This paper considers the scheduling of several different items on a single machine, in literature known as the economic lot scheduling problem, ELSP. One of the characteristics of this problem is that the demand rate is deterministic and constant. However, in a practical situation demand usually varies. In this paper we examine if a deterministic model can be used if demand is stationary stochastic. A dynamic programming approach from Bomberger (Manage. Sci. 12(11) (1966) 778) and a heuristic method from Segerstedt (Int. J. Production Econom. 59(1–3) (1999) 469) are used to calculate lot sizes for four items. The production of these items is simulated with different variations in demand rates. Our conclusion is that a deterministic model of this kind can be used in a practical situation where the demand rate is stationary stochastic, but the models must be complemented by a decision rule; which item to produce and when to produce it. In our tests the heuristic method and the dynamic programming approach perform rather similarly with respect to costs and inventory levels, but the dynamic programming approach results in more backorders when there is small variation in demand rates. This study indicates that the model used for determination of lot sizes is of less importance than the decision rule used for identification of the item to produce and when to produce it.
A heuristic scheduling policy is introduced for a multi-item, single-machine production facility. The scheduling policy uses the presumed optimal order quantities derived from solving an Economic Lot Size Problem and checks that the quantities obtain a feasible production schedule according to current inventory levels and expected demand rates. If not, the scheduling policy modifies the order quantities to achieve a possible solution without shortages. The scheduling policy is inspired by modification of the similar heuristic Dynamic Cycle Lengths Policy by Leachman and Gascon from 1988, 1991. The main characteristics of this scheduling policy are successive batches of the same item are treated explicitly, due to that it is quite possible that one item be manufactured several times before one other item is manufactured once more; the batches are ordered in increasing run-out time; if the existing situation creates stock-outs with ordinary order quantities, then the order quantities are decreased with a common scaling factor to try to prevent inventory shortages; in case the decrease of the order quantities changes expected run-out times, the batches are reordered after new run-out times; no filling up to an explicit inventory level is done, the filling up is done by the desirable order quantity; to prevent possible excess inventory the policy suggests time periods where no production should be performed. The scheduling policy contains no economical evaluation; this is supposed to be done when the order quantities are calculated, the policy prevents shortages and excess inventory. A numerical example illustrates the suggested scheduling policy. Finally, it is discussed as to how the policy can also take into account stochastic behaviour of the demand rates and compensate the schedule by applying appropriate safety times.
This paper considers an inventory control system, primarily for a finished goods inventory. The purpose is to create a procedure that can handle both fast-moving items with regular demand and slow-moving items. The suggested procedure should be easy to implement in a modern computerized ERP-system. Essentially, the system is a periodic review system built around a Croston forecasting procedure. An Erlang distribution is fitted to the observed data using the mean and variance of the forecasted demand rate. According to probabilities for stock shortages, derived from the probability distribution, the system decides if it is time to place a new order or not. The Croston forecasting method is theoretically more accurate than ordinary exponential smoothing for slow-moving items. However, it is not evident that a Croston forecasting procedure (with assumed Erlang distribution) outperforms ordinary exponential smoothing (with assumed normal distribution) applied in a “practical” inventory control system with varying demand, automatically generated replenishment, etc. Our simulation study shows that the system in focus will present fewer shortages at lower inventory levels than a system based on exponential smoothing and the normal distribution.
Polarica is a company in northern Europe buying, refining and selling wild berries and other specialty foods. During the last couple of years the volume of wild berries, mostly blueberries, has increased a lot. This expansion forces Polarica to consider investments in freezing-in capacity and cold-storage capacity. A simple heuristic model was constructed, from which it was concluded that additional volumes of frozen blueberries require more storage facilities and it was also recommended that the location should be based on load-distance analysis. The way this problem is tackled and solved can be copied and hopefully it presents ideas for other similar studies.
Croston (1972) presented an idea and method to separate ordinary exponential smoothing in two parts; in the time between demand, or withdrawals, and demand size. The forecasts are then updated only when there is a demand. Syntetos and Boylan (2005) recommended an adjustment of the Croston method due to a systematic error notified by Syntetos and Boylan (2001). Levén and Segerstedt (2004) suggested a modification of the Croston method where a demand rate is directly calculated when a demand has happened. In this paper real demand data is used to compare these variants of the Croston method. The idea with the modification of Levén and Segerstedt (2004) is that time between demand and demand size is not independent. But this modification has shown poor results. Therefore Wallström and Segerstedt (2010) suggest another modification, a “forward coverage” different to the previous “backward coverage”. Teunter et al (2011) suggest another method, a combination of updates every time period, for estimating the probability of a demand occasion, and every time a demand occurs for estimating demand size. All these different techniques are tested here. The different tech-niques are compared with Mean Squared Error (MSE), Cumulated Forecast Error (CFE) and with the new bias measure “Periods in Stock” (PIS). The tests show that modified Croston with “forward cover-age” in most cases seems to be to prefer compared to “backward coverage”, but still it overestimates demand. Ordinary Croston may be to prefer; therefore our tests show that Croston according to Syntetos and Boylan and also Teunter et al show a tendency to underestimate demand.
An often-adopted technique for short-term forecasting is the single exponential smoothing (SES); it is available in most computer systems for material- and production control. With help from it, it is possible to update continually reorder points for an efficient inventory control. However, the ability for SES to forecast an item when the forecasting time periods often have zero demand is questioned; i.e. slow moving items or when demand is intermittent. Croston presented an idea and method to separate ordinary exponential smoothing in to two parts; the time between demand (withdrawals), and demand size. The forecasts then update only when there is a demand. Since then modifications of Croston’s idea have been suggested and also an idea to measure the fraction of periods with zero demand compared to non-zero demand, i.e. techniques to handle intermittent demand. From the beginning four different suggestions to treat intermittent demand are tested. The tests showed that some complementary modifications were interesting to investigate. The different techniques are compared with Mean Squared Error (MSE), Cumulated Forecast Error (CFE) and with a new bias measure “Periods in Stock” (PIS). Our tests show that Croston’s original idea may be to prefer; some techniques overestimate and others underestimate demand in certain circumstances; and one technique is not to prefer at all.