THE EFFECTS OF PRODUCT STRUCTURE COMPLEXITY ON MRP LOT SIZING

Abstract

1be amount of research on MRP lot sizing is enormous. Despite this, there has been no comprehensive research to test the effects of the product structure complexity on the relative performance of lot sizing rules. The effects of product structure complexity on MRP lot sizing is important because the product structure is an important input into the MRP planning process and may affect the overall system performance. The purpose of this academic exercise is to address this gap in the MRP lot sizing research by conducting an experiment to test extensively the effects of product structure complexity on the relative performance of lot sizing rules in a multi-level environment. Three parameters are proposed to characterise product structure complexity. These parameters are: Number of Items, Number of Levels and Commonality Index. Eleven lot sizing rules are evaluated in the experiment these rules include the Lot-For-Lot, Periodic Order Quantity, Economic Order Quantity, Wagner-Whitin, Part Period Balancing, Incremental Part Period Balancing, Least Unit Cost, Incremental Order Quantity, Groff's marginal rule, Silver-Meal and Bookbinder-Koch. A full factorial simulation experiment with 1980 simulation runs is conducted using the product structure complexity parameters as factors. A computer model is developed to perform this simulation. The model consists of a main program, a routine to generate different problem environments and separate routines for each lot sizing rule. A new index called the Reduction Index is also developed to track the degree of batching in the problems to ensure that the effect of lot sizing in a multi level environment is captured. The experimental results show that the number of items and the number of levels in the product structure have no significant effect on the performance of lot sizing rules. Only the relative cost difference between the rules changes. Relative ranking of the rules do not change across different problem sizes and different number of levels. The commonality index, however, affects the lot sizing rules. Relative cost differences in the performance of the 5 best lot sizing rules tend to reduce as commonality index increases. This shows that differences in lot sizing rule performance is less important when commonality index is high. Among the rules tested, Groff's rule improves with increasing commonality while POQ and LUC become worse. Bookbinder and Koch's rule converges towards the performance of WW. The best lot sizing rule is Bookbinder and Koch, which is consistently the best rule over most of the problems. 1be Wagner-Whitin benchmark, IOQ rule and IPPB rule also perform well. An analysis of the reduction index indicates that lack of batching appears to be the primary cause for bad performance. This suggests the effectiveness of incremental cost rules since they are more likely to batch.