|
Activity Number:
|
65
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 9, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Cmte on Statisticians in Defense and National Security
|
| Abstract - #300287 |
|
Title:
|
Discriminant Analysis Models for Unified Damage Prediction Across Failure Modes
|
|
Author(s):
|
Thomas F. Curry*+ and Elisabetta L. Jerome
|
|
Companies:
|
Northrop Grumman Information Technology and Sverdrup Technology/TEAS
|
|
Address:
|
5450 Tech Center Dr., Colorado Springs, CO, 80919,
|
|
Keywords:
|
predictive modeling ; Operation Iraqi Freedom ; weapons effects ; discriminant analysis
|
|
Abstract:
|
The development of precision guided munitions and the need to minimize collateral damage have brought damage modeling to the forefront of modern warfare. For years, weapons effects modelers have struggled with the problem of predicting damage modes and damage parameters within each mode. The usual solution was to try to predict the failure mode, and then select a model appropriate for that mode. This resulted in discontinuities in the damage prediction as failure modes changed. Incorrect damage model selection, combined with uncertainties within failure modes, frequently caused large prediction errors. An approach is presented that uses discriminant functions to produce an optimized (minimum error) damage model that is continuous across failure modes. Mode transition points are explicitly defined by mode-weighting. Initial comparisons to older models show the discriminant analysis model can reduce damage prediction error by 75%. Warfighters stated the models were very useful in Operation Iraqi Freedom.
|
