Document Type

Article

Publication Title

Energy and AI

Abstract

Thermoacoustic Stirling-like cycle engines offer a compelling solution for converting residual heat into usable energy through the generation of acoustic waves, providing a mechanically simple and environmentally friendly alternative. While recent developments have demonstrated the feasibility of using artificial neural networks for performance prediction, limited research has addressed the effect of training data volume or the experimental validation of these models. This study investigates the use of artificial neural networks to predict the acoustic power delivered to the power extraction branch in thermoacoustic engines operating at low drive ratios, using data generated through DeltaEC (Design Environment for Low-Amplitude Thermoacoustic Energy Conversion) simulations. Two training approaches were implemented: Python programming and SPSS (Statistical Package for the Social Sciences); and three datasets of increasing size were used to evaluate the impact of data volume on prediction performance. Seven design variables were analyzed, of which three emerged as the most critical. The proposed methodology enables fast and reliable prediction of thermoacoustic performance, reducing reliance on computationally expensive simulations and supporting the integration of artificial intelligence techniques into energy system design.

First Page

100837

DOI

10.1016/j.egyai.2026.100837

Publication Date

9-2026

Language

eng

Rights

open access

Download

Share

COinS