High-dimensional Life-Cycle Modeling for Naval Survey Assessments
Abstract
A conceptual flowchart of the data modeling framework and analytical approach, consisting of three core aspects, is illustrated in F"igure 4 in the proposal. The first and most central aspect is a photorealistic and geometrically accurate spatial point cloud that reflects the geometric configuration and visual appearance of the structure at a given survey interval. Survey information is then em"bedded into this model through pattern mapping, providing spatiotemporal context for the otherwise localized information. The combin"ed result is a HD model that richly contextualizes information in a 3D modeling environment (Task 1). Through spatial analytics and" pattern analysis, survey measurements and changes in the model itself are parametrized and fused (Task 2). The fused information is"" then modeled over time through a combination of unsupervised machine learning and dynamical systems analysis, in order to support l""ife-cycle assessments (Task 3). A research program that explores, prototypes, and validates this survey information modeling concept" is proposed.The research plan includes the following core tasks:1. Creation of the information modeling process2. Formulation of analytics techniques suitable to the now structured information3. Establishment of dynamic models for time-history analyses4. Validation testing on laboratory-scale structural modelsExperimental validation will take place concurrent with the remainder of the" research program, as well as through a separate validation study (Task 4). For each task, representative survey data will be collec"ted on laboratory scale structural components (e.g. simple bending and tensile specimens) to generate data for algorithm prototyping and testing. Laboratory structures of increasing complexity and scale will then be used to provide validation testing. In all exper"iments, a variety of structural configurations will be evaluated. This is essential, as many data analysis techniques are prone to s""tatistical overfitting, working effectively for one specific problem (or structural component) but proving inaccurate or unreliable"" for other components or structural systems. The PI also intends to use information collected during full-scale ship surveys, throug"h an ongoing collaboration with NSWC Carderock researchers.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 06, 2017
- Source ID
- N000141812014
Entities
People
- David Lattanzi
Organizations
- George Mason University
- Office of Naval Research
- United States Navy