Aeroelastic stability of piezoelectric Timoshenko pipes conveying fluid in yawed supersonic flow: A hybrid data-driven and multiphysics approach

This study presents a hybrid numerical and machine learning (ML) framework to analyze the vibration and stability of piezoelectric-coated fluid-conveying Timoshenko pipes subjected to yawed external supersonic flow in a multiphysics environment. The governing equations, derived and discretized via the Galerkin method, account for a laminar parabolic internal flow profile, supersonic damping, and coupled magneto-thermo-hygroscopic effects. Solving the resulting eigenvalue problem yields the system's dynamical response and stability maps, which are validated against established experimental and theoretical benchmarks. Subsequently, Decision Tree Regression and ε-Support Vector Regression (ε-SVR) models are trained to predict the pipe's vibrational frequencies and instability thresholds under varied operational parameters. Performance metrics, error distributions, and Taylor diagrams confirm the high predictive accuracy of the ML models. A comprehensive parametric study elucidates the influence of flow-related factors, including internal velocity profile, supersonic damping, yaw angle, and external aerodynamic pressure, on the dynamic characteristics and instability boundaries. Key findings reveal that considering a non-uniform internal flow velocity profile reduces the critical flow velocity and pipe stability. Among the ML models, the cubic ε-SVR demonstrates superior accuracy and reliability, establishing it as an efficient surrogate for rapid parametric analysis. The proposed hybrid framework proves to be a reliable and effective computational tool for stability optimization and parameter design of complex fluid-conveying structures under supersonic flow.