Real-time, embedded diagnostics and prognostics in advanced artillery systems

This paper explores an integrated modeling and reasoning approach to real-time, embedded diagnostics and prognostics called the Armament Diagnostic And Prognostic Tool (ADAPT). In addition, an approach for using the real-time diagnostic and prognostic information for degraded operation control of armament systems is described. The application focus of this paper is on advanced armament system gun mounts; however, the ADAPT approach has general applicability to a large class of complex systems. It is powered and enabled by the integration of three modeling and reasoning technologies Prognostics Framework (PF) model-based reasoning, Statistical Network (StatNet) modeling, and a time domain gun mount simulation. The model embodied in the PF reasoning is called a fault/symptom matrix, which is a connectivity matrix that represents the linkages of anomalies or faults (rows in the matrix) to observable measurements and the coverage of tests that pass or fail (columns in the matrix). StatNet is a modeling algorithm in the ModelQuest Analyst data mining tool. This algorithm combines the effective 'network of functions' concept in neural networks with proven statistical learning techniques.