The Application of Maximum Entropy Signal Processing to Ultrasonic Surface Parameterisation

Abstract

The Maximum Entropy Method (MEM) has been employed successfully in previous work at Aberdeen for the deconvolution of ultrasonic pulses reflected from rough surfaces. The aim is to establish a new technique for surface parameterisation by obtaining and analysing the height (or amplitude) distribution functions of rough surfaces. Data obtained from an experimental system is inherently noisy: Linear signal processing techniques cannot produce a noise and artefact free deconvolution under these circumstances. The MEM, on the other hand, is more suitable, and the surface amplitude parameters can be easily derived from the computed amplitude distribution function. In practical cases good agreement with stylus profilometry is obtained. In the work presented here, a simple extension has been applied to the MEM to allow it to produce images containing both positive and negative data points with little compromise to the MEM’s frequency extrapolation property. This permits consideration of surfaces with layers of media with differing acoustic impedances. Simulations of a surface layer with differing acoustic properties have been carried out to gauge how accurately the MEM can reproduce the expected image peak positions and heights—the results are presented. To conclude, results of experimental tests of a single layered paint surface are presented, showing how the calculation of the acoustic parameters and thickness are significantly aided by the MEM.