A Computational System to Solve the Warehouse Aisle Design Problem
Metadata Field | Value | Language |
---|---|---|
dc.contributor.advisor | Smith, Alice | |
dc.contributor.author | Ozden, Sabahattin Gokhan | |
dc.date.accessioned | 2017-05-17T20:08:58Z | |
dc.date.available | 2017-05-17T20:08:58Z | |
dc.date.issued | 2017-05-17 | |
dc.identifier.uri | http://hdl.handle.net/10415/5761 | |
dc.description.abstract | Order picking is the most costly operation in a warehouse. However, current warehouse design practices have been using the same design principles for more than sixty years: straight rows with parallel pick aisles and perpendicular cross aisles that reduce the travel distance between pick locations. Gue and Meller (2009) altered this design mentality by proposing a fishbone layout that offered reductions in travel distance of up to 20% in unit-load warehouses. However, research in finding alternative layouts for order picking warehouses is lacking. The main result of this dissertation is to show that there are non-traditional designs that reduce the cost of order picking operation by changing the aisle orientation, aspect ratio, and placement of depot simultaneously. We present designs that achieve reductions in travel distance of up to 5.3%. Order picking warehouse layout optimization is computationally complex. Three problems need to be solved: layout design, product allocation, and pick routing. In this dissertation, we focus on layout optimization but we propose new methods for product allocation, routing, and certain speed-up techniques for routing algorithms. We need to solve multiple traveling salesman problems (TSP) to find the expected distance traveled for order picking. We develop parallel and distributed computing techniques to solve large batches of TSPs simultaneously. We compare two well known TSP solvers and various machine settings (serial, parallel, and distributed). Distributed computing techniques only show their real benefits when the TSP instances have more than 50 locations so that the network and file read/write overhead is relatively low. Our results also show that for both real data and generated data, a scheduling algorithm like longest processing time performs better than a naïve method like the equal distribution rule even though the method used for estimating processing times of TSPs is crude (TSP size, in this case). In all of the layout literature, simulation and analytical models assume a simple travel rule: order pickers follow the aisle centers. Following aisle centers leads to longer travel distances when an order picker picks items within the pick aisles that have angles other than 90 degrees between cross aisles. By using a visibility graph, we show that paths are more reasonable in most layout settings, and comparisons between traditional and non-traditional layouts for order picking operations are affected. Our results show that the visibility graph method impacts the assessment of well-known non-traditional layouts compared to traditional counterparts. Moreover, it also changes the rank order of the three most common traditional layouts. Most importantly, we develop a warehouse layout optimization system that models layouts, allocates products to storage locations, calculates routing distances, and performs heuristic optimization over a comprehensive set of layout design parameters. The system searches over 19 different design classes simultaneously by using a layout encoding. We propose improved non-traditional aisle designs for different pick list sizes and demand skewness. The proposed designs can shorten the average walking distance up to 5.3% compared to traditional layouts. | en_US |
dc.subject | Industrial and Systems Engineering | en_US |
dc.title | A Computational System to Solve the Warehouse Aisle Design Problem | en_US |
dc.type | PhD Dissertation | en_US |
dc.embargo.status | NOT_EMBARGOED | en_US |
dc.contributor.committee | Gue, Kevin | |
dc.contributor.committee | Murray, Chase |