Investigation into the use of ultrasonics for surface texture evaluation
This thesis presents a preliminary investigation into the determination of enginerring important surface texture (of around a few microns) using ultrasonic pulse-echo signals. Current texture measurements and parameterisation techniques are reviewed, identifying their deficiencies, the requirements of a texture measurement device and the potential benefits of an ultrasonic technique. Plane compressional wave, high resolution, commercial NDT probes are considered, with the received signal being modelled as a first type Fredholm equation. The characteristics of these probes are considered in detail and realistic numerical simulations are presented, showing the variation in sound field and receiver voltage with the geometry and electrical condition of the test configuration. Inversion of the signal model is considered as a direct and useful route to characterising the surface. Techniques for this inversion are reviewed indicating their salient features and showing their inter-relationships. Several linear inversion techniques are considered in detail, using numerical simulations. These demonstrate their characteristics and the predictability of their performance. A simple 'interpolation' technique is shown to yield excellent results when applied under appropriate conditions. The Maximum Entropy algorithm is shown to yield significant improvements over linear techniques, in terms of reduced noise and filter artefacts and also possibly in improved resolution. Various characteristics of the algorithm performance are considered. Finally, preliminary results using real pulse echo signals are considered, with Maximum Entropy again yielding considerably superior results. While the achievable resolution is still somewhat below that desired, these results clearly demonstrate that the use of ultrasonics for surface texture evaluation is highly feasible and that the continued investigation of these, and other, ultrasonic techniques is fully justified.