The Wigner-Ville Distribution and its smoothed versions, i.e., Choi-Williams and gaussian kernels, are evaluated for propulsion systeni diagnostics. The approach is intended for off-line kernel design by using the ambiguity domain to select the appropriate gaussian kernel. The features produced by the Wigner-Ville Distribution and its smoothed versions correlate remarkably well with documented failure indications. The selection of the kernel on the other hand is very subjective for our unstructured data.