The goal of the structural health monitoring (SHM) research at Los Alamos National Laboratory (LANL) is to develop robust and cost-effective SHM solutions by integrating and extending technologies from various engineering and information technology disciplines. More specifically, sensing and data processing hardware must be integrated with data interrogation algorithms. This integration often requires some form of predictive modeling to aid in the definition of the sensing system parameters and to quantify the ability of the data interrogation algorithms to identify damage. To facilitate this integrated approach, the SHM research group at LANL has adopted a statistical pattern recognition paradigm for the development of SHM solutions. This paradigm can be described as a four-part process:
Inherent in Steps 2-4 of the process are Data Normalization (removing the effect of operational and environmental variability from the damage-sensitive features), Data Cleansing (selectively choosing which data to pass on to the feature extraction process) Data Fusion (combining information from multiple sensors in an effort to enhance the fidelity of the damage detection process), and Data Compression (reducing the dimensionality of the feature vector or the volume of data to be stored). These additional tasks can be implemented either in hardware or software and in most cases some combination of these two implementation modes will be used. The LANL staffs believe that this four-step paradigm must be implemented in a manner where approaches to the data acquisition, feature extraction and statistical modeling portions of the process are developed in close coordination.
For more information on this process, please refer to the full statistical pattern recognition paradigm page.
Graduate Students doingthe dirty work.