Statistical Validation of Shipyard Scheduling Software

Oliver Dain, Matthew Ginsberg, Erin Keenan, Iain Pardoe, John Pyle, Tristan Smith, and Andrew Stoneman

Abstract

The ARGOS scheduling system is designed to reduce labor costs in large facilities such as shipyards. Theoretical results suggest that ARGOS is capable of reducing total labor costs for a large ship construction project by between 7 and 14 percent. SimYard is a stochastic shipyard simulation tool that can model a wide variety of shipyard production conditions, including problems that arise and how the shipyard reacts to those problems. It was built to evaluate ARGOS under actual production conditions to validate labor savings estimates. This article describes the major statistical hurdles involved in the design and subsequent validation of SimYard. The main challenge is to take performance data describing how a particular shipyard performs and tune SimYard to behave accordingly; this requires a model of SimYard itself. This problem can be framed as one of multivariate calibration, which can then be tackled using inverse regression methods. Predictive sampling from the resulting model provides an appropriate adjustment for statistical uncertainty, and shows that ARGOS schedules can be expected to save between 7 and 13 percent on a large ship construction project, confirming the theoretical results.

Keywords: calibration, inverse regression, predictive sampling, stochastic simulation

Back to Research page.


Send me e-mail at ipardoe at lcbmail.uoregon.edu

Last updated: November 16, 2006


The views and opinions expressed in this page are strictly those of the page author. The contents of this page have not been reviewed or approved by the University of Oregon.

© 2006, Iain Pardoe, Lundquist College of Business, University of Oregon

Valid XHTML 1.0! Valid CSS!