An Intelligent Sensor Fusion Architecture for Navy Auxiliary Systems

The defense and industrial communities are faced with major challenges to monitor, process, and optimize the operation of complex systems/processes. Of particular interest to the Navy are critical shipboard auxiliary systems that support chilled water and electrical processes. The complexity of modern shipboard systems and the requirements for their reliable and safe operation suggest that optimum means must be deployed to make effective use of multiple sensors providing an enormous amount of raw data. Impact Technologies, in collaboration with the Georgia Institute of Technology and its industrial partners, is proposing to develop, test and evaluate an intelligent/hierarchical sensor data fusion architecture to improve the information derived from raw data at various levels of abstraction: the data, feature, sensor and knowledge levels. The data/sensor fusion approach is based on a hybrid analytical/intelligent methodology that exploits the concept of"focus of attention"via active perception in order to optimize auxiliary system component classification accuracy while reducing substantially the computational burden. The proposed fusion scheme incorporates several levels of abstraction: fusion at the data level, the feature level, the sensor level and the knowledge level. The enabling technologies build upon the fundamental premise that fusion is an optimization problem and rigorous computational models, coupled with intelligent search algorithms (Genetic Programming, Particle Swarm Optimization, etc.) are employed to fuse sensor data, resolve conflicts and arrive at a verifiable solution to data fusion. Our targeted testbed is a typical shipboard auxiliary system supported by the water chiller/electrical processes. Our target application domain for the data/sensor fusion algorithms to be developed in this program is equipment damage assessment and process control.