Probabilistic Error Estimation in Model-Based Predictions
Abstract
This report was developed under a SBIR contract award for Solicitation topic # N04-137. Modern ship designs can only be successful if the ship can survive in a hostile environment. External threats for a ship can originate from underwater detonations, anti-ship missiles, and even low tech weapons like in the terrorist attack of USS Cole. These threats become even more important as the focal point for naval operations has shifted towards littoral areas, where ships are exposed to a higher risk. Current analysis methods do not provide high level of confidence and large structural design safety factors are used. Thus, ships are heavier and more expensive to construct and maintain than may actually be required. Models that quantify the level of confidence are required in order to provide meaningful and reliable information during the ship design stage. This SBIR project developed a system for probabilistic Error Estimation in model based predictions. It can utilize commercially available non-linear codes for shock analysis (LS-DYNA, ABAQUS, LS-DYNNUSA, etc.) or Navy developed codes (DYSMAS), a limited amount of test data, the level of uncertainty in model parameters and in the test data, in order to update the numerical model for improved probabilistic correlation to the test data. Once the numerical model has been updated, the new development also provides error bounds for the numerical results.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 31, 2004
- Accession Number
- ADA427630
Entities
People
- Nickolas Vlahopoulos