Non-destructive methods play a fundamental role in analyzing theintegrity of civil and mechanical components. This work presents an integrative approach that can be applied in structural assessments, involving the study of ultrasonic waves guided with machine learning techniques,together with alternative methods to expand the data set necessary fortraining prediction models. A methodology for estimating stress in platesis proposed, incorporating synthetic data generated by linear interpolationfrom experimental results. In parallel, using the Semi-Analytical Finite Element Method (SAFE), we examined wave propagation in cylindrical guides,comparing results with numerical simulations and subsequently, applyingthe SAFE method to generate dispersion curves that serve as data to feeda neural network for predicting nominal pipeline thickness. This study contributes to the efficiency and precision of methodologies applied in structuralintegrity, encompassing experimental and computationally simulated dataapproaches.