Detection of fastener failure in a thermal protection system panel

Abstract

This paper presents development methods and application results for a structural health monitoring (SHM) system for assessing the condition of mechanical fasteners in a test article representing a realistic portion of a thermal protection system. The test article is a carbon-carbon panel bolted through 15 brackets to a backing structure. Mechanical states considered include all bolts fastened to a nominal torque value, or one of the 15 bolts loosened. Four transducers on the backing structure provide actuation and sensing signals. Spectral functions are computed from all single and pair-wise signal combinations: (cross) power spectral densities, transfer functions, and coherence functions. Automated analysis of the spectral functions shows frequency intervals exist over which the function values are indicative of the mechanical state of the test article. These frequency intervals are used to provide features for the SHM classifier. Statistical pattern recognition methods select a subset of the features. The status of the test article is determined as "undamaged" or "bolt j is loose." The overall localization accuracy of the SHM system on test data is 99.1% with 99.7% probability of detecting a damaged condition at a 0.2% probability of a false alarm.