Application of industrial pipelines data generator in the experimental analysis: Pipe spooling optimization problem definition, formulation, and testing. (January 2020)

• Record Type: Journal Article
• Authors: AL-Alawi, Mubarak; Mohamed, Yasser; Bouferguene, Ahmed
• Is Part Of: Advanced engineering informatics. Volume 43(2020)
• Publication Date: 2020-01
• Subjects: Branch-and-bound -- Data generator -- Bin packing -- Optimization -- Pipe spooling; Computer-aided engineering -- Periodicals; Engineering -- Data processing -- Periodicals; 620.00285
• DOI: 10.1016/j.aei.2019.101007
• Languages: English
• ISSNs: 1474-0346

Abstract

Abstract: Experimental analysis of algorithm performance can generally be obtained by running the algorithm of interest on a large number of diverse datasets from which statistical information regarding scalability and efficacy are obtained. In addition, these datasets can also be used to gain insight into the impact of a local modification on the global performance of a procedure. However, the main challenge in this area is related to the availability of real-world instance projects from which useable data can be collected. In fact, not only real-life data collection, documentation and management is expensive but more importantly they are generally confidential. As a result, building data simulators capable of generating instance datasets exhibiting features similar to those collected from real-life projects can help alleviate the challenge of availability and confidentiality of data for research. Building on previous work (Al-Alawi et al., 2018), this contribution illustrates the application of the industrial pipelines data generator in the experimental analysis of a pipe spooling optimization problem. The industrial project-based problem in the form of pipe spooling process was defined and projected as a three-dimensional bin-packing class of optimization problem. A branch-and-bound heuristic was proposed to solve the optimization problem and tested on 1000 instance problems generated using the industrial pipeline data generator. Two scenarios were tested the run time performance was reported and recorded as benchmark results for future use.