2019_programme: UNCERTAINTY ANALYSIS IN SONAR PERFORMANCE MODELLING BY LATIN HYPERCUBE SAMPLING

• Session: 20. Sonar performance modeling and verification. Applications to active and passive sonar
  Organiser(s): Colin Mathieu, Prior Mark, Heaney Kevin, Ellis Dale
• Lecture: UNCERTAINTY ANALYSIS IN SONAR PERFORMANCE MODELLING BY LATIN HYPERCUBE SAMPLING
  Paper ID: 808
  Author(s): Sigg Robert, Pihl Jörgen
  Presenter: Pihl Jörgen
  Presentation type: oral
  Abstract: We have studied the possibility to use stratified sampling for uncertainty analysis applied to underwater applications. The method is called Latin Hypercube Sampling (LHS) and the idea behind this approach is to sample only a representative part of the probability distributions so that the number of runs can be reduced significantly. Usually, the number of runs are in the order of 50-100 compared to traditional Monte Carlo sampling where you have millions of runs. An important issue is the difference between uncertainty and sensitivity analysis, which we demonstrate with an active acoustic scenario. Thereafter, we show how we can use LHS in an area where you have quite large uncertainties in the geo-acoustic properties. We have identified the need of presenting the uncertainties to the user in an easy understandable way in a single plot. A new visualization parameter is defined, TPOD (Total Probability of Detection), to meet this requirement. Our studies using LHS show that it is easy to apply and give useful insight in complex uncertainty dependencies. However, for relevant output we need to have good knowledge about the individual parameter probability distributions, both regarding absolute values and representative distributions.
    Download the full paper
• Corresponding author: Dr Pihl Jörgen
  Affiliation: Swedish Defence Research Agency
  Country: Sweden
  e-mail: