Xijzt ≤ δijzVijzt ∀i,j, z,t Next equation ensures that only one supplier can be active for a particular product in a period. However, same supplier can be active again in next period. J∑ j=1 Vijzt = 1 ∀ i, t,z Goods are reaching at zth processing point from all the suppliers. Yizt = J∑ j=1 Xijzt ∀i, t,z At Receiving & Scanning level in processing point, 2% from each lot is re- jected and removed. Aizt = 0.98Yizt ∀i, t,z Quantity dispatched from zth processing point is being transported to all distribution centres. Aizt = M∑ m=1 Bizmt ∀i, z, t 48 Fuzzy bi-objective optimization model Goods reaching at mth distribution centre are transported from all the pro- cessing points. Eimt = Z∑ z=1 Bizmt ∀i, m, t Goods are transported from mth distribution centre to all the retail outlets. Eimt = O∑ o=1 Fimot ∀i, m, t Goods reaching at oth retail outlets Eiot are transported from all the distri- bution centres Giot = M∑ m=1 Fimot ∀i, o, t Following three equations explain the capacities in processing point at all the levels respectively i.e. Receiving and Scanning level, Sorting & Packaging level and Scanning and Dispatching level. Yizt ≤ αizrt ∀i, z,t,r Aizt ≤ βizst ∀i, t,z,s Aizt ≤ γizdt ∀i, t,z,d Next three equations show balancing equations at Processing Point, which also takes care of no shortages assumption. First two equations of the set calculate inventory at end of the period with respect to quantity reached at receiving and scanning level from the supplier and quantity sent to sorting & packaging level. The third equitation takes care of the shortages by balancing the quantity between the two levels discussed above. Iizt = Iizt−1 + Yizt −Aizt ∀i, t > 1,z Iizt = INizt + Yizt −Aizt ∀i, t = 1,z T∑ t=1 Iizt + T∑ t=1 Yizt ≥ T∑ t=1 Aizt ∀i,z Balancing at distribution centres have been discussed in next three equation, where assumption of no shortages has also been taken care of. Iimt = Iimt−1 + Eimt − O∑ o=1 Fimot ∀i, t > 1,m 49 Gandhi, Jha Iim1 = 0 ∀i,m T∑ t=1 Iimt + T∑ t=1 Eimt ≥ T∑ t=1 O∑ o=1 Fimot ∀i,m At retail outlets also, inventory has been balanced with respect to the received quantity and demand. Iiot = Iiot−1 + Giot − ∼ D iot −ηIiot ∀i, t > 1,o Iio1 = 0 ∀i,o (1 −η) T∑ t=1 Iiot + T∑ t=1 Giot ≥ ∼ T∑ t=1 ∼ D iot ∀i,o Following equation is an integrator and calculates the weighted quantity which is to be transported from processing point to distribution centres. Lzmt = P∑ i=1 ωiBizmt ∀z,t,m The next equation finds out transportation policy as per the weighted quan- tity. Here, the costs of TL policy and TL<L policy are compared as per the weight. Lzmt ≤ (Qzmt + jzmtw) uzmt + (jzmt + 1) w (1 −uzmt) ∀z,m,t The calculation of overhead quantity in TL<L policy is calculated by comparing total weighted quantity with total number of full truck loads as per weight is discussed in following equation. Lzmt = Qzmt + jzmtw ∀z,m, t Lastly, describing the nature of decision variables and enforcing the binary and non-negative restrictions to them. Xijzt, Yizt, Aizt, Eimt, Fimot, Giot, Lzmt ≥ 0; Vijzt, uzmt ∈ [0, 1] ; Iimt, Iiot, Iizt, Qzmt, jzmt are integer. 50 Fuzzy bi-objective optimization model 4.5.3 Formulated Model Minimize ∼ C = T∑ t=1 Z∑ z=1 J∑ j=1 P∑ i=1 ϕijztXijztVijzt + T∑ t=1 Z∑ z=1 P∑ i=1 [ R∑ r=1 CizrtYizt + ( S∑ s=1 Cizst + D∑ d=1 Cizdt ) Aizt ] + T∑ t=1 M∑ m=1 P∑ i=1 ∼ HD imt Eimt + T∑ t=1 M∑ m=1 Z∑ z=1 [(AQzmt + jzmtKzmt) uzmt + (jzmt + 1) Kzmt (1 −uzmt)] + T∑ t=1 O∑ o=1 M∑ m=1 P∑ i=1 TCimotFimot + T∑ t=1 O∑ o=1 P∑ i=1 ∼ HR iot Iiot + T∑ t=1 O∑ o=1 P∑ i=1 λiotGiot Maximize ∼ PR = T∑ t=1 Z∑ z=1 J∑ j=1 P∑ i=1 ( ∼ DT ijzt + ∼ AC ijzt ) Vijzt. Subject to Xijzt ≤ δijzVijzt ∀i,j, z,t J∑ j=1 Vijzt = 1 ∀ i, t,z Yizt = J∑ j=1 Xijzt ∀i, t,z Aizt = 0.98Yizt ∀i, t,z Aizt = M∑ m=1 Bizmt ∀i, z, t Eimt = Z∑ z=1 Bizmt ∀i, m, t Eimt = O∑ o=1 Fimot ∀i, m, t Giot = M∑ m=1 Fimot ∀i, o, t Yizt ≤ αizrt ∀i, z,t,r Aizt ≤ βizst ∀i, t,z,s Aizt ≤ γizdt ∀i, t,z,d Iizt = Iizt−1 + Yizt −Aizt ∀i, t > 1,z Iizt = INizt + Yizt −Aizt ∀i, t = 1,z T∑ t=1 Iizt + T∑ t=1 Yizt ≥ T∑ t=1 Aizt ∀i,z 51 Gandhi, Jha Iimt = Iimt−1 + Eimt − O∑ o=1 Fimot ∀i, t > 1,m Iim1 = 0 ∀i,m T∑ t=1 Iimt + T∑ t=1 Eimt ≥ T∑ t=1 O∑ o=1 Fimot ∀i,m Iiot = Iiot−1 + Giot − ∼ D iot −ηIiot ∀i, t > 1,o Iio1 = 0 ∀i,o (1 −η) T∑ t=1 Iiot + T∑ t=1 Giot ≥ ∼ T∑ t=1 ∼ D iot ∀i,o Lzmt = P∑ i=1 ωiBizmt ∀z,t,m Lzmt ≤ (Qzmt + jzmtw) uzmt + (jzmt + 1) w (1 −uzmt) ∀z,m,t Lzmt = Qzmt + jzmtw ∀z,m, t Xijzt, Yizt, Aizt, Eimt, Fimot, Giot, Lzmt ≥ 0; Vijzt, uzmt ∈ [0, 1] ; Iimt, Iiot, Iizt, Qzmt, jzmt are integer. 5 Solution Algorithm 5.1 Fuzzy Solution Algorithm In following algorithm by Zimmermann (1976) specifies the sequential steps to solve the fuzzy mathematical programming problems. Step 1. Compute the crisp equivalent of the fuzzy parameters using a defuzzification function. Here, ranking technique is employed to defuzzify the parameters as F2(A) = (al + 2am + au)/4, where al,am,au are the Triangular Fuzzy Numbers (TFN). Let − D iot be the defuzzified value of ∼ D iot and (D1iot,D 2 iot,D 3 iot) for each i,o & t be triangular fuzzy numbers then,D iot = (D1iot + 2D 2 iot + D 3 iot) /4. Similarly, − HD imt and − HR iot are defuzzified aspired holding cost at warehouse and destination. Step 2. Since industry is highly volatile and customer demand changes in every short span, a precise estimation of cost and performance aspirations is a major area of discussion. Hence, a better way to come out of such situation is to incorporate tolerance and aspiration level with the main objectives. The model discussed in section 4.5.3 can thus be re-written as follows: Find X, X ∈ S 52 Fuzzy bi-objective optimization model (1 −η) T∑ t=1 Iiot + T∑ t=1 Giot ≥ ∼ T∑ t=1 − D iot ∀i,o C(X)≤ ∼ C0 PR≥ ∼ PR0 Xijzt, Yizt, Aizt, Eimt, Fimot, Giot, Lzmt ≥ 0; Vijzt, uzmt ∈ [0, 1] ; Iimt, Iiot, Iizt, Qzmt, jzmt are integer. Step3. Define appropriate membership functions for each fuzzy inequal- ities as well as constraint corresponding to the objective functions. µC (X) =   1 ; C(X) ≤ C0 C∗0−C(X) C∗0−C0 ; C0 ≤ C(X) < C∗0 0 ; C(X) > C∗0 , µP R(X) =   1 ; PR ≥ PR0 P R−P R∗0 P R0−P R∗0 ; PR∗0 ≤ PR < PR0 0 ; PR < PR∗0 µIiot (X) =   1 ; Iiot(X) ≥ D0 Iiot(X)−D ∗ 0 D0−D ∗ 0 ; D ∗ 0 ≤ Iiot(X) < D0 0 ; Iiot(X) > D ∗ 0 Where D0 = T∑ t=1 O∑ o=1 Diotis the aspiration and D ∗ 0is the tolerance level to inventory constraints. Step4. Employ extension principle to identify the fuzzy decision, which results in a crisp mathematical programming problem given by Maximize α Subject to µc(X) ≥ α, µP R(X) ≥ α, µIiot (X) ≥ α, X ∈ S Where α represents the degree up to which the aspiration of the decision- maker is met. The above problem can be solved by the standard crisp math- ematical programming algorithms. Step5. Following Bellman and Zadeh (1970), while solving the problem following steps 1-4, the objective of the problem is also treated as a constraint. Each constraint is considered to be an objective for the decision-maker and the problem can be looked as a fuzzy bi-objective mathematical programming problem. Further, each objective can have a different level of importance and can be assigned weight to measure the relative importance. The resulting 53 Gandhi, Jha problem can be solved by the weighted min max approach. On substituting the values for µP R(x) and µC (x)the problem becomes Maximize α subject to PR(x) ≥ PR0 − (1 −w1α)(PR0 −PR∗0) C(x) ≤ C0 + (1 −w2α)(C∗0 −C0) (P1) µIiot (X) ≥ α X ∈ S w1 ≥ 0, w2 ≥ 0, w1 + w2 = 1,α ∈ [0, 1] Step6. If a feasible solution is not obtained for the problem in Step 5, then we can use the fuzzy goal programming approach to obtain a compro- mised solution given by Mohamed (1997). The method is discussed in detail in the next section. 5.2 Fuzzy Goal Programming Approach On solving the problem, we found that the problem (P1) is not feasible; hence the management goal cannot be achieved for a feasible value of α[0,1]. Then, we use the fuzzy goal programming technique to obtain a compro- mised solution. The approach is based on the goal programming technique for solving the crisp goal programming problem given by Mohamed (1997). The maximum value of any membership function can be 1; maximization of α[0,1] is equivalent to making it as close to 1 as best as possible. This can be achieved by minimizing the negative deviational variables of goal program- ming (i.e., η) from 1. The fuzzy goal programming formulation for the given problem (P1) introducing the negative and positive deviational variables ηj & ρj is given as Minimize u subject to µP R(X) + η1 −ρ1 = 1 µC (X) + η2 −ρ2 = 1 u ≥ wj ∗ηj j = 1, 2 ηj ∗ρj = 0 j = 1, 2 w1 + w2 = 1 α = 1 −u ηj,ρj ≥ 0; X ∈ S; u ∈ [0, 1]; w1,w2 ≥ 0 6 Case Study Fish is a highly perishable food which needs proper handling and preser- vation if it is to have a long shelf life and also retain a desirable quality and 54 Fuzzy bi-objective optimization model its nutritional value. The central concern of fish processing is to prevent fish from deterioration. When fish are captured or harvested for commercial purposes, they need some pre-processing so they can be delivered to the next part of the supply chain in a fresh and undamaged condition. This means, for example, that fish caught by a fishing vessel need handling so they can be stored safely until the boat lands the fish on shore. Some of the methods to preserve and process fish and fish products include control of temperature using ice, refrigeration or freezing, sorting and grading, chilling, storing the chilled fish. The model is validated for the case on fish and fish products. Case is taken for two suppliers, two processing points, three distribution cen- tres and three retail outlets for three time periods. Each processing point has its own internal three stages i.e. Receiving & Scanning, Sorting & Packing and Scanning & Dispatching. At processing point, fish products are received and scanned, which have been pre-processed to reduce the deterioration per- centage. Afterwards, they are sorted as per quality checks and packed and further sent to the next stage for final scanning before dispatching to the dis- tribution centres. The objectives include minimizing the cost of procurement, processing, transportation and inventory by obtaining the optimal ordered quantity, transportation weights & minimum inventory and maximizing the performance of procurement by choosing the best supplier on the basis of delivery and quality. The data on cost of procurement from suppliers, pro- cessing cost, transportation cost from one stage to another, cost of inspection and inventory carrying cost has been discussed. Three types of fish have been discussed in the case are Rohu, Katle and Pomfret which are ranging from Rs.80 to Rs.190 per kg. In the case, uncertain parameters are performance parameters, holding cost and demand. Further, defuzzified holding costs at all distribution centres and retail outlets are Rs.14, Rs.8 and Rs.8 for three fish types respectively in all the periods. The capacity at both the suppliers is 300 and 380 packets for fish type ‘Rohu’, 370 and 390 packets for fish type ‘Katle’ and 360 and 380 packets for fish type ‘Pomfret’. In processing stage, the costs of receiving & scanning, sorting & packing and scanning & dispatching are Rs.1, Rs.2 and Rs.2.5 respectively per packet. Inspection cost per packet is Rs.2 and deterioration percentage is constant with 3% deterioration cost. 55 Gandhi, Jha Product Type Supplier Rohu Katle Pomfret Supplier 1 134 90 190 Supplier 2 185 85 185 Table 1: Purchase Cost in all periods and at all processing points Product Type Processing Point Rohu Katle Pomf PP 1 320 310 300 PP 2 355 275 245 Table 2: Capacity at all stages in processing point for all periods Supplier 1 to PP1 & PP2 Product Type Period 1 Period 2 Period 3 AC DT AC DT AC DT Rohu 0.93 0.98 0.93 0.98 0.93 0.98 Katle 0.99 0.98 0.99 0.98 0.99 0.98 Pomfret 0.95 0.98 0.95 0.98 0.95 0.98 Supplier 2 to PP1 & PP2 Product Type Period 1 Period 2 Period 3 AC DT AC DT AC DT Rohu 0.95 0.99 0.95 0.99 0.95 0.99 Katle 0.93 0.97 0.93 0.97 0.93 0.97 Pomfret 0.95 0.97 0.95 0.97 0.95 0.97 Table 3: De-fuzzified Delivery time (DT) and Acceptance (AC) Probabilities Distribution Centre Processing Point DC 1 DC 2 DC 3 PP 1 2000 2500 2500 PP 2 2200 2900 2400 Table 4: Transportation cost per truck 56 Fuzzy bi-objective optimization model Retail Outlet Distribution Centre RO 1 RO 2 RO 3 DC 1 2 2.2 1.9 DC 2 2.2 2.5 2.1 DC 3 1.9 1.8 2 Table 5: Transportation cost per packet from DC to RO Product Type Retail Outlet Rohu Katle Pomfret RO 1 100 160 140 RO 2 110 150 135 RO 3 105 170 150 Table 6: De-fuzzified demand in all time periods Truckload per truck is 250kg. Overhead quantity transportation cost is Rs.9 per packet. 6.1 Results and Managerial Implications The model helps company to provide minimum total cost incurred co- ordinating all the entities. Rs. 1085767 is the total cost which consists of holding cost at distribution centres as Rs.65758, procurement cost of Rs.856600, processing cost of Rs.33001, cost of transportation from process- ing point to distribution centres of Rs.76588, holding cost at retail outlets of Rs.28015.63, cost of transportation from distribution centres to retail outlets of Rs.13848.80 and finally inspection cost of Rs.11956. It is observed from the results that highest proportion is of the cost of procurement, which clearly validates the requirement of supplier selection. Further, keeping a valid track of transportation polices is equally important as the second highest portion in the cost is due to the transportation cost only. Next observation is towards the impact of the product’s nature as holding cost at distribution centre con- tributes towards the third highest portion in the cost. To prevent the over valuation of cost, the aspiration and tolerance level have been considered as Rs.950000 and Rs.1220000. As validated with the help of cost, the suppli- ers’ performance is second objective of the model which is a combination of 57 Gandhi, Jha on-time delivery and acceptance percentage of the suppliers. The higher the performance of the supplier, better the performance of the company. Keeping the aspiration level of suppliers’ performance as 39 and tolerance as 30, the performance level of suppliers obtained is 35.04. The model tries to activate the high performers to procure ordered quantity so that uncertainty in the environment can be managed. Nearby 78% of the aspiration level of cost and performance has been attained which makes the environment more certain and crisp for future decisions. Processing Point 1 Per. 1 Per. 2 Per. 3 Pr.T. S1 S2 S1 S2 S1 S2 Rohu 0 350 0 350 0 350 Katle 350 0 350 0 350 0 Pomfret 350 0 350 0 350 0 Processing Point 2 Per. 1 Per. 2 Per. 3 Pr.T. S1 S2 S1 S2 S1 S2 Rohu 0 350 0 350 0 150 Katle 350 0 350 0 350 0 Pomfret 350 0 350 0 350 0 Table 7: Optimum ordered quantity from supplier (S1-S2) In Table 7, the positive ordered quantity indicates the active supplier to supply goods as he has the highest performance percentage between the two suppliers on the bases of on-time delivery, acceptance percentage and capacity. It can help in reducing the procurement cost and making the process smooth in further echelon. Tables 8 and 9 shows ending inventory at processing points and retail outlets, which ensures no shortages in the case of unexpected demand. It is observed that at second retail outlet, storage capacity and infrastructure is better as well as the cost of holding is also low, hence inventory is higher at this outlet in comparison to others. Inventory at distribution is not discussed as no inventory was leftover at any of the distribution centres. While transporting weighted quantity to distribution centres, the policy type, number of trucks and overhead weights are to be checked as each of them incurs cost. In the Table 10 it is observed that while transporting 58 Fuzzy bi-objective optimization model Processing Point Period 1 Period 2 Period 3 Product Type PP1 PP2 PP1 PP2 PP1 PP2 Rohu 7 7 14 14 21 21 Katle 7 7 14 14 21 21 Pomfret 3 7 10 14 17 21 Table 8: Inventory at processing points (in packets) Retail Outlet Period 1 Period 2 Period 3 ProductType RO1 RO2 RO3 RO1 RO2 RO3 RO1 RO2 RO3 Rohu 0 0 0 112 171 78 11 698 1 Katle 0 0 0 131 69 0 2 317 75 Pomfret 0 0 0 144 58 51 5 487 8 Table 9: Inventory at retail outlets (in packets) from processing point 1 to distribution centre 1 in period 2, only Truckload (T*) policy is used as 250kg can be transported by 1 truck. In this case, LTL policy will become expensive. On the other side, transporting from processing point 1 to distribution centre 1 in period 1, TL & LTL? policy is used as 49kg should be transported as per unit weight. In the case of TL<L policy, if overhead weighted quantity is transported through full truckload, the cost of transportation will become much higher than using LTL policy. Where TL & LTL is indicated as TLT and only TL is indicated as T. Some more operational variables who helped in smooth process of goods from one level to other are as follows: 7 Conclusion In the emerging business scenario, the concepts of time, volume and ca- pacity become even more essential to the managerial decision-making. Cus- tomers are more sensitive to delivery times and service quality. The coordi- nation among the members of the chain helps them to make a cost-effective procurement and distribution network as well as better response to the cus- 59 Gandhi, Jha Distribution Centre 1 Period 1 Period 2 Period 3 PP1 PP2 PP1 PP2 PP1 PP2 Tpt Quantity 49 7 250 0 329 250 No. of Trucks 0 0 1 0 1 0 Tpt Mode TLT? TLT T* T TLT T Qty Overhead 49 7 0 0 79 0 Distribution Centre 2 Period 1 Period 2 Period 3 PP1 PP2 PP1 PP2 PP1 PP2 Tpt Quantity 749 761 752 1000 686 500 No. of Trucks 2 3 3 4 2 2 Tpt Mode TLT TLT TLT T T T Qty Overhead 249 11 2 0 186 0 Distribution Centre 3 Period 1 Period 2 Period 3 PP1 PP2 PP1 PP2 PP1 PP2 Tpt Quantity 35 261 27 29 14 279 No. of Trucks 0 1 0 0 0 1 Tpt Mode TLT TLT TLT TLT TLT TLT Qty Overhead 35 11 27 29 14 29 Table 10: Transported quantity, no. of trucks, transportation mode, over- head quantity E imt Period 1 Period 2 Period 3 Dis.C. Rohu Katle Pomf Rohu Katle Pomf Rohu Katle Pomf DC 1 18 16 22 94 144 12 1 0 578 DC 2 451 634 425 544 534 674 657 421 108 DC 3 217 36 43 48 8 0 28 265 0 Table 11: 60 Fuzzy bi-objective optimization model Giot Period 1 Period 2 Period 3 R.O. Rohu Katle Pomf Rohu Katle Pomf Rohu Katle Pomf RO 1 0 35 0 215 295 288 0 31 1 RO 2 641 647 426 286 221 195 658 408 578 RO 3 45 4 64 185 170 203 28 247 107 Table 12: tomers’ demand. The authors explain the coordination among many entities of supply chain. As mentioned in the objectives of this study, the main plan of this research is to find optimum quantity from the best suppliers under fuzzy environment to develop an optimum coordination among multi supplier, multi processing points, multi distribution centres and multiple number of retail outlets. To attain the objective, a fuzzy bi-objective mathe- matical model is formulated with objective functions of cost and combination of timely delivery & acceptance of lot, keeping the constraints as supplier ca- pacity, processing capacity, deteriorating nature of the product and truck capacity. The parameters in study as holding cost, consumption, delivery time and acceptance percentage are fuzzy in nature. To handle the issues of uncertainty and fuzziness, the model is converted into crisp form with the help of membership functions of fuzzy modeling. The parameters are also converted into crisp form by using triangular fuzzy numbers. To obtain the solutions, a fuzzy goal programming is employed. Hence, the current study is able to find a balance between minimum cost and best performed supplier. The proposed model was validated by applying to the real case study data. References [1] Ho, W., Xu, X., and Dey, P.K. (2010). Multi-criteria decision making approaches for supplier evaluation and selection: a literature review. European Journal of Operational Research 202 (1), 16-24. [2] Hassini, E. (2008). Order lot sizing with multiple capacitated suppliers offering lead time-dependent capacity reservation and unit price dis- counts. Production Planning & Control, 19(2), 142–149. [3] Ravindran, A. R., Bilsel, R. U., Wadhwa, V., and Yang, T. (2010). Risk adjusted multi criteria supplier selection models with applications. International Journal of Production Research, 48(2), 405–424. 61 Gandhi, Jha [4] Liao, Z., and Rittscher, J. (2007). Integration of supplier selection, pro- curement lot sizing and carrier selection under dynamic demand condi- tions. International Journal of Production Economics, 107, 502–510. [5] Razmi, J., and Maghool, E. (2010). Multi-item supplier selection and lot-sizing planning under multiple price discounts using augmented e- constraint and Tchebycheff method. International Journal of Advances in Manufacturing Technology, 49, 379–392. [6] Zhang, J.-L., and Zhang, M.-Y. (2011). Supplier selection and purchase problem with fixed cost and constrained order quantities under stochas- tic demander. International Journal of Production Economics, 129, 1–7. [7] Jolai, F., Yazdian, S. A., Shahanaghi, K., and Khojasteh, M. A. (2011). Integrating fuzzy TOPSIS and multi-period goal programming for pur- chasing multiple products from multiple suppliers. Journal of Purchasing & Supply Management, 17, 42–53. [8] Pal, B., Sana, S. S., and Chaudhuri, K. (2012). A multi-echelon sup- ply chain model for reworkable items in multiple-markets with supply disruption. Economic Modelling, 29(5), 1891–1898. [9] Kilic, H. S. (2013). An integrated approach for supplier selection in multi-item/ multi-supplier environment. Applied Mathematical Mod- elling, 37, 7752–7763. [10] Lee, A. H. I., Kang, H. Y., and Chang, C. T. (2009). Fuzzy multiple goal programming applied to TFT-LCD supplier selection by downstream manufacturers. Expert System with Applications, 36, 6318–6325. [11] Madronero, M. D., Peidro, D., and Vassant, P. (2010). Vendor selection problem by using an interactive fuzzy multi-objective approach with modified S-curve membership functions. Computers and Mathematics with Applications, 60,1038–1048. [12] Wu, D. D., Zhang, Y., Wu, D., and Olson, D. L. (2010). Fuzzy multi- objective programming for supplier selection and risk modeling: A pos- sibility approach. European Journal of Operations Research, 200, 774– 787. [13] Arikan, F. (2011). An Augmented Max–min Model For Multiple Ob- jective Supplier Selection, 9th International Congress on Logistics and 62 Fuzzy bi-objective optimization model Supply Chain Management, 27–29 October, Yas ar University, Ces me- Izmir-Turkey. The congress proceeding including full-texts is in the pub- lication process. [14] Feng, B., Fan, Z., and Li, Y. (2011). A decision method for supplier se- lection in multi-service outsourcing, International Journal of Production Economics, 132, pp 240-250. [15] Seifbarghy, M. and Esfandiari, N. (2013). Modeling and solving a multi-objective supplier quota allocation problem considering transac- tion costs, Journal of Intelligent Manufacturing, 24(1), 201-209. [16] Aghai, S., Mollaverdi, N., and Sabbagh, M. S. (2014), A fuzzy multi- objective programming model for supplier selection with volume dis- count and risk criteria, The International Journal of Advanced Manu- facturing Technology, 71(5-8), 1483-1492. [17] Zimmermann, H. J. 1976. “Description and optimization of fuzzy sys- tems.” International Journal of General Systems 2 (4): 209–215. [18] Bellman, R. E., and Zadeh, L. A. (1970). “Decision-making in a fuzzy environment.” Management Science 17 (4): 141-164. [19] Mohamed, R. H. 1997. “The relationship between goal programming and fuzzy programming.” Fuzzy Sets and Systems 89 (2): 215-222. 63 Gandhi, Jha 64