As a nanoporous material with high specific surface area and high porosity， aerogels exhibit irreplaceable application value in fields such as thermal insulation， adsorption， and catalysis. However， the traditional research and development mode based on experimental trial-and-error is limited by complex preparation variables and nonlinear coupling of properties， making it difficult to efficiently meet the demand for rapid iterative optimization and improvement of aerogels. In recent years， researchers have significantly enhanced the research and development efficiency of aerogels by constructing an association chain of “data-features-models-properties”. This review systematically summarizes the recent progress in the application of machine learning in aerogel research： first， it elaborates on the basic framework of machine learning-assisted aerogel research； subsequently， according to the three major algorithm categories （supervised learning， unsupervised learning， and reinforcement learning）， it sorts out their specific applications and typical cases in aerogel performance prediction， multi-property correlation analysis， aerogel performance grading， dataset quality improvement， multi-objective optimization， and dynamic process regulation， respectively； finally， it summarizes the core achievements of current research and prospects the future research directions combining multi-modal data fusion， interpretable physical modeling， cross-scale model optimization， automated experimental platforms， and interdisciplinary collaborative innovation. Machine learning can provide a new paradigm of data-driven research for aerogel studies.

Contents

1 Introduction

2 Basic framework of machine learning-assisted aerogel research

2.1 Data collection and preprocessing

2.2 Feature engineering

2.3 Model selection and training

2.4 Performance evaluation and optimization

3 Application of supervised learning in aerogel research

3.1 Specific surface area prediction

3.2 Thermal conductivity prediction

3.3 Mechanical properties prediction

3.4 Electromagnetic performance optimization

3.5 Other performance optimizations

4 Application of unsupervised learning in aerogel research

4.1 Revealing intrinsic multi-property correlations

4.2 Achieving aerogel performance grading

4.3 Enhancing dataset quality

5 Application of reinforcement learning in aerogel research

5.1 Systematic optimization of multi-objective performances

5.2 Dynamic process control throughout the entire workflow

6 Conclusion and outlook