DEPARTMENT OF MANAGEMENT SCIENCE MScs in Management Science 2007-08 MRPII - Manufacturing Resource Planning Introduction Statistical inventory control, which is concerned with economic lot sizes and probability of stockout, is inappropriate in situations such as the control of components used in the manufacture of a finished product. At least three demonstrations of this can be presented. First, consider a product, which costs £30, whose demand in the next ten weeks is 60. Assume also that the set-up cost for production of this product is £25 and that the inventory carrying cost is 20% per annum. A straightforward application of the formula for the Economic Batch Quantity (EBQ), assuming a 50 week year, gives a production quantity of 50. One assumption behind the EBQ is that the demand is constant, in this case 6 per week. If, however, the demand pattern over the next 10 weeks is 20-0-20-0-00-0-0-0-20, production of 50 units will lead to a 'remnant' of 10 units being carried for 7 weeks without any purpose, as production will be required in the 10th. week. Similarly, a demand pattern of 20-0-40-0-0-0-0-0-0-0 would imply that the EBQ will fail to cover the third week's requirements. Therefore, in situations of 'lumpy' demand it would seem that alternatives to the fixed lot size produced by the EBQ would be more appropriate. Second, to demonstrate how 'lumpy' demand may arise, consider the production of four tools from two forgings, which are both made from a particular steel. Tool 1, with a smooth demand of one per week, and Tool 2, with a smooth demand of six per week, are both made from Forging A. Tools 3 and 4, with smooth demands of three and seven per week respectively, are each produced from Forging B. If we suppose that the EBQ for the four tools are 5, 15, 10 and 25 respectively, then Tool 1 will be produced in weeks 1, 6, 11 etc., Tool 2 will be produced in weeks 1, 3, 6, 8, 11 etc., Tool 3 will be produced in weeks 1, 4, 7, 11 etc. and Tool 4 will be produced in weeks 1, 4, 8, 11 etc. The implication for the production of the forgings, assuming lot sizes of 25 and 50 for Forgings A and B respectively, is that Forging A will be produced in weeks 1, 3, 6 etc. and Forging B in weeks 1, 4 etc. The demand for the steel will therfore be 75-0-25-50-0-25 in the first six weeks, and the smooth constant demand of the four final products has produced 'lumpy' demand for the raw material. Third, if the probability of having one item in stock when required (the service level) is 90%, then two components, required simultaneously for the same end item, will both be in stock with a service level of 81%. If ten components are required, the odds against all of them being available has risen to 2:1 (34.8%). Even if the service level is set at 95%, the probability of simultaneous availability of ten components is 60%, and for 14 components it drops below 50% (48.8%). Thus, when components are ordered, independently of each other, their inventories will tend not to match assembly requirements, and the cumulative service level will be significantly lower than the service levels of the parts taken individually. An alternative to statistical stock control, therefore needs to be used when item demand is dependent. The alternative is Material Requirements Planning (MRP). The principle of MRP is that a time phased schedule of requirements over the, pre-determined, planning horizon, is generated for each component/sub-assembly/raw material on the basis of independent forecasts of demand for the end products. In a full requirements planning system, such schedules of item demand will take account of consolidated demand for common components existing physical stock levels, production/purchasing lead times and production batching policy. MRP takes no account of available capacity, a defect which MRPII seeks to rectify. Inputs to an MRP System 1. The Master Production Schedule The overall plan of production during the planning horizon is detailed in the master production schedule. The planning horizon normally equals or exceeds the cumulative purchasing and production lead time for components of the items in question. It will therefore include time to assemble the product, time to manufacture the components, time to purchase raw materials, and, often, time to allow the purchasing department to consolidate orders to take advantage of opportunities for bulk purchasing. In practice this frequently will require an horizon of a year to eighteen months. It is not possible to allow no changes to be made over this period of time, but it is important that changes in the immediate future of the schedule are kept to a minimum. For this reason, the planning horizon is often split into time zones, with different constraints on the changes that can be made. In the immediate future, in which finished products are currently being assembled, the schedule should only be changed in emergency, since the component parts will have already been manufactured to schedule. Moving forward in time, to the period in which parts are currently being manufactured, it may be possible to make some changes, for example to the sequence of manufacture. In the part of the horizon in which orders for materials have been placed on suppliers, it may be possible to alter production quantities, if the consequent changes to orders can be made. Finally, in the last time zone it will normally be possible to make changes without to much difficulty. In addition to the sales of finished products, a company may also have demand for spare parts; demand which is independent. Such demand can be incorporated into an MRP system by adding them to the dependent demands in the appropriate time period. The essential purpose of an MRP system is to translate the schedule into individual component requirements for specific time periods. In order to take into account the different time zones, it is possible to have variable length time periods. In the first time zone the time period (or time bucket) may be a day, in the second time zone a week may be more appropriate, in the third time zone a month might be used, and, finally, in the fourth time zone, a quarter may be used. 2. The Inventory Record File The individual item stock records are kept in a file which is kept up-to-date by the posting of transactions (e.g. stock receipt, disbursement, scrap, etc) which change the status of the respective items. An important aspect of an inventory record for an MRP system is that it is time-phased, e.g. information on either specific dates or planning periods with which the respective quantities are associated is recorded and stored. In many stock recording systems the status of a given item is normally shown as consisting of only the quantity on hand and the quantity on order. In addition to status data, the inventory records also contain "planning factors", used principally for determination of the size and timing of planning orders, such as lead time and scrap allowance. 3. Bill of Material file The B-O-M file, as it is known, contains information on the relationships of components and assemblies. Its conventional representation is shown in Figure 1 where the product in question (the parent item) is assembled from four different parts. It uses only one of each of three of the components, but uses two of the fourth. End Product | Part A Quantity 1 Part B Quantity 1 Part C Quantity 2 Part D Quantity 1 Figure 1: Bill of Material When the individual bills defining a product are linked together they form a hierarchical pyramid-like structure, and the levels emerge into view. The items on the highest level in a B-O-M file are the end products. Calculation of Requirements 1. Gross requirements The gross requirements for an item equals the quantity of demand for that item. It is the quantity of the item that will have to be disbursed to support a parent order, rather than the total quantity that will be consumed by the end product. These two quantities may or not be identical. Assume that 50 lorries are to be produced and in stock there are 2 transmissions 5 gearboxes 7 gears 20 forging blanks The simple product structure (B-O-M) is illustrated in Figure 2 Lorry | Transmission | Gearbox | Gears | Forging blanks Figure 2 Product structure The gross requirements of the forging blank are calculated as follows: No. of lorries to be produced Quantity of forging blanks needed Stocks of gears containing forging blanks Stocks of gearboxes containing gears Stocks of transmissions containing gearboxes 50 50 7 5 2 14 36 Gross requirement for forging blanks. If there are multiple sources of demand, and therefore of gross requirements, they are combined and summarised, by planning period, in the gross requirements schedule which is represented by a row of time-buckets as in Figure 3. Period Gross requirements 1 2 15 3 10 4 5 22 6 7 13 8 12 Total 72 Figure 3 Gross requirements schedule 2. Net requirements The net requirements of an item are simply found by allocating the available inventory on hand and on order to the gross requirement. So, in the lorry example, the net requirement of forging blanks is 16 (Gross requirement - stock). In the example given in Figure 4 there are 18 units of this item on hand and 24 on order, due in period 4. This information can be displayed in time bucket format as in Figure 4 and the net requirements calculated. Period Gross requirements Scheduled receipts On hand 18 Net requirements 1 2 15 3 10 4 5 22 6 7 13 8 12 Total 72 24 11 12 30 24 7 Figure 4 Net requirements schedule 3. Explosion of requirements The 'explosion' of requirements from the master production schedule down into the various component-material levels, is guided by the logical linkage of inventory records. Net requirements for the high-level items are covered by planned orders. The quantity and timing of these planned orders determine, in turn, the quantity and timing of component gross requirements. This procedure is carried out for items on successively lower levels until a purchased item is reached. The computation of requirements is complicated by six factors: 1. The structure of the product, containing several manufacturing levels of materials, component parts and subassemblies. 2. Lot sizing, i.e. ordering inventory items in quantities exceeding net requirements, for reasons of economy or convenience. 3. The different individual lead times of inventory items that make up the product. 4. The timing of end-item requirements during a planning horizon of, typically, a year's span or longer, and the recurrence of these requirements within such a time. 5. Multiple requirements for an inventory item due to its 'commonality' i.e. usage in the manufacture of a number of other items. 6. Multiple requirements for an inventory item due to its recurrence on several levels of a given end product. Lot sizing Much of the Operational Research literature related to MRP systems is devoted to discretedemand, time-series lot sizing. Some of the new techniques which have been developed are described and evaluated in this section, along with the more traditional approaches to lot sizing. Given a particular net requirements schedule the question arises as to what ordering schedule should be followed. 1. Some simple approaches A fixed batch quantity may be specified and determined arbitrarily, or, if determined using cost factors an economic batch quantity (EBQ) can be calculated. This will be determined by using the 'square-root' formula, but the known future demand, rather than historical demand, can be used. In contrast to all the other approaches that will be described these two approaches create 'remnants', i.e. quantities that could be carried in stock for some length of time without being sufficient to cover a future period's requirements in full. (NB. If there is a net requirement during a period that is higher than the order quantity, the order quantity is increased to the net requirement). All other approaches are discrete lot-sizing techniques because they generate order quantities that equal the net requirements in an integral number of consecutive planning periods. The simplest is lot-for-lot ordering when the planning order quantity always equals the quantity of the net requirements being covered. Fixed period requirements is equivalent to the primitive rule of ordering 'X months' supply except that here the supply is determined not by forecasting but by adding up discrete future net requirements. The period of coverage every planned order should provide is specified. Lot-for-lot ordering is, obviously, a particular case of fixed period requirements lot sizing. Figure 5 illustrates the effect of these different policies for a particular net requirements schedule. For the calculation of the EBQ the set-up (A) is £100, the unit cost (C) is £50, the carrying cost (I) is £0.24 per annum and D = 200 is obtained by annualizing the 9 month demand. It is assumed that the fixed period is 2. Net requirements Fixed batch quantity Economic batch quantity Lot-for-lot Fixed period requirements 1 35 60 58 35 45 2 10 10 3Month4 40 60 58 40 40 5 6 20 20 25 7 5 8 10 5 58 10 40 9 30 60 30 Total 150 180 174 150 150 Figure 5 Some simple lot-sizing approaches 2. Period order quantity (POQ) POQ is based on the logic of the EBQ, modified for use in an environment of discrete period demand. The EBQ is calculated, as described in the previous section and the number of orders per year calculated. The number of planning periods during a year is then divided by this quantity to determine the ordering interval. So the POQ technique is identical to the fixed period requirements approach, except that the ordering interval is computed, and is more effective than the EBQ, as set-up cost per year is expected to be the same but the carrying cost will tend to be lower, as 'remnants' are avoided. However, the predetermined ordering interval might prove inoperative, if several periods show zero requirements and force the technique to order fewer times per year than intended. Using the previous example with D = 200 and Q* = 58 the number of orders per year is 3.4 and the ordering interval is 3.5 months. The effect of this policy, (assuming the interval alternates between 3 and 4 periods), is shown in Figure 6. Net requirements Period Order Quantity 1 35 85 2 10 3Month4 40 5 6 20 35 7 5 8 10 9 30 30 Total 150 150 Figure 6 POQ 3. Least unit cost and least total cost approaches These approaches, and those that follow, all share the assumption of discrete inventory depletions at the beginning of each period, i.e. a portion of each order, equal to the net requirements in the first period covered by the order, is consumed immediately upon arrival in stock and thus incurs no inventory carrying charge. Inventory carrying cost, under these lot-sizing approaches, is computed on the basis of this assumption rather than on average inventories in each period. In determining the order quantity, the least unit cost (LUC) technique calculates the 'unit cost' (i.e. set-up plus inventory cost per unit) for each of the order quantities obtained by covering the next period's requirements, the next two periods' requirements and so on. The one with the least unit cost i chosen to be the lot size. The calculation of the first lot in the example is as follows. (A = £100, IC = £1 per unit per month). The results of the whole calculation are given in Figure 7, and illustrate the limitation of the approach in that only one lot is considered at a time. Tradeoffs between consecutive lots can sometimes be made that would reduce the total cost of two or more lots. For instance, if the requirement in period 7 were added to the second lot, its inventory carrying cost would increase by £15, but that of the third lot would decrease by £40. Net requirements LUC 1 35 45 2 10 3Month4 40 60 5 6 20 7 5 45 8 10 9 30 Total 150 150 Figure 7 LUC The least total cost approach (LTC) attempts to overcome this flaw in LUC logic, on the basis that the total cost for all lots within the planning horizon will be minimised if the setup cost per unit and the carrying cost per unit are as nearly equal as possible (as in the classical EBQ approach). The approach involves calculating the 'economic part-period' (EPP) factor, which is defined as that quantity of the item which, if carried in stock for one period, would result in a carrying cost equal to the cost of set-up. It is computed simply by dividing the inventory carrying charge per unit per period (IPC) into set-up cost. In our example: EPP = A/IPC = 100 The order quantity selected by the LTC approach is that for which the part period cost most nearly equals the EPP. The calculation for the first lot in the example is Month 1 2 4 Net Requirements 35 10 40 Part-periods(Cumulative) 0 10 130 So the quantity chosen for the first lot would be 85, because the 130 part-periods that it would cost most nearly approximate the EPP of 100. The second order, 65, would cover the requirements of months 6-9. The LTC logic has a flaw of its own because the premise that 'the least total cost is where the inventory cost and setup cost are equal' holds true for the EBQ but not for a discrete lot-sizing approach which assumes that inventory depletions occur at the beginning of each period. 4. An incremental heuristic This heuristic simply sequentially increases the lot size by the requirements of successive periods until the incremental cost of carrying the next period's requirements in inventory exceeds the cost of set-up. Month 1 2 4 Net requirements 35 10 40 Inventory cost 10 120>100 so set-up in period 4 The use of this heuristic indicates that lots be produced in periods 1,4 and 9. 5. The Silver-Meal heuristic The objective of this algorithm is to minimise the total relevant costs per unit time. The approach is similar to the least unit cost calculation, but this time the total relevant costs are divided by the number of time periods, rather than the lot size. The lot size is the sum of the requirements in all periods up to and including that period for which the total cost per unit time is a minimum. The calculation of the first lot in the example is as follows: Month 1 2 4 Net requirements 35 10 40 Inventory cost 0 10 120 Total cost 100 110 230 Total cost/No of periods 100 55 57.5 Thus the first lot will be for the demand in periods 1 and 2, and is of size 45. The use of this heuristic indicates that lots will be produced in periods 1,4 and 8. 6. Comparison between the various approaches Every one of the lot-sizing approaches reviewed above suffers from some deficiency. Furthermore any comparison based on an example, such as the one that has been considered, is meaningless because a change in the data can produce a different order of performance. For the example, the results given below are obtained (all inventory carrying costs are based on discrete depletion at the beginning of the period) . If the set-up cost were £300, however, the POQ would outperform LTC and match LUC in effectiveness. In fact, the requirements data can be changed to produce practically any results desired. Therefore, there does not seem to be one 'best' lot-sizing algorithm that could be selected for a given manufacturing environment, for a class of items, or even a single specific item. If set-up costs are low then the lot-for-lot approach may as well be used and in cases of significant set-up cost LUC, LTC or POQ may be chosen. Algorithm Silver-Meal Incremental LUC LTC Fixed period requirements POQ EBQ Fixed batch quantity Lot-for-lot 7. Set-up cost 300 300 300 200 400 300 300 300 700 Carrying cost 95 105 120 245 45 155 206 220 - Total cost 395 405 420 445 445 455 506 520 700 When to use the heuristics If there is no variability of demand then it doesn't make sense to use an algorithm that allows for such a situation. Clearly the variability should exceed some threshold value before a heuristic is used. Silver suggests that the Silver-Meal heuristic should be used when a Variability Coefficient (VC) is greater than one quarter, where VC is the variance of demand per period divided by the square of the average demand per period, ie. N N VC = ((N D 2i ) / ( D i ) 2 ) 1. i=1 i=1 N In the example, N Di 150 and D i=1 i=1 2 i 4350 So VC = (9 x 4350/(150)2 )-1 = 0.74. As this is greater than one quarter, it would be sensible to use the heuristic in this case. System Outputs An MRP system can provide a great number of outputs because the data base, and the inventory status records in particular, contain a wealth of information that provides an opportunity for extracting or further processing the data. Four examples are given here. 1. Inventory order action These types of output are generally self-explanatory. Outputs for inventory order action are based primarily on planned orders becoming mature for release. The MRP system detects such orders by examining the contents of planned-order release buckets in the time-phased inventory records. Other types of inventory order action are increases, reduction, and cancellations of order quantities. 2. Replanning order priorities When there are cases of divergence between open-order due dates and dates of actual need, as indicated by the timing of net requirements, an MRP system has the capability to indicate precisely by how many time periods each item affected should be rescheduled, and in what direction. 3. Safeguarding priority integrity To keep priorities honest, the master production schedule must reflect the realities of production, i.e. it must not contain end product requirements that cannot be met for lack of capacity, material or lead time. Some companies use reports in this category to provide guidance in accepting customer orders for guaranteed delivery. Such reports are generated by 'trial fit' of the order into the master production schedule, and then letting the MRP system determine component-material and lead-time availability. If the order does not fit, the report indicates a best delivery date alternative. 4. Capacity requirements planning Outputs for purposes of capacity requirements planning are based on quantities and due dates of both open and planned shop orders, which serve as input to the capacity requirements planning (or loading) system. The MRP system makes it possible for the load report to be complete, valid, and extending far enough into the future to allow capacity-adjustment action to be taken in time. To keep the load projection up-to-date and valid, it must be repeatedly recomputed as the order schedules in the MRP system change.