A memosducer based biquaternionic signal processing for nonlinear ultrasonics imaging applied to nondestructive testing

Optimization of memristor based ultrasonic transducers for mesoscopic characterization of biomaterials has been presented during the last IEEE 2022 ISAF. In parallel, the development of quaternionic Clifford based transforms for improving deep learning algorithms within nondestructive testing (NDT) has been recently suggested.Practical implementatiion of this signal processing tool was tested on the time reversal based nonlinear elastic wave spectroscopy (TR-NEWS) experiments. The (2+1)D image data is suggested using quaternions H, detecting anomalous scattering positions in media with hysteresis, while in the (3+1)D image data analysis, an extension was necessary.The (3+1)D image processing was conducted using Clifford algebra ${{{\mathcal{A}}}_{3,1}}$ isomorphic to M<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>(H), where ultrasonic signals are represented in the biquaternionic space.The optimal weight function of paths associated with this biquaternionic signal processing was obtained by modifying the Echo State Network (ESN) method, and comparison with TR-NEWS experiments was conducted on complex samples (biomaterials or NDT samples) with intrinsic hysteretic nonlinearities.Stability of the weight function of the ultrasonic (US) wave path in (3+1)D is checked by the Machine Learning technique.