Abstract:
At a previous meeting (June 1995) a report was given on anomalous behavior of the Rayleigh--Ritz method for calculating the normal mode frequencies of elastic solids (the method that forms the basis for most resonant ultrasound work), when those solids had the shapes of slightly perturbed parallelepipeds. These shapes are of interest to understand the effects of sample preparation errors on the overall elastic constant determination. In an effort to understand these effects, a systematic study of the various interrelated computational issues in RUS has been undertaken. Attempts are made to address the following issues: (1) whether the anomalous behavior of the Rayleigh-Ritz solutions is algorithmic or numerical; (2) whether these anomalies may result from problems with the inverse calculation (where the frequencies are inverted to obtain elastic constants) as well; and (3) what relationship exists between the RMS error in the frequencies and the confidence in the elastic constant determination. Results will be shown comparing analyses of actual data with and without the inclusion of a correction for nonparallelism/nonperpendicularity in the sample. [Work was supported in part by the Office of Naval Research.]