JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 332 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. DEVELOPMENT OF FUZZY LOGIC APPROACH TO OPTIMIZE SAFETY STOCK LEVEL IN DETERIORATED PRODUCTS/A SUPPLY CHAIN DAIRY INDUSTRIES CASE STUDY Riyadh Jamegh PhD Candidate. University of technology riyadh872005@yahoo.com Dr. AllaEldin Kassam University of technology/ Department of production engineering and metallurgy/ Industrial Engineering branch, allakassam@yahoo.com Dr. Sawsan Sabih University of technology/ Department of production engineering and metallurgy/ Industrial Engineering branch, sawsanaa2006@yahoo.com Abstract: In today's complex environment, a high responding ability represents a core for each organization to survive in a competitive environment. To grip your position in intense competition market, the organization must design high efficiency inventory system that has the ability to respond to changes in demand and at the same time reduce holding cost of accommodation to the lowest possible value by controlling inventory drivers such as safety stock level (SS). The traditional approaches of safety stock are limited to deal with dynamic behavior of market. Advanced approaches based on soft computing allow the dynamic updating of SS level. In this paper, a highly advanced dynamic fuzzy logic (DFL) has been suggested as an innovation step to identify safety stock level in dairy industries with objective of minimizing total cost and meet with customer requirements. The proposed approach consists of three main steps firstly, identifying demand uncertainty conditions by applying fuzzy logic steps embedded by identifying dynamic (N) factor which represents the increasing level in demand in period time. Secondly, identifying of raw material availability conditions by applying fuzzy logic steps, and finally, identification of inventory on hand conditions by applying fuzzy logic steps. It is necessary to identify the level of SS dynamically in fuzzy logic as an output embedded with identifying of period specification concept which describes states of demand in a specific period in which the demand is high, medium, or low which leads to identify maximum values of universe of discourse of output (safety stock). Here Matlab program was used. The provided solution demonstrates the proposed model validity. There has been a significant reduction in safety stock level ranging from (7-98)% depending on product type and period specification with a reduction also in holding cost, while keeping the requirements fulfillment of customers demand. Key words: dynamic fuzzy logic, Inventory optimization, safety stock, dairy industries, supply chain. mailto:riyadh872005@yahoo.com mailto:allakassam@yahoo.com mailto:allakassam@yahoo.com mailto:sawsanaa2006@yahoo.com JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 333 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. INTRODUCTION: Deterioration is defined as decay, damage, spoilage, or perishability and its effect cannot be disregarded in inventory models(Yang & Tseng, 2015). The performance of supply chain is greatly affected by the inventory policy followed by organizations (Beheshti, 2010). So, if inventory strategy is poor, this will lead to high burdened enterprises and high holding costs. Therefore, many organizations work to find the most suitable inventory strategy (Yang & Tseng, 2015). The shelf life of perishable products like meat , fruits, and dairy or any cooked vegetable, or grain products depends upon the specification of products and environment conditions in which that products are stored, so an efficient system of inventory is a crucial element to ensure that the products are still fresh and satisfy the customer's demand efficiently(Chiadamrong & Lhamo, 2017). Dairy supply chain takes more attention to the products perishability, so poor network integration may lead to excessive inventory and then enormous losses(Touil, Echchtabii, & Charkaoui, 2016). For example, M.Kummu et al. (2012) noted that 25% of the global food produced was lost within the food supply chain before consumption and, roughly estimated that global food losses could be 30% to 40%. Lack of effective supply chain management practices could be one of the major reasons for food losses, hence analyzing and improving food supply chains may reduce food losses(Kummu et al., 2012). According to Gruen et al. (2002), the rate of average stock-out is 8.3% (Duan & Liao, 2013), Inventory annual holding costs today can be as high as 40% of inventory value (or, in some situations, even higher), while competition continually puts pressure on companies to achieve higher service levels; therefore, efficient inventory management is essential to all companies except the few that do not deal with any type of inventory(Wanke, Alvarenga, Correa, Hadi-Vencheh, & Azad, 2017). The Products with finite shelf life which are subject to perishability are considered as an important issue and force the organizations to manage carefully. Expired products consuming time and cost to rework (if it is possible) or destroy. This problem emerges significantly in healthcare or food industry where all products lose their value easily during the manufacturing process, storage process or distribution, and this explains the fact that a one third of food production is damaged(Duong, Wood, & Wang, 2015). Ghare and Schrader are the first researchers who considered deterioration concept in their research. They indicated that inventories are depleted not only by demand but also by their work. They presented a model that explained how deterioration affects the inventory model(Yang & Tseng, 2015). The aim of this research is to provide a dynamic approach to identify safety stock and it can be implemented in different industries. 1. FUZZY INVENTORY IN PERISHABLE PRODUCTS Fuzzy logic is considered an alternative approach to the traditional probabilistic approach which deals with ambiguity and uncertainty in inventory. Fuzzy logic was emerged for the first time by Lotfi A. Zadeh (1965) while traditional probabilistic inventory models were used for many decades. The first use of inventory fuzzy logic was started fairly in the 90’s (Kao & Hsu, 2002) as that reported by (Wanke et al) in 2017 (Wanke et al., 2017). Fuzzy inventory control for highly perishable products was presented by Hideki Katagiri & Hiroaki Ishii 2002). The objective was to maximize the profit. Fuzzy shortage cost and fuzzy outdates cost were the variables of the proposed model while the fuzzy profit is the output(Katagiri & Ishii, 2002). Multi-objective model of inventory deterioration products was developed by Savita Pathak & Seema Sarkar in 2012. The main objective was to maximize the profit of different items(Pathak & Sarkar, 2012). 2. LITERATURE REVIEW An improved model to identify reorder point and order quantity in dairy industry was developed by [E. Khanlarpour et al 2013]. The researchers presented a distinctive model where they began using fuzzy logic to calculate the time of re-ordering rather than the amount of re-ordering as in most researches. The researchers designed a fuzzy logic system consisting of six inputs which are (air condition, competitor https://www.sciencedirect.com/science/article/pii/S0048969712011862?via%3Dihub#! JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 334 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. company, customer income, religious and non-religious ceremonies, passenger & vacation, and customer satisfaction and one output (order quantity). Genetic algorithm was also used to identify optimal level of order quantity. Thus, researchers were able to introduce an intelligent warehouse system, which was an important shift in research(Khanlarpour, Fazlollahtabar, & Mahdavi, 2013). The study of pricing strategy effect on inventory level was studied by Panda et al in (2013) where they thought that a perishable product was a very important element of inventory system management. They also began their study by investigating the impact of price discounts policy and used dynamic pricing rather than using static pricing. Dynamic pricing policy for pre and post deterioration of products was employed in order to increase the rate of inventory depletion which leads to decrease in inventory level, reduction in holding cost, and then maximization in profit. Discount in prices is presented before and after starting of deterioration process. The discount with cumulative way is presented also for reduced units(Panda, Saha, & Basu, 2013). Ming-Feng Yang andWei-Chung Tseng (2015) presented a study on the effect of product perishability on inventory system. Where the researcher combined traditional deterioration model with quality prediction model in order to quantize the quality and determine the remaining value of each product. (Yang & Tseng, 2015). Rigzin Lhamo & Navee Chiadamrong developed a simulation optimization model to study replenishment policy for each scenario in order to manage the inventory of perishable products taking into account the age of these products. ARENA software simulation is used in order to identify the best inventory policy taking into consideration FIFO and LIFO withdrawal behavior by customer. The computational results concluded that policies based products age will lead to great reduction in cost compared to the case of without inventory age of product taken into consideration(Chiadamrong & Lhamo, 2017). Integration of location-inventory problem into SCN is presented by (Zhuo Daia et al in 2018). The model develops optimization conditions for perishable products based on fuzzy statues for capacity and carbon emission. The authors used hybrid genetic algorithm (HGA), hybrid harmony search (HHS), and Lindo software to solve optimization problem under different scenarios of different capacity confidence levels and carbon emission confidence levels. Through the numerical experiments, it is easy to conclude that the Lindo programming is faster than HGA & HHS while the last two are considered more efficient than Lindo. (Dai, Aqlan, Zheng, & Gao, 2018). According to our survey about safety stock, it is clear to notice the scarcity of research that deals with safety stock by using fuzzy logic. Table (1) below shows the approaches used in different researches to calculate safety stock level. Table (1). Summary of safety stock researches approaches. No Researcher Year Reference Tool Items Single Multi 1 Ming-Feng Yang & Wei- Chung Tseng 2015 (Yang & Tseng, 2015) quality prediction model ے 2 Linh N. K. Duonga et al. 2015 (Duong et al., 2015) simulation perishable and substitutable products 3 Hui-Ming Wee 1997 (Wee, 1997) price-dependent demand ے 4 Savita Pathak and Seema Sarkar 2012 (Pathak & Sarkar, 2012) Fuzzy logic ے 5 S. Panda et al. 2013 [14] price discounts policy Applicable for different cases 6 Qinglin Duan,T.WarrenLiao 2013 (Duan & Liao, 2013) Optimization based simulation ے JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 335 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. 7 1RIGZIN LHAMO& 2NAVEE CHIADAMRONG 2017 (Chiadamrong & Lhamo, 2017) Simulation by ARENA ے 8 Zhuo Daia et al. 2018 (Dai et al., 2018) Optimization model with fuzzy logic ے 9 E. Khanlarpour et al 2013 (Khanlarpour et al., 2013) fuzzy logic & genetic algorithm Just dairy firm 3. DESCRIPTION OF PROPOSED APPROACH The general architecture of the proposed model is shown in Figure (1). The proposed model has three steps: Identifying demand uncertainty level, identifying raw material level, and identifying inventory level. 3.1 GENERAL DESCRIPTION OF THE PROPOSED APPROACH The proposed approach takes into consideration three variables and treats them by applying fuzzy logic system to identify appropriate level of safety stock. The sub section below will explain this process. 3.1.1 DEMAND UNCERTAINTY CONDITIONS The process of employment demand uncertainty was done by following the steps below;  Identification of dynamic factor (N) factor which represents maximum rate of demand increasing for a specific time period (monthly basis values were identified).  Building the set of membership functions to describe the status of demand as shown in figure (2).  Identification maximum value of universe of discourse which obtained by multiplying dynamic (N) factor by demand level of previous day (D(i-1) ) and this value is changed daily. Identify demand of previous period D (i-1) [set input] of (FL) Identify demand of current period D (i) [daily input] of (FL) N factor Identify maximum value of inventory. Max universe discourse. Input of inventory value. Daily input of inventory. Identify maximum value of raw material. Max universe discourse. Input of raw material value. Daily input of raw material. Dynamic Fuzzy logic system of safety stock calculation Safety Stock level Figure (1). Architecture of the proposed model JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 336 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved.  Input the current day demand (D(i) ) and this value is also changed daily. 3.1.2 RAW MATERIAL AVAILABILITY Raw material availability is considered as an important part in identifying safety stock level because unavailability of raw material will lead to unmet customer requirements. The steps below show how raw material availability was employed in this model;  Building set of membership functions to describe the statues of raw material availability as shown in figure (3).  Identifying maximum value and minimum values of raw material availability which represent universe of discourse of this variable.  Input daily value of raw material availability current day (D(i) ) and this value is also changed daily. 3.1.3 INVENTORY LEVEL ON HAND The level of on hand inventory is an crucial part and must be taken in account in order to identify safety stock level. Suitable management of inventory leads to reduction the level of safety stock. Next steps explain how inventory on hand was employed in the proposed model approach:  Building a set of membership functions to describe the status of on hand inventory as shown in figure (4). 0 0.375 0.5 0.625 0.75 KKKKKKKKKKKAssume N* d (i-1)= DM Figure (2) Fuzzy set of demand stability level. D e g re e o f m e m b e rs h ip S M H UAV RAV AV 0 0.125 RM 0.34 RM 0.75 RM 0.9375 RM RM Assume Max RM=RM Figure (3) Fuzzy set of availability of raw material. D e g re e o f M e m b e rs h ip JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 337 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved.  Identifying the max value and min values of on hand inventory to represent universe of discourse of the variable.  Input of daily value of on hand current day inventory (D(i) ) and this value is also changed daily. 3.1.4 SAFETY STOCK LEVEL Safety stock level (SSL) is the unique output of the model. Where the rules of reasoning, are applied to the system after entering of all the variables. Maximum values of universe of discourse are changed daily allow high flexibility for the system and dynamic status based on period specification and this change of maximum value will be done by executing the following steps:  Identifying the (dynamic value which represents period specification) ranging from (10-30)% of current period demand D(i), and the value chosen based on season conditions of product, the value is (10%) for weak period, (20)% for normal period, and (30)% for peak period.  Building a set of membership functions to describe the status of demand as shown in figure (5).  Identifying the maximum value of universe of discourse by multiplying (dynamic value) by demand of current the day (D(i) ) and this value is changed daily. 3.2 MECHANISM OF SAFETY STOCK IDENTIFICATION The safety stock is calculated by applying dynamic fuzzy logic for the three pre-identified variables where the maximum value of universe of discourse of the first variable (demand stability level) is changed daily to control dynamic states for the system. The mechanism of changing this value is shown by following the procedures described in section (4.1.1), whereas for maximum universe of discourse of safety stock, the 0 0.05SS 0.08SS 0.17SS 0.2SS 0.3SS 0.32SS 0.4SS 0.42SS 0.51SS 0.54SS 0.62SS 0.65SS SS Assume MAX SS =SS Figure (5). Fuzzy sets of safety stock (output variable). Low little medium High v.high extreme v.extreme Figure (4). Fuzzy sets of inventory level. 0 0.08INV 0.16INV 0.32INV 0.4INV INV. Assume maxINV =INV D e g re e o f m e m b e rs h ip Low Medium High JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 338 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. procedures are presented in section (4.1.4). By calling the procedures of sections (4.1.2) & (4.1.3) the maximum values of universe of discourse of raw material and inventory on hand are identified respectively. The algorithm called Dynamic Fuzzy Safety Stock (DFSS) is used to identify safety stock level based on dynamic fuzzy logic; the steps below will explain this algorithm. Step (1): Identification of dynamic (N) factor for each product at each period. Step (2): Identification of universe of discourse of current period for the first variable (demand stability level) by multiplying of demand of previse period D (i-1) by dynamic (N) factor. Step (3): Identification of universe of discourse of the second variable (raw material availability). Step (4): Identification of universe of discourse of the third variable (inventory on hand level). Step (5): Identification of period specification for all products (PS). This specification includes the description of the period in which the demand is high, medium, or low in order to integrate with demand stability level which is identified later to know the maximum safety stock. Step (6): Identification of universe discourse of the unique output (safety stock level) based on period specification which was identified in step (5). 3.3 CASE STUDY The examination of the proposed model validation is presented in this case study. The dairy products are considered good example of perished products, so the process of inventory optimization is crucial. High demand variability and short shelf life forced toward adapting suitable inventory approach which can deal with this complexity. Demand stability level, raw material availability, and inventory on hand as inputs and safety stock as output form the fuzzy logic system to solve the problem of dairy industries company. Three products are selected which are (butter 100g, yoghurt 1 kg, and cheddar cheese 100gm). It must be called the algorithm of Dynamic Fuzzy Safety Stock (DFSS) which is used to optimize inventory level by identifying safety stock based on dynamic fuzzy logic as shown in steps below. In order to simplify the process we will use sample of data related to third product (cheddar 100gm) as shown in Table (2). Table (2). Snap of required data of third product (cheddar cheese 100gm). Step (1): Identification of dynamic (N) factor for each product at each period. The identification of dynamic factor represents a heart of the model; it is the identification of the dynamic state of the system through which the universe of discourse of demand stability level (first variable) is identified. It is determined based on daily basis. (if- function) Excel program was employed in order to identify these values, where the change rate in demand is determined between each two adjacent periods (previous and next day). For example if the demand changed from 50 to 100, it means that the change from the first order and (N) dynamic factor is equal to (2). After applying the function for demand columns of all products for all periods, we can obtain (N) dynamic factor as shown in Table (3). Product Month Day (N) factor d(i)[deman d] Max raw material daily raw material max inventory On hand inventory [first day] P3 T1 1 3 1080 30 17.09 750 128 P3 T1 2 3 45 30 9.22 750 P3 T1 3 3 1305 30 6.33 750 P3 T1 4 3 1575 30 17.5 750 P3 T1 5 3 1329 30 19.3 750 JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 339 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. Table (3), the values of (N) for all selected products at (12 months). (N) dynamic factor T1 T2 T3 T4 T5 T6 T7 T8 T9 10 T11 T12 P1 0 0 5 3 5 5 4 2 3 0 4 4 P2 0 3 3 2 5 2 5 2 4 2 0 3 P3 3 3 3 2 6 2 4 3 3 3 3 3 Step (2): Identification of universe of discourse of the current period for the first variable (demand stability level) The universe of discourse for demand stability level is obtained by multiplying the demand of previous period D (i-1) by dynamic (N) factor. Equation (1) shows the application of this process. Table (4) represents universe of discourse of first variable (demand stability level) while Table (5) shows the implementation of fuzzy set boundary conditions for cheddar cheese (P4) for first day. The demand of previous day is (100). Max. Universe of discourse for second day (DM) = D (i-1) *N (1) Table (4), universe of discourse of demand uncertainty conditions. System variables Linguistic variable Linguistic values Numerical ranges Demand uncertainty condition Low L (0-0.5 DM) Medium M (0.375-0.75 DM) High H (0.625-1DM) From Tables (2 &3), the universe of discourse for second day according to Eq. (1) equals to (3*100=300) where (3) represents (N factor) for cheddar cheese in January and (100) represents the demand of previous day. Table (5) shows the results of demand uncertainty universe of discourse. Table (5), universe of discourse of demand uncertainty conditions for the third day. System variables Linguistic variable Linguistic values Numerical ranges Demand uncertainty condition Low L (0-0.5 ) *300 Medium M (0.375-0.75 ) *300 High H (0.625-1) *300 Step (3): Identification of universe of discourse of the second variable (raw material availability). Universe of discourse of second variable (raw material) is identified on monthly basis, while daily level of raw material is entered to the model to identify the impact of this variable on safety stock (SS) level. Table (6) shows the universe of discourse of this variable while Table (7) presents the value of universe of discourse of the variable in January where max level of raw material is (30ton). Table (6), universe of discourse of raw material availability conditions. Raw material availability Unavailable UAV (0-0.34 RM) Rare available RAV (0.125-0.9375 RM) Available AV (0.75-1RM) JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 340 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. Table (7), universe of discourse of raw material availability conditions for January. Raw material availability Unavailable UAV (0-0.34) *30 Rare available RAV (0.125-0.9375) *30 Available AV (0.75-1) *30 Step (4): Identification universe of discourse of the third variable (inventory on hand level). Inventory on hand universe of discourse and fuzzy sets are identified in this step, Table (8) shows the universe of discourse of this variable and Table (9) presents the value of universe of discourse of the variable in January where maximum level of on hand inventory is (75 packages). Table (8). Universe of discourse of inventory on hand conditions. On hand inventory Low L (0-0.16 INV) Medium M (0.08-0.4 INV) High H (0.32-1 INV) Table (9). Universe of discourse of inventory on hand for January. On hand inventory Low L (0-0.16)*750 Medium M (0.08-0.4) *750 High H (0.32-1) *750 Step (5); Safety Stock Level (Output) Only one output i.e. safety stock level is determined by this model. Seven sets are used to identify the SS level. It is crucial to identify period specification. In January for the cheddar cheese the period is peak, so the percentage is (30%) of D(i), while for yogurt 400gm the period specification is (20)% of D(i). For example, the max universe of discourse of the first day of January of cheddar cheese is (0.3*1080) where (0.3) represents the period specification value and (1080) represents the demand of first day. Table (10) shows the universe of discourse of SS level while table (11) presents the value of universe of discourse of the first day for the cheddar cheese in January where max level of SS is changed daily. Table (10). Universe of discourse of SS level. Safety stock level Low L (0.-0.08 SS) Little LI (0.05-0.2 SS) Medium M (0.17-0.32 SS) High H (0.3-0.42 SS) Very high VH (0.4-0.54 SS) Extreme E (0.51-0.65 SS) Very extreme VE (0.62-1 SS) JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 341 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. Table (11). Universe of discourse of SS level for soft cheese in third day of January. Safety stock level Low L (0.-0.08)*324 Little LI (0.05-0.2)*324 Medium M (0.17-0.32)*324 High H (0.3-0.42)*324 Very high VH (0.4-0.54)*324 Extreme E (0.51-0.65)*324 Very extreme VE (0.62-1)*324 Referring to Table (2) and in order to simplify the process, third product (cheddar 100gm cheese (P3)) parameter was selected for (January), when calculate SS for first day. The demand of the previous period equals to (100) package with inventory on hand equals to (128) package and raw material available of (30) tone of raw milk, and demand of current day is (1080) package, then the safety stock is (116) package. Depending upon the specific percentage which mentioned in proposed model i.e. (MAX.SS= 30% of D(i)), SS= 30%*1080=324 and then distribute all percentage values as shown in figure (5). Figure (6) shows the results of SS level for (P3) in the first day of January The mechanism of dynamic fuzzy logic represents continuous change (depending on the specific time period) of fuzzy system which depends on the change occurred in universe of discourse of one or more of included variables, which results in a flexible system that deals with these changes that occurred periodically and thus lead to obtain high accuracy results. Figure (7) illustrates the dynamic behavior of fuzzy logic. Figure (6); implication process for identifying SS level for third day of January for soft cheese 0 92 140 203 324 Fuzzy inputs Applying AND operator (min) Applying implication process Aggregation results If demand stability is high and raw material is rare and inventory is medium R23 JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 342 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. By applying the proposed model on the case study parameters, great results were obtained. Clear reduction in SS level for all products is gained while keeping customer satisfactions, the reduction of SS level is varying between (7-98)% as shown in table (12); Table (12). Percentage reduction in SS level for three products in 12 months. In order to provide detailed description about the obtained results, Table (13) shows the values of the original demand, current SS level (identified by company), and safety stock of proposed model for butter 100gm product in January. The (current SS) for butter 100gm in March is (1219) and the (developed SS) is (74), so the reduction is about 93% as mentioned in table (13):- Table (13). Demand, current SS and developed SS for P1 (butter 100gm ) in (T3). Butter 100gm (March) Butter 100gm (March) Percentage impact of the proposed approach on Safety Stock T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 T11 T12 P1 0 0 93% 93% 93% 96% 96% 93% 92% 0 87% 88% P2 97% 96% 97% 96% 96% 96% 96% 91% 92% 97% 97% 98% P3 35% 7% 86% 68% 72% 82% 87% 86% 81% 95% 74% 72% Figure (7) dynamic mechanism of proposed fuzzy logic approach 0 0.375 0.5 0.625 0.75 KKKKKKKKKK KAssume N* d (i-1) = DM M e m b e r sh ip fu n c ti o n S M H N factor Demand 3 1080 3 45 3 1305 3 1575 3 1320 3 1435 3 577 3 1166 3 1665 3 1150 3 1295 3 1365 3 770 3 1380 3 1570 3 1345 specification Demand 0.3 1080 0.3 45 0.3 1305 0.3 1575 0.3 1320 0.3 1435 0.3 577 0.3 1166 0.3 1665 0.3 1150 0.3 1295 0.3 1365 0.3 770 0.3 1380 0.3 1570 0.3 1345 Low little med. High v.high extreme v. extreme M e m b e r sh ip fu n c ti o n 0 0.05SS 0.08SS 0.2SS 0.32SS 0.42SS 0.54SS 0.62SS SS KKKKKKKKKK KAssume MAX SS = SS JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 343 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. Day Demand Current SS Developed SS Day Demand Current SS Developed SS 1 14 61 2 12 18 32 2 2 92 19 23 13 22 60 2 3 66 3 3 14 15 45 2 4 41 112 2 15 16 79 2 5 18 94 2 16 24 55 2 6 48 46 6 17 6 49 1 7 40 6 3 18 20 79 5 8 23 83 2 19 13 66 1 9 21 62 2 20 20 46 2 10 46 66 5 21 8 38 1 11 16 50 1 22 20 68 3 Sum=1219 Sum=74 Reduction= 93% CONCLUSIONS In today's dynamic and global market environment, it is very important to use advanced approaches rather than traditional ones to identifying safety stock (SS) level. This seems more important when applied in the dairy industry due to the short shelf life of products and high competitive situations, which made the application of soft computing techniques absolutely indispensable. As the experience of the model on more than one product which has different specifications in terms of the dynamic factor and the amount of demand stability level proved the importance and success of the proposed model. Moreover the implementation of dynamic fuzzy logic in identifying SS level leads to more control on demand variation by employing an expert knowledge which enables the owners to overcome the problem of excessive inventory level due to reliance only on common statistical equations. The dynamic fuzzy logic has an excellent ability to deal with the advanced global market which is characterized by dynamic and competitive nature. It is crucial to notice that the dynamic fuzzy logic is considered as flexible tool because the dynamic input variable i.e. (demand level) and dynamic output variable (safety stock) can be readjusted and modified easily. Despite the success of the proposed model, there are some future areas of research, such as the use of advanced techniques in identifying inference rules, which gives the models high robustness in the identification of outputs. JOURNAL OF AL-QADISIYAH FOR ENGINEERING SCIENCE Vol. 11, No. 3 ISSN: 1998-4456 Page 344 Copyright  2018 Al-Qadisiyah Journal For Enginnering Science. All rights reserved. REFERENCES 1.Beheshti, H. M.. A decision support system for improving performance of inventory management in a supply chain network. International Journal of Productivity and Performance Management, (2010), 59(5), 452-467. 2.Chiadamrong, N., & Lhamo, R.. Inventory management of perishable products in a retail business: a comparison with and without in-store replenishment policies under different purchasing batch sizes. International Journal of Logistics Systems and Management, (2017), 26(2), 224-252. 3.Dai, Z., Aqlan, F., Zheng, X., & Gao, K.. A location-inventory supply chain network model using two heuristic algorithms for perishable products with fuzzy constraints. Computers & Industrial Engineering, (2018), 119, 338-352. 4.Duan, Q., & Liao, T. W.. A new age-based replenishment policy for supply chain inventory optimization of highly perishable products. International Journal of Production Economics, (2013), 145(2), 658-671. 5.Duong, L. N., Wood, L. C., & Wang, W. Y. (2015). A multi-criteria inventory management system for perishable & substitutable products. 6.Katagiri, H., & Ishii, H.. Fuzzy inventory problems for perishable commodities. European Journal of operational research, (2002), 138(3), 545-553. 7.Khanlarpour, E., Fazlollahtabar, H., & Mahdavi, I.. Designing an Intelligent Warehouse Based on Genetic Algorithm and Fuzzy Logic for Determining Reorder Point and Order Quantity. Computer Science and Information Technology, (2013), 1(1), 1-8. 8.Kummu, M., de Moel, H., Porkka, M., Siebert, S., Varis, O., & Ward, P. J.. Lost food, wasted resources: Global food supply chain losses and their impacts on freshwater, cropland, and fertiliser use. Science of the total environment, (2012), 438, 477-489. 9.Panda, S., Saha, S., & Basu, M.. Optimal pricing and lot-sizing for perishable inventory with price and time dependent ramp-type demand. International Journal of Systems Science, (2013), 44(1), 127-138. 10.Pathak, S., & Sarkar, S.. Fuzzy Inventory Models of Perishable Multi-items for Integrated and Non- integrated Businesses with Possibility/Necessity Measure of Trapezoidal Fuzzy Goal. International Journal of Modeling and Optimization, (2012), 2(2), 119. 11.Touil, A., Echchtabii, A., & Charkaoui, A.. Possibilistic programming approach for production and distribution problem in milk supply chain. Paper presented at the Logistics Operations Management (GOL), 2016 3rd International Conference on. 12.Wanke, P., Alvarenga, H., Correa, H., Hadi-Vencheh, A., & Azad, M. A. K.. Fuzzy inference systems and inventory allocation decisions: Exploring the impact of priority rules on total costs and service levels. Expert Systems with Applications, (2017), 85, 182-193. 13Wee, H.-M.. A replenishment policy for items with a price-dependent demand and a varying rate of deterioration. Production planning & control, (1997), 8(5), 494-499. 14.Yang, M.-F., & Tseng, W.-C. Deteriorating inventory model for chilled food. Mathematical Problems in Engineering, 2015.