電子資訊產業供應鏈管理---子計畫二：晶圓製造設備備用零件及光罩需求管理之研究

全文

(1)

(2)

(3)

(4) !" A Study of Demand Management for Spare Parts and Mask in Wafer Manufacturing NSC87-2213-E-009-044 86 8 1

(5) 87 7 31

(6) . 1.#$%&(

(7) ) !"

(8) #$ %&'()*+, VMI -.*/ 01ti, s, S2*3456789#$: ;*<=">?@A1s, S2B%) C&/D EFG* HI/JKLMN% +,OP=Q*&RST,U 0*V3WX*H VMI YZ[.O 01ti, s, S2\]^1s, S2 _`a\b 9\FG %cdW*e 1ti, s, S2f(/D 9EFG&g;hi%. Abstract. (Keywords: inventory policy, spare part, VMI(vendor managed inventory), customer servic e level, inventory cost, inventory turns.) The objective of this resear ch is to develop an appropriat e inventory policy for the spar e part supplier. A VMI-bas ed inventory policy, (ti, s, S), which can be properly re fer red to as a stocking policy for customer i, for joint consideration of demand and supply by incorporating a continuous review (s, S) type inventory policy. The potential advantage of the developed inventory policy is to improve customer service level, to reduce total inventory cost, and to increase inventory turns. The result of this resear ch pres ents suffi cient evidenc e to indicate that the VMI-bas ed (ti, s, S) policy results in higher customer service level and lower total inventory cost than (s, S) policy does under same stock levels. In other words, when adopting the (ti, s, S) policy, positive effects in the increase of customer service level and decreas e of total inventory cost can be obtained.. 2.Introduction. !"#. This research explor es the inventory policy of a spare part supplier in semiconductor equipment industry in Taiwan. The supplier sells equipment to semiconductor manufa cturing fa ctories. Most of the equipment the supplier sells is very expensive and costs millions of U.S. dollars. The equipment is critical to the processes in semiconductor manufacturing and its failure may lead to huge lost. Not only selling the equipment, the supplier also provides spare parts of the equipment. Effi cient service is essential as it affects the sales of the equipment directly. Inventory management of the spare pa rts, there for e, is one of the most important activities for the supplier. I f parts are understocked, customer demands cannot be satisfi ed and customer complaints will emerge. On the other hand, if parts are overstocked, inventory car rying costs will be high which may cause financial problems. Situations where some parts have ve ry high inventory and some others a re in shortage are not unusual. In such a service system, an efficient inventory policy is essential. The spares depa rtment of the supplier is responsible for spare part supply. For the ordinary failures of spare pa rts or at the preventive maintenanc e, the supplier sells the corresponding spare parts to meet customer demand. The spare parts are procur ed mostly from the overseas he adquarte rs and partly from local vendors depending on the part chara cteristics and the existence of the cor responding vendor. At present, the inventory policy is ca rried out with the so called (s, S) policy, based on the planner’s experience. The objective of this resear ch is to develop an appropriat e inventory policy for the spar e part supplier under the given environment described above. The potential advantages of the developed inventory policy will be : to improve customer service level; to reduce total inventory cost; to increase inventory turns. In recent years, the practice of VMI(Vendor Managed Inventory) has be en widely discussed. jRobin, 1995kVMI repres ents a partnership betwe en the supplier and the customer. It lowers mutual operating costs, optimizes shipment quantities, and. 1.

(9) achieves excellent servic e level for the suppliers’ customers. Based on the concept of VMI, this resear ch proposes a new policy, (ti, s, S), which can be properly re ferr ed to as a stocking policy for customer i, for joint consideration of demand and supply by incorporating a continuous review (s, S) type inventory policy, where s is the reorder level and S is the maximum stock level. The operating principle for the inventory policy can be stated as follows : An amount of spare units must be ordered from the vendor to bring inventory up to a maximum stock level S when inventory depletes to a reorder level s. In addition, a delivery time to customer i is scheduled for ti days aft er the corr esponding customer orde r issued date. When the scheduled periodic delive ry time arrives, spare units are delivered to customer i in advanc e provided the spar es are available. Otherwise, the spares are delivered as the stock is replenished. I f spa res a re order ed by customer i be for e the scheduled time, the spares will be deliver ed as soon as they are available. Optimal values of the decision variables, (ti, s, and S), are determined by maximizing customer service level and minimizing total inventory cost per period, where the cost components include inventory related costs for both the supplier and customers. Because of the complexity in formulating a mathematical model for this multi-unit inventory environment, a simulation procedure is employed. Although a user-written simulation program in a gener al purpose language like C, FORTRAN, or PASCAL could be more flexible in conjoining ce rtain analytical functions, it is usually prefe rable to use a simulation language such as SLAM, SIMAN, SIMSCRIPT, etc., because of the many built-in functions.jKabir and Farr ash, 1996kFor modeling a complex system like an inventory management system, the use of a simulation language can save time and e ffort needed for modeling and progr am development. jPritsker, 1986k This research will describe the construction of two SLAM network models for the (s, S) policy and (ti, s, S) policy respectively and develop a simulation procedure to select the appropriate inventory policy for the supplier based on the result of simulation experiment. Experimental design is conducted to compare the per formance betwe en (s, S) policy and (ti, s, S) policy. The dominant policy will then be adopted and the level of ea ch decision variabl e leading to the best per form ance will be dete rmined by the result of experimental design. The effects of di fferent fa ctors including cost elements, item demand characte ristics, and lead time distributions will also be considered.. 3.Methology The objective of this resear ch is to develop an appropriat e inventory policy for the spar e part supplier. in semiconductor equipment industry. The inventory policy may be formulated in a mathematical model. The objective function of the inventory policy is to maximize customer service level and minimize total inventory cost. The constraint may include the limitation of maximum stock quantity, the restriction of inventory r eplenishment lead time, the stochastic customer demand to fa ce, the minimum inventory turns to maintain, the minimum customer service level to ful fill, the maximum inventory cost to endure, etc. The solution is to find the decision variabl e of the inventory policy which can achieve the objective function and satisfy the constraint simultaneously. Howeve r, a mathematical model may be diffi cult to formulate to deal with differ ent demand distributions and the optimal solution of the decision variables may be hard to obtain. On the contrary, it is possible to construct inventory policy in a simulation model. The use of simulation to construct inventory model in this resear ch is suitable because it can clearly and fully model the environment and control the factors affecting the inventory system. The continuous or periodical review (s, S) policy is the most common policy adopted widely in industry because it is easy to execute. The task is to monitor the inventory position continuously or periodically. An amount must be ordered to bring inventory up to a desired level S when inventory depletes to a reorder level s. However, this policy may suffe r by overstocking or understocking when customer demand fluctuates intensively. (ti, s, S) policy is the spare parts stocking policy originally adopted by factories which have many types of manufa cturing equipment in the shop floor. Kabir and Farrash [1996] made much improvement on this policy. Their research jointly concerns age replac ement and spar e provisioning by incorporating a continuous review (s, S) inventory policy. The (ti, s, S) policy which this resear ch proposes is diffe rent from the original one. This rese arch attempts to explore the VMI-based (ti, s, S) policy for the spare part supplier that has many customers. The basic idea of this policy is to reduce the risk of facing sudden customer demands or inventory shortages by means of pe riodic delivery in advanc e. The supplier can periodically deliver spare parts to its customers in order to reduc e the dem and uncertainty. Consequently, it gives safety stock more capability to serve as a buffer of dem and fluctuations. Thus, the spare part supplier can prevent overstocking or unde rstocking while storing the minimum stocks needed and maintaining a satisfactory customer service level. The spare part suppliers of semiconductor equipment industry often fac e sudden and unce rtain customer demands. In order to penetrate the market, many high-tech companies in Taiwan tend to provide high customer service level by means of incr easing the sa fety stock level. It results in high inventory cost and interests losses. In view of this, we believes that. 2.

(10) the VMI-bas ed (ti, s, S) policy is appropriate for the spare part supplier in this industry and this resea rch is worth doing.. 4 Research Procedure This research will be proceeded step by step as follows. 1. To consult relevant people of the supplier and collect nec essary dat a for the conveni ence o f identifying key fa ctors of inventory management policy. The historical customer demand patterns, for example, are the important and necessary data for this research. 2. To list key factors which a ffe ct the inventory management system in order to determine the necess ary components in the simulation model. The key factors include controllable and uncontrollable factors for the spare part supplier. Customer demand patterns and lead time patterns are uncontrollable fa ctors, while the inventory policies are controllable ones. 3. To adopt the key fa ctors into appropriate simulation components so as to build the inventory model incorporating the inventory policy. The components include attributes and variables of whi ch the simulation entities carry. Part ID, time to demand, type of supply, replenishment lead time, and so on, are some components to be considered. 4. To build two simulation models, one for the (s, S) policy and the other for the VMI-based (ti, s, S) policy. Because the two inventory policies ar e distinct, two simulation models need to be constructed and run separately. 5. To run simulation using the inventory models constructed in step 4 under SLAM system. The relevant pa ramet ers for the inventory polici es ar e selected in the r ange of m anagem ent speci fic ation. Customer demand and lead time distributions are acquired from fitting the historical data provided by the supplier. 6. To design experiments to analyze the outcome of each simulation run and compare the per form anc e of the two inventory policies. The key per form ance indices of the inventory policies comprise customer service level and total inventory cost. The inventory turns will also be evaluated. 7. To have discussion and make conclusions. From the analysis of simulation outcome, the feasibility of VMI -based (ti, s, S) policy for the supplier will be evaluated, and the appropriat e inventory policy for spare part suppliers in semiconductor equipment industry will then be suggested.. 5 Key Factors and Performance Indices to Inventory Policies. An inventory management system is composed of policies and processes that work together to accomplish the goal of a company. Thus, in dealing with an inventory system, it is significant to first identify pertinent key factors which influence the policies and processes. From the supplier’s point of view, key fa ctors can be divided into two categories, controllable fa ctors and uncontrollable ones. Customer demand patterns and lead time patte rns ar e uncontrollable fa ctors becaus e the supplier cannot dominate customer demands and its vendor’s delivery lead time. Howeve r, the inventory policies are controllabl e factors. The inventory policy is specified by the management and the reason is self-evident. It is significant to choose appropriate per form anc e indices for inventory policies. The conce rns in this research include custome r servic e level and total inventory cost. The inventory turns will also be evaluated. The customer servic e level is measur ed as a whole from the ave rage ful fillment of individual customer orders. If inventory shortages oc cur, this means that the delivery cannot be executed within a predete rmined order promising date or the orde r quantity cannot be met, the customer service level of the supplier drops. Therefore, the customer service level can be de fined as : CSLl. m, where CSL . repr esents customer service level, M represents the number of customer orde rs which are satisfi ed by the supplier in a given period, and O represents total number of customer orde rs which the supplier promises in a given period. Total inventory cost contains inventory related costs for both the supplier and its customers. The major concern of these costs is carrying cost becaus e of the high price char acte ristics of spar e parts for semiconductor equipment. In addition, the inventory car rying cost for customers is also counted in the total inventory cost while adopting the VMI-based (ti, s, S) policy. The conception is that when the supplier delivers spare parts to a customer in advance, the inventory carrying cost of the customer can be treated as a penalty cost for the supplier until the time when the customer really needs it. Inventory replenishment cost and stockout cost are also included in the cost function. Therefore, the cost function c an be de fined as : TClCSnCCnRCnSC, where TC is total inventory cost, CS is inventory carrying costs of the supplier, CC is the inventory carrying costs of customers, RC is inventory replenishment cost, and SC is stockout cost. Inventory turns is a convenient per formanc e measure of how e ffe ctively inventories are being used. It is calculated as : ITlSoI, where IT repr esents inventory turns, S represents the annual cost of goods sold, and I represents average inventory in dollars. Theoretically, the less inventory in the war ehouse, the. 3.

(11) higher inventory turns will be obtained which results in less inventory carrying cost as well. However, customer service level may be degraded while attempting to increase inventory turns. Therefore, this research intends to select a suitable inventory policy1(s, S) or (ti, s, S)2to reduce total inventory cost and satisfy the desired customer se rvice level simultaneously while maintaining the inventory turns at a reasonable level.. 6 Discussion and Conclusion In the industrial case study, a perfo rmanc e comparison between (s, S) policy and the VMI-based (ti, s, S) policy applying experimental design is made. The result presents suffici ent evidence to indicate that (ti, s, S) policy results in higher customer service level and lower total inventory cost than (s, S) policy does under the same stock level. In other words, adopting the policy of pe riodic delivery in advanc e, positive 1. e ffects in the incre ase of customer s ervic e level and in the decreas e of total inventory cost can be obtained. It means that the supplier can improve customer service level and decre ase total inventory cost while maintaining the level of inventory turns without increasing stock level by adopting (ti, s, S) policy. There fore, the VMI-based (ti, s, S) policy is identified as the dominant policy in this case study. Although only three critical spare parts are analyz ed, the inference can be applied to gener al spare parts as long as the demand variations of the spa res a re not too large. A 33 fa ctorial design of (ti, s, S) policy for the three pe r formance indic es is made and the following conclusion is drawn. Both ti and s affect the customer service level but S does not. Furthermore, all three decision variables affect the total inventory cost. In respect to inventory turns, both s and S have significant effect but ti does not. From the result of Duncan’s multiple range test, it reveals that combinations of high s, low S, and low to medium ti will result in high customer service level and low inventory cost. Similarly, as long as s and S are at medium levels, high customer service level and low inventory cost will be gene rated no matte r what ti value is. In general, all three levels of ti will give higher customer servic e level and lower inventory cost if s and S are properly set at the corresponding levels. To sum up, the contribution of this research c an be stated as follows. We propose a VMI-based inventory policy, (ti, s, S), for a spare part supplier in semiconductor industry in Taiwan. The result of simulation experiment concludes that applying periodic delivery in advanc e c an e ffe ctively improve customer service level and decrease total inventory cost while maintaining a desired level of inventory turns. The performance of (ti, s, S) policy proposed by this research is signific antly better than that of (s, S) policy widely used in industry under the sam e stock level. There fore, the VMI-based inventory policy is. favorable. Future resea rch on the inventory policy of spar e part supplier can be focused on : Formulate the impact on the competitors of the supplier since customers may purchase spare parts from a second sourc e if the custome r service level cannot be satisfied. Consider the relationship between customer interarrival time and the corresponding order quantity while describing the customer demand characteristics. Investigate the limitation on different types o f spares and the corresponding customer demand patterns while adopting the VMI-bas ed (ti, s, S) policy. Discuss the degrees of the e ffects from differ ent factors, such as cost elements, item demand characteristics, lead time distributions, etc.. 7. 1.

(12) 50 95 2. /0. !"#$% &'()*+ ,#-.. 12345. 6789:;3<=. >?3@A9BC<=. DEFG. HIJK3LM. *N( ). 2HIOP3Q5. RSTUV( ). 3. V3WIXYZ[ \]. ]. '. ^_. . ^_. . `abcXYdefg 4. V3bh@iZ[j \]. ]. '. kl<mcnopqr9stXYj uvbhwxyz k={@ i @i. |} ~ . 5. Vk. k. 4. . 3 j. 28. 2C . k j

(13) !"# 6. V43<mj jd . 4. V|Q .

(14) Proceedings . W4. k4*NP s W9 IJPE 7. 3¡¢£¤¥

(15) %&3¦V §¨§ ,#45©*Nª«Z[¬®¯ $%&'() *+,-. /0&'"1 23456789:# °±V%&²¬³j\´ ´ ' k µ. 8. References Kabir, A. B. M. Z. and S. H. A. Farrash, “Simulation of an integrated age replac ement and spare provisioning policy using SLAM”, Reliability Engineering and System Safety, 52 (1996), pp129-138. Pritsker, A. A. B., Introduction to Simulation and SLAM, Systems Publishing Corporation, Indiana, Third Edition, (1986) Robin, G., “Pushing the limits of VMI”, Stores, Vol. 77, NO. 3, (1995), pp42-44. 5.

(16)