Damage Detection Using Time-frequency Decay-rate based Features

Some automotive components, for instance a clutch housing, need to fit the assembly (dimensional constraint), and must be perfectly weight balanced for proper functioning (functional fidelity), which could be achieved by removing material from the body (relaxed geometry). While such mission critical components require 100% inspection, current non-destructive methods are often inadequate, trying to trade-off cycle time with inspection thoroughness. In this paper, a non-destructive global structural damage detection method has been demonstrated at its infancy, to be fast, accurate, and immune to inconsistent resonant frequencies which is an outcome of relaxed geometry. This method requires a traditional wide band exogenous acoustic excitation of the component being evaluated, followed by a joint time-frequency characterization of the response such that it is immune to dominant modal frequencies. A quadratic discriminant analysis on the time-frequency features successfully detected all damaged samples and falsely identified 1 among the 6 healthy samples available for testing, as damaged.