An Optimal Returned Policy for a Reverse Logistics Inventory Model with Backorders

Environmental responsibility plays a significant role in the firm’s agendas nowadays. In this paper, we address the environmental operations of reverse logistics. Here we developed an integrated supply chain model with coordinated production and remanufacturing due to time-dependent rates. To study the problem we consider the demand to be satisfied with newly manufactured produced and the remanufactured products, so there is no difference between manufactured and remanufactured items. The shortages are allowed and excess demand is backlogged as well. The returned items are collected from the end user to be remanufactured. Optimal expression is obtained for the acceptable returned quantity, maximum inventory level, production and remanufacturing scheduling period, and the total average cost. Illustrative examples, which explain the application of the theoretical results as well as their numerical verifications, are given. The sensitivity of these solutions to change in underling parameter values is also discussed.


Introduction
Environmental issues are gaining reasonable attention among society, worldwide.Consumer demand for clean manufacturing and recycling is increasing.Consumers expect to trade in an old product when they buy a new one.Hence, for the past few decades the reverse flow of products from consumers to upstream businesses has received much interest.Due to the governmental regulations and consumer concerns regarding these environmental issues, an increasing number of companies have focused on reduction efforts in the amount of waste stream, diversion of the discarded products, and disposition of the retired products properly whereas reverse logistics is the process of retrieving the product from the end consumer for the purposes of proper disposal.To facilitate the reverse flow of used products from consumers to manufacturers in an efficient manner, the most appropriate approach is to create a reverse supply chain network.Wherein reverse logistics can take place through the Advances in Decision Sciences original forward channel, through a separate reverse channel, or through combinations of the forward and the reverse channel.Generally, companies focus on setting up a reverse supply chain either because of environmental regulations or to reduce their operating cost by remanufactured products or components.For companies that utilise a reverse supply chain deals with handling and reprocess of repairable used products withdrawn from production and consumption process.Such a reuse is, for example, recycling or repair of spare parts.It has an advantage from economic point of view, as reduction of environmental load through return of used items in the manufacturing process.
In the recent years researchers paid much attention to reverse logistics inventory models.There have been numerous studies and research on reverse logistics.In the past one approach adopted by many authors to the study of recovery systems is the use of Economic Ordering Quantity EOQ technique.The EOQ models are simple and they usually lead to closed-form solutions.The first reverse logistic model was investigated by Schrady 1 .He analyzed the problem in the EOQ model for repairable items which assumes that the production and repairing rates are instantaneous without disposal cost.Nahmias and Rivera 2 considered the model of Schrady 1 for the case of finite repair rate and limited storage in the repair and production shops.A very good review on quantitative models for recovery production planning and inventory control is given by Fleischmann et al. 3 .In this review they subdivided the field into three main areas, namely, distribution planning, inventory control, and production planning for each of these they discuss the implication of the emerging reuse efforts.There is a multiproduct generalization of EOQ-type reverse logistics models published by Mabini et al. 4 .They have extended the basic model of Schrady 1 with capital budget restriction.Richter 5,6 investigated a modified version of the model of Schrady 1 by assuming multiple production and multiple repair cycles within a time interval.Most of the models investigated earlier are governed by two extreme bang-bang strategies, that is, "dispose all" or "recover all" 7 .In a similar work to Richter 5-7 , Teunter 8 developed a deterministic EOQ inventory model with disposal option where recoverable and manufactured items have different holding costs and obtained a general finding similar to Richter 7 .Koh et al. 9 generalized the model of Nahmias and Rivera 2 by assuming a limited repair capacity.In a later study, several researchers have developed models along the same lines as Schrady and Richter, but with different assumptions, for example, Teunter 10 ,Inderfurth et al. 11 ,Dobos,and Richter 12,13 .Dobos and Richter 14 explored their previous model by assuming that the quality of collected used items is not always suitable for further recycling.Konstantaras and Papachristos 15 have investigated an inventory model for stability.In his next work 16 he extended the work of Koh et al. 9 and followed a different analysis to obtain closed form expressions for both optimal number of set up in the recovery and the ordering processes.Konstantaras   work Jaber and El Saadany 18 by assuming that unfulfilled demand for remanufacturing and produced items is either fully or partially backordered.This paper also considered the scenario of overlapping of one production and one remanufacturing cycle to minimize the effect of stock outs.A closed-loop supply chain inventory model is developed by Yang et al. 24 ,in press .In this paper he considers price-sensitive demand and multiretailer and analysis of the problem with three optimization methods sequential optimization, a centralized optimization without benefit sharing, and a centralized optimization with benefit sharing.The comparative review is given in Table 1 .
Usually in most of the models shortages are not permitted to occur.However, in many practical situations, stock out is unavoidable due to various uncertainties.Therefore, the occurrence of shortages in inventory is a natural phenomenon.In this paper we use this phenomenon.In the proposed model we determined the coordination of reverse manufacturing with the forward supply chain in the inventory management.The reverse logistics operations deal with the collection of returns, cleaning of the collected returns, and remanufacturing of the reusable collected items.The quality of the remanufactured items is assumed to be good as those of new products hence the demand is to be satisfied with newly manufactured produced and the remanufactured products.A general framework of the system is shown in Figure 1.In Section 1, a comprehensive literature review and background of the model are presented.Section 2 is for assumption and notations.Section 3 demonstrates the model development.Section 4 presents the solution procedure to solve the optimization problem.Section 4.1 shows three numerical examples to illustrate the model and sensitivity analysis is presented in Section 4.2 .A particular case of the given problem is given in Section 4.3 .Concluding remarks are derived and future research topics are suggested in Section 5 .

Assumptions and Notations
In this paper the subscript "m" is used to indicate the quantity corresponds to the remanufactured stock, we will use the subscript "r" to indicate the quantity corresponds to the remanufactured stock and the subscript "R" to indicate the quantity corresponds to the returned stock.
The model is developed with the following assumptions and notations.
i New products are produced at a rate of P m t .
ii Repairable used products are collected at a rate of R t and then remanufactured at a rate of P r t .All the returned items are remanufactured.iii Demand is satisfied from the newly produced and remanufactured items at a rate of D t .
iv P m t , R t , P r t , and D t are assumed to be arbitrary functions of time.
v We will require that vi Shortage is allowed and c s is the unit shortage cost.
vii Cost parameters related to manufactured products are as follows.
I m t is the inventory level at time t related to manufactured products.
q 2 is the maximum inventory level in the production process.
c m is the unit procurement cost.
s m is the unit production cost.
h m is the unit holding cost per unit time.
k m is the setup cost per production cycle.
viii The cost parameters related to remanufactured products are as follows.
I r t is the inventory level at time t related to remanufactured products.
q 1 is the maximum inventory level in the remanufacturing process.
s r is the unit remanufacturing cost.
h r is the unit holding cost per unit time.
k r is the setup cost per remanufacturing cycle.
ix The cost parameters related to returns are as follows.
I R t is the inventory level at time t related to returns.
c R is the unit acquisition cost.
h R is the unit holding cost per unit time.
k R is the setup cost per returned cycle.

Mathematical Modelling and Analysis
The cycle now starts at time t 0 with backorders.At this instant of time, remanufacturing starts to clear the backlog by the time t 1 and the inventory level I r t increases at a rate P r t − D t until the time t 2 where stock level reaches its maximum value.Then the remanufacturing is stopped and hence the demand depletes the inventory level I r t during the period t 2 , t 3 and falls to zero at t t 3 , thereafter shortages occur during the period t 3 , t 4 due to the absence of stock.At this instant of time, fresh production starts to clear the backlog by the time t 5 .Production raises the inventory level I m t at a rate P m t − D t and reaches its maximum at time t t 6 .Then the production is stopped and hence the demand depletes the inventory level I m t until the time t 7 by which it becomes zero.Now shortages start developing and accumulate to their maximum equal to the shortage level at time t t 0 at the time t t 8 .However for each returned cycle the inventory level I R t is affected by the returned rate and the remanufacturing rate, as the remanufacturing process starts at t 0 , the stock level declines at a rate R t − P r t and falls to zero at t t 2 by which the remanufacturing stops.Now the stock level increases at a rate R t by the time t t 8 .This is depicted in the Figure 2.
The differential equations governing the stock level during the period t 0 ≤ t ≤ t 8 can be written as

3.1
Advances in Decision Sciences 7 Solutions of the above differential equations using their boundary conditions are 3.2 3

3.4
Let I x 1 , x 2 x 2 x 1 I u du, then from 3.3 -3.6 we have 3.5

3.9
Advances in Decision Sciences 9 Here we have a cost function of the system in terms of t  3.14 shows that the backorders at the time t 0 and t t 8 are alike as we have already assumed in the formulation of model.According to the presented model all the units buyback from the market will be remanufactured no disposal hence the total remanufacturing during the cycle is equal to the total returned items in the complete cycle which is demonstrated in 3.15 .

Solution Procedure
Let Q be the acceptable returned quantity for used items in the interval 0, t 8 : And let we have the maximum inventory level q 1 and q 2 in remanufacturing and production process, respectively, so that from 3.11 and 3.13 we get Therefore the total variable cost function will be the function of three variables Q, q 1 , and q 2 which is Z Q, q 1 , q 2 : where the values of t 1 , t 2 , t 3 , t 4 , t 5 , t 6 , t 7 , and t 8 are given in the next section.
To obtain the optimal solution of the proposed problem, we minimize the functions Z Q, q 1 , q 2 with respect to Q, q 1 , and q 2 Taking the first order derivatives of Z Q, q 1 , q 2 with respect to Q, q 1 , and q 2 .Thereafter setting it equal to zero gives ∂Z Q, q 1 , q 2 ∂Q 0, ∂Z Q, q 1 , q 2 ∂q 1 0, ∂Z Q, q 1 , q 2 ∂q 2 0.

4.5
The optimal value of Q, q 1 , and q 2 can be derived from the above equations when the following conditions of Hessian matrix are satisfied.The Hessian matrix is The

4.12
By which we can find t 1 , say

Advances in Decision Sciences
From 4.2 we have

4.14
From which we can find t 3 , say From 3.14 we have

4.16
Hence the value of t 7 is given by  From which we can find t 5 From 3.12 we have

4.22
From which we can find t 4 Applying the solution procedure given in the last section we derive the optimal solution and results are presented in Table 2 .
The convexity of the reverse logistics inventory model is shown in Figure 3 .The three dimensional graph shows that the integrated expected total annual cost is convex, and that there exists a unique solution minimizing the integrated expected total annual cost.

Sensitivity Analysis
To study the effects of the parameter changes on the optimal solutions derived by the proposed method, this investigation performs a sensitivity analysis by increasing or decreasing the parameters, one at a time.
The main conclusions drawn from the sensitivity analysis given above are as follows.
i From Tables 3 and 4 we have observed that as the production rate increases the optimum value of total acceptable returned quantity decreases and hence the total Figure 3: Convexity of the reverse logistics inventory model is derived, and when the optimal value of acceptable returned quantity Q, maximum inventory level q 1 and q 2 is taken.minimum cost slightly decreases.However we observe a little but unexpected increment in the total minimum average cost and this increment happened due to the reduction in total cycle length.
ii Similarly from Tables 5 and 6 we have observed that as the remanufacturing rate increases the optimum value of total acceptable returned quantity decreases and hence the total minimum cost slightly decreases.But due to the reduction in total cycle length we observe a little increment in the total minimum average cost.
iii From Tables 7 and 8 it is observed that as the return rate increases the optimum value of total acceptable returned quantity increases while the total time required to produce the optimum manufactured quantity decrease.Hence the total procurement and production cost will decrease.This reduction resulted in the decrease in the total minimum average cost.
iv Similarly from Table 9 it is observed that as the constant demand rate increases the optimal values of total acceptable returned quantity and the total time required to produce the optimum manufactured quantity increase.Hence the total procurement, acquisition, and production cost increases.This increment resulted in the increase in the total minimum average cost per cycle.
v Now from Table 10 it is observed that as the variable demand rate increases the optimal values of total acceptable returned quantity decrease and hence the total  minimum cost slightly decreases.But due to the reduction in total cycle length we observe a little increment in the total minimum average cost.

A Particular Case When q 1 and q 2 Are Constant
In a particular case let us have q 1 and q 2 the maximum inventory level in the remanufacturing and production process, as a constant then from 4.9 -4.23 the values of t 1 , t 2 , t 3 , t 4 , t 5 , t 6 , t 7 , and t 8 can be found in the form of single variable Q, say From 4.4 the total cost function will be the function of one variable, say Q and the problem will be converted into the following unconstrained problem with one variable Q:

4.25
Now the necessary conditions for having a minimum for the problem  Taking the first derivative of both sides of 3.11 , 3.12 , 3.13 , 3.14 , and 3.15 with respect to Q.We obtain

4.31
As we have q 1 and q 2 are constant so that from 4.2 and 4.3 both sides of 4.29 and 4.30 will be equal to zero Form the above equation we can find the values of

4.34
From which and 4.28 we obtain where W is given by 4.25 and w Q is given by 4.34 .Equation 4.35 can now, be used to determine the optimal value of Q.If we get more than one solution we choose the one for which the condition d 2 W/dQ 2 > 0 is satisfied.Behavior of the inventory cost function, when the values of q 1 and q 2 are taken to be constant.
Then the optimal values of t 1 , t 2 , t 3 , t 4 , t 5 , t 6 , t 7 , and t 8 can be determined by using 4.9 -4.23 .From the above-mentioned results we can determine the value of w Q and w.Hence from 4.25 we can get the minimum total average cost.
Example 4.3.The same set of input data are considered as in the Example 4.2 except that the constant values of q 1 and q 2 are as follows: q 1 800 units, q 2 200 units.
Applying the solution procedure given in last section we find that the optimal value of the acceptable returned quantity Q is 1737 and hence the corresponding optimal results are presented in Table 11 .
The convexity of the total cost function has been established graphically in Figure 4 it has also been verified numerically that the cost function is convex.

Conclusion
In this paper designed model for a supply chain network with reverse flows is proposed.This paper generalises a reverse logistics inventory model for integrated production of new items and remanufacturing of returned items with shortages.In many practical situations, stock out is unavoidable due to various uncertainties.Therefore, the occurrence of shortages in inventory is a natural phenomenon.The developed model is solved by using the Hessian matrix.Results presented herein provide a valuable reference for decision makers in production, storage remanufacturing, and returning planning.A numerical example demonstrates that applying the proposed model can minimize the total average cost.In addition, sensitivity analysis is performed to examine the effect of parameters.According to those results, the presented model is least sensitive with respect to the production and remanufacturing rates while enough sensitive with respect to the returned and demand rates.It can be concluded from the obtained results that the higher returned rate can provide more profit.A future study should incorporate more realistic assumptions into the proposed model, for example, stochastic nature of demand, production, and remanufacturing rates follow learning and forgetting curves, and consider multiple production and remanufacturing batches per interval.

Figure 1 :
Figure 1: The flow of material in the integrated inventory model.
and Papachristos 17 extended the work of Teunter 10 by introducing the exact solution method for the same model.Jaber and El Saadany 18 extend the work of Richter 5, 6 by assuming the newly produced and remanufactured items are perceived differently by customers.Omar and Yeo 19 developed a production model that satisfies a continuous time varying demand for finished goods over a known and finite planning horizon by supplying either new products or repaired used products.The extended version of Dobos and Richter 12, 13 is made by El Saadany and Jaber 20 by assuming the returned rate of used items follows a demand like function depend on the purchasing price and acceptance quality level of returns.Alamri 21 derives a general reverse logistics inventory model for deteriorating items taking the returned rate as a decision variable.Another extension of Koh et al. 9 is made by Konstantaras et al. 22 by introducing the inspection and sorting of returned items.Hasanov et al. 23 extended the

Table 1 :
The comparative review in the tabular form.
Holding cost h r I r t 1 , t 2 I r t 3 , t 4 r t 2 , t 3 h m I m t 5 , t 6 I m t 6 , t 7 h R I R 0, t 2 I R t 2 , t 8 , Shortage cost c s −I r 0, t 1 − I r t 3 , t 4 − I m t 4 , t 5 − I m t 7 , t 8 .3.8 Hence the total cost per unit time of the given inventory model during the cycle 0, t 8 as a function of t 1 , t 2 , t 3 , t 4 , t 5 , t 6 , t 7 , and t 8 say Z t 1 , t 2 , t 3 , t 4 , t 5 , t 6 , t 7 , t 8 is given by Z t 1 , t 2 , . . ., t 1 , t 2 , t 3 , t 4 , t 5 , t 6 , t 7 , and t 8 .To find the optimum solution we have to find the optimum value of t 1 , t 2 , t 3 , t 4 , t 5 , t 6 , t 7 , and t 8 that minimize Z t 1 , t 2 , . . ., t 8 but we have some relations between the variables as follows.0 ≤ t 1 ≤ t 2 ≤ t 3 ≤ t 4 ≤ t 5 ≤ t 6 ≤ t 7 ≤ t 8 , 3.10 Equation 3.10 is an essential condition for the existence of the model.Equation3.11showthat the inventory levels I r t are same at the time t t 2 , similarly 3.12 and 3.13 show that the inventory level I m t and I r t are same at the time t t 4 and t t 6 , respectively, depicted in Figure2.Equation first principal minor determinant of H, |H 11 | > 0. The second principal minor determinant of H, |H 22 | > 0 and the third principal minor determinant of H, |H 33 | > 0.

Table 2 :
The optimal results for the inventory model under the above parametric values as in Example 4.2.

Table 3 :
For the different constant production rates optimal results for the same set of values as in Example 4.2.

Table 4 :
Optimal results for the same set of values as in Example 4.2 for different variable production rates.

Table 5 :
Optimal results for the same set of values as in Example 4.2 for different constant remanufacturing rates.

Table 6 :
Optimal results for the same set of values as in Example 4.2 for different variable remanufacturing rates.

Table 7 :
Optimal results for the same set of values as in Example 4.2 for different constant return rates.

Table 8 :
Optimal results for the same set of values as in Example 4.2 for different variable return rates.

Table 9 :
Optimal results for the same set of values as in Example 4.2 for different constant demand rates.

Table 10 :
Optimal results for the same set of values as in Example 4.2 for different variable demand rates.

Table 11 :
The optimal results for the inventory model under the parametric values given in Example 4.2.