Two new batch construction heuristics called K-means Batching (KMB) and Self-organisation Map Batching (SOMB) are developed and verified by simulation experiments. Both KMB and SOMB have a preferment of superior performance in total travel distance and average picking vehicle utility, and even a conspicuous improvement in total CPU running time. Besides, this paper investigates the overall performance of order picking systems integrating storage assignment, order batching and picker routing to find the optimal policy combinations under different order types. The sensitivity analysis is performed to distinguish the relative importance of the various strategies to enhance the performance of operations management.
In the current fluctuant market, how to use limited resources to satisfy the needs of customers has become the key to enterprises' success in sustainable operation. The development in information technology and the advancement of the Internet have not only speeded up the global competition for local industries, but also shortened the life cycle of products. The market became more fluctuant and difficult to predict. To quickly adjust to the needs of customer and adapt to the changes, enterprises must utilise their distribution centre to integrate and connect all the partners on the supply chain, allowing the products to be delivered to the customer even faster. Amongst the internal operation processes in the distribution centre, the order picking process is the core procedure, taking up over 55% of the storage operation cost (Bartholdi and Hackman, 2006). Recent trends show that customer orders changed from few kinds and large quantity orders to many kinds and small quantity ones, which need to be picked and distributed in short time. These changes require efficient and effective order picking systems in warehouses for companies to remain competitive. It was discovered that all the related strategies including storage assignment, order batching and the picker routing would affect the performance of order picking systems, and the order batching is a key strategy especially. Therefore, appropriate order batches can enhance the performance of order picking procedures to effectively reduce the cost of operations management, and then improve the quality of service.
It is believed that if focus on the layout design, storage assignment, order batching and picker routing strategies separately, and discuss the optimisation of either, it often causes local optimisation of the order picking system. De Koster et al. (2007) give a comprehensive literature review on these topics. They indicated the key factors affecting the performance of order picking systems include the layout of the warehouse, the storage strategy, the order picker routing policy, the zoning method and the order batching policy. Petersen and Aase (2004) pointed out, overly complicated picker routings or storage assignments would cause unnecessary waste of costs, their experiment also showed that order batching instead of picker routing planning and storage assigning, has the most prominent effects on decreasing the order picking time; however that research used a traditional warehousing environment as its simulation environment. Ruben and Jacobs (1999) indicated that the methods used for constructing batches of orders and for assigning storage space to individual items can significantly impact order retrieval efforts in warehouses. Thus, the objectives of this study are as follows:
(1)
To develop two new batch construction heuristics called K-means Batching (KMB) and Self-organisation Map Batching (SOMB), since both the amount of stock keeping units (SKUs) in an order and the variation of the order size for an order picking system will affect the order batching procedure. This study contemplates the effects of KMB and SOMB by further comparisons amongst KMB, SOMB, Association rule and Particle Swarm Optimisation Batching Method (PSOBM) under different order types. The Association rule and PSOBM are suggested for order batching by the literature.
(2)
To ensure the feasibility and preciseness of SOMB, the optimal weights for applying SOMB are found to further improve the solution quality of SOMB.
(3)
To evaluate the performance of order picking systems integrating different policies for storage assignment, order batching and picker routing under small, medium, large and mixed order environments, respectively. The optimal policy combinations under different order types are found by the simulation experiment.
(4)
Lastly, sensitivity analysis is performed on order type, storage assignment, order batching and picker routing. Based on the worst policy combination, this paper analyses the improvements made by varying one strategy at a time.
