QUANTUM-INFORMED AI SURROGATE MODELING FOR NANO-BEAM VIBRATION ANALYSIS USING NONLOCAL STRAIN-GRADIENT THEORY

Abstract

 There is a need to address the computational modeling challenges of accounting for classical bending behavior, nanoscale size effects and modal quantities of the quantum harmonic oscillators while maintaining efficiency with respect to frequent parametric prediction in order to successfully run a nano-beam vibration analysis. In this work, an artificial-intelligence surrogate framework based on EulerBernoulli beam theory, nonlocal strain-gradient correction (in reduced order form), surface elasticity, and modal quantum oscillator relations is developed with the aid of quantum information. The physicsgenerated database contains 5,000 of these nanobeam cases that vary in the geometric, material, boundarycondition and temperature settings and are produced over nanoscale length-scale variations. Multi-output surrogate regression models for prediction of NSGT-corrected modal frequencies, corrective factors, zero point energy, zero point displacement and thermal RMS displacement are trained using log-scaled PI features. The best model performed the Log-Target test with R2 of 0.862, the Physical-Scale performed R2 of 0.762 and the average mean percentage error for the best model was 7.69%. Inside the physical domain sampled, the surrogate model is explicitly viewed as an interpolation and prediction technique that cannot compete with the governing beam, NSGT, or quantum oscillator but it does complement these models. Other verification checks, information used for generating the datasets, limitations and interpretations of errors in the proposed framework are provided for clarification to enhance the confidence of the proposed framework.