<span class="var-sub_title">High-Accuracy Scalable Solutions to the Dynamic Facility Layout Problem</span> SC18 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

High-Accuracy Scalable Solutions to the Dynamic Facility Layout Problem

Authors: Apan Qasem (Texas State University), Clara Novoa (Texas State University), Chandra Kolla (Texas State University), Samantha Coyle (Texas State University)

Abstract: The dynamic facility layout problem (DFLP) is concerned with finding arrangements of facilities within plant locations that minimize the sum of material handling and relocation costs over a planning horizon. DFLP is relevant in manufacturing engineering; accurate solutions can reduce operational costs by as much as 30%. We present a new scalable solution that formulates the task of finding the optimal arrangement as a shortest-path (SP) problem. The new parallel algorithm to find the SP employs a problem-specific heuristic to substantially cut down the search space. Compiler-level optimizations improve the performance across different execution platforms, including an auto-tuning strategy to derive the optimal SMT configuration on a POWER8 system. Results show a factor of 13 speedup over existing methods. For the six-facilities problems the best known solution is reached and for sets with 15 and 30 facilities the solution is within 2.83% and 5.66% of the best solution, respectively.

Best Poster Finalist (BP): no

Poster: pdf
Poster summary: PDF

Back to Poster Archive Listing