Experimental Design and Statistical Analysis: Improving Testing of the AN/BQQ-10 Submarine Sonar System

• Activity Number: 370
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Defense and National Security
• Author(s): Laura Freeman*
• Companies: IDA
• Keywords: DOE ; Censored Data ; Regression Analysis ; Department of Defense ; Laboratory Testing

Abstract

In this case study Design of Experiment (DOE) was used to develop the In-Lab Operational Testing of the AN/BQQ-10 Submarine Sonar System. The case study shows how DOE and lab-based, but operationally relevant testing can be used to show improvements in software versions, which have not been possible in the past because of the high variability of at sea measurements. This test used recorded real data played back on two different versions of the sonar system. This new test methodology had several benefits: (1) the laboratory setting allowed for the use of design of experiments to control factors that are traditionally unfeasible to control during an at sea test; (2) the direct comparison between the two systems resulted in demonstrating a statistically significant reduction in the detection time for the new system. Several challenges in the data analysis resulted in the need to incorporate censored data analysis on skewed distributions. This statistical engineering case study brings together DOE, missing data, skewed data, censoring, and inverse prediction to draw defensible conclusions about sonar system performance.