Intelligent backpropagated predictive neural networks approach for acoustic wave-driven gas bubble dynamics in non-Newtonian fluid

Abstract

Ultrasound imaging, often known as sonography, employs high-frequency sound waves to generate images of inside structures of human body. Its non-aggressive nature and real-time capabilities make it commonly used in medical diagnostics. Ultrasound waves are projected into the body and their echoes are recorded to produce intricate images of organs, tissues, and fetuses in utero, among other things. An essential aspect of enhancing image quality and safety involves the examination of how sound waves interact with biological tissues, including the phenomena of propagation, reflection, and absorption of ultrasound. It is commonly employed to monitor the well-being and growth of the fetus throughout pregnancy. Various organs, such as liver, kidneys, heart, and blood arteries, are also examined to detect abnormalities, tumors, and other disorders. This work investigates the behavior of gas bubbles with a spherical shape in non-Newtonian fluid when subjected to an external sonic field. Artificial intelligence has greatly impacted fluid dynamics by enhancing mesh efficiency, reducing manual intervention, offering dependable predictions, aiding in data analytics, and allowing for machine automation. This research investigates the behavior of bubbles in the flow of a tangent hyperbolic fluid model (THFM) through the application of artificial intelligence methods. The system employs Nonlinear Autoregressive with Exogenous inputs (NARX) networks trained with the Levenberg–Marquardt technique (LMT), known as NARX. The NARX-LMT model was applied to data produced using the Adams numerical approach for the THFM by systematically adjusting parameters such as Weber number, Weissenberg number, pressure affecting velocity, and bubble radius. The effectiveness of projected THFM is demonstrated comprehensively through mean square error generated iterative learning curves, error histogram plots, analysis of adaptive control factors, regression, and time series response metrics for different versions of nonlinear differential equations of THFM based on bubble dynamics.