Machine learning-assisted design of lightweight refractory high-entropy alloys: A comprehensive review

Lightweight refractory high-entropy alloys (LRHEAs) represent an emerging class of structural materials that integrate low density with exceptional strength and outstanding high-temperature stability, positioning them as promising candidates for aerospace and advanced industrial applications. Nevertheless, the design of LRHEAs is challenged by their vast compositional space, complex multi-objective performance trade-offs, and the inefficiency of conventional trial-and-error experimental approaches. In recent years, machine learning (ML) has emerged as a transformative tool in this domain, offering the capacity to analyze high-dimensional datasets and uncover hidden correlations between composition, processing, microstructure, and properties. This review systematically examines both conventional design strategies—including empirical parameters, phase diagram calculations, and first-principles simulations—and the emerging ML-aided design framework, with a focus on bridging traditional knowledge and data-driven methodologies. We critically survey recent advances in ML applications across three key areas: compositional optimization, mechanistic interpretation, and atomic-scale simulation. Target-driven ML models facilitate efficient navigation of the alloy design space, while interpretable algorithms integrated with atomic simulations provide fundamental insights into strengthening and toughening mechanisms. The review concludes by summarizing current achievements and identifying persistent challenges related to data scarcity, model transferability, and physical interpretability. Looking forward, we envision that a deeper integration of high-throughput experiments, multi-scale simulations, and artificial intelligence will establish a robust, systematic, and accelerated design paradigm for next-generation LRHEAs. • Systematic overview of conventional and machine learning–assisted design strategies for LRHEAs. • Critical assessment of ML-enabled advances in composition optimization, mechanistic insights, and atomic-scale simulations for LRHEAs. • Perspectives on integrated high-throughput, multiscale, and AI-driven design of LRHEAs.