University of Twente Student Theses


Lot Sizing at AkzoNobel Polymer Chemicals: Improving the Quantity and Timing of Production Orders

Hoogen, D.J.F. van den (2010) Lot Sizing at AkzoNobel Polymer Chemicals: Improving the Quantity and Timing of Production Orders.

[img] PDF
Abstract:Introduction At AkzoNobel Polymer Chemicals (ANPC), production takes place at several production sites across the world. Each production site contains multiple Production Units (PUs) that are only suitable for producing a fraction of the complete product portfolio of ANPC. The production process implies changeover and inventory costs, which we define as the lot sizing costs. Currently, ANPC has insufficient insight in the relation between the decision for the lot sizes and the resulting costs: 1. There is no unambiguous rule regarding how to determine the changeover costs. 2. The validity of the methods that support ANPC in deciding on the lot sizes is unknown. In this report we define these methods as the current lot sizing policy. In addition, internal research shows that ANPC scores low on inventory costs with respect to the competition. Moreover, AkzoNobel’s Board of Management announced that ANPC should focus on cost reduction, due to the impact of the credit crunch. For this reason, we study the relation between the decision for the lot sizes and the resulting changeover and inventory costs: We study how we can support ANPC in improving the lot sizes and reducing the lot sizing costs. We reflect this need in the goal of our research: To support AkzoNobel Polymer Chemicals in improving the quantity and timing of production orders, in order to reduce the lot sizing costs. Current situation Every week, the Planning Department (PD) of ANPC determines the Production Plan (PP). This PP indicates the quantity and timing of production runs of product families and is determined based on the current lot sizing policy. ANPC classifies a Stock Keeping Unit (SKU) as one product of the total product portfolio and classifies a product family as multiple SKUs with similar chemical characteristics. When deciding on the PP, the PD tries to ensure that the due dates of customers are respected. At the same time, the sum of the production time and the changeover time is limited by the available production capacity of the PU: A changeover between two product families requires a changeover in the PU. Moreover, the inventory levels cannot exceed the available inventory storage capacity for some SKUs, while it is possible to rent or buy additional storage capacity for other SKUs. Consequently, the problem that ANPC faces when deciding on the lot sizes is defined as the Capacitated Lot Sizing Problem (CLSP) with changeover times. Due to the restrictions and the magnitude of the number of product families that ANPC produces in a PU, developing the PP can be quite a puzzle. The PD uses the PP to develop the Weekly Schedule (WS), on SKU level. Heuristic In order to attain our research goal we develop a heuristic that is based on the Simulated Annealing (SA) algorithm, an iterative improvement algorithm. This heuristic calculates a PP for a PU in which the lot sizing costs are minimized, while the capacity restrictions and due dates are respected. We consider the planning on product family level and assume that all data is timevarying, but known with certainty. In accordance with the requirements from ANPC, our heuristic is easy to understand and execute in practice by employees of ANPC. Results We design six different heuristics that are all based on the SA algorithm and calculate the results for twelve different problem instances. In addition, we design two mathematical models that describe the CLSP with changeover times at ANPC and calculate the PP for every problem instance by using optimization software for solving the mathematical models. Next, we compare this output with the output of the heuristics: We compare the lot sizing costs. Unfortunately, the optimization software does not attain the optimal solution for most problem instances. Yet, for two problem instances, which contain five product families, the optimization software does attain the optimal solution. The heuristics approach the optimal solution to respectively 3% and 6%. 3 One heuristic outperforms the other five heuristics. This heuristic, which we define as KK-NS1, first develops a starting solution based on the algorithm of Silver and Meal (1973) and Kirca and Kökten (1992). Next, the heuristic searches for improvements via the SA algorithm by iteratively rearranging two succeeding runs for a product family. The results show that the performance of the heuristic decreases when the number of product families, which need to be included in the PP, increases. In addition, the performance decreases when the demand level increases. Next to assessing the performance of the heuristic by using the output of the optimization software, we analyze whether the heuristic could be used to improve the lot sizing at ANPC: We use the actual data of four PUs and calculate the PP by using KK-NS1. The heuristic attains a valid PP for two problem instances. Following, we compare the lot sizing costs for these valid PPs with the current lot sizing costs: We conclude that the lot sizing costs in both PUs could be decreased with respectively 45% and 15%. For the other two PUs, the heuristic did not attain a valid planning. Yet, we present several possibilities for realizing a valid planning by adapting the demand data. After adapting the demand data, we indicate that the current policy could be improved by 30% and 45% respectively. Still, in practice, adapting the demand could lead to additional costs: We cannot quantify the improvements for these two PUs. Conclusions The results show that our heuristic could be used in decreasing the total lot sizing costs at ANPC. Since the results show a possible cost reduction for two PUs, we recommend that the heuristic is implemented in other PUs as well. There are 31 other PUs around the globe in which the heuristic can be applied. As such, we anticipate that the contribution that the heuristic can deliver, with respect to decreasing the lot sizing cost, is promising. In order to support the implementation of the heuristic, we develop a manual and provide functional training for the end-users.
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:85 business administration, organizational science
Programme:Industrial Engineering and Management MSc (60029)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page