Properties Prediction of Composites Based on Machine Learning Models: A Focus on Statistical Index Approaches
Barshan Dev,Md Ashikur Rahman,Md. Jahidul Islam,Md Zillur Rahman,Deju Zhu
DOI: https://doi.org/10.1016/j.mtcomm.2023.107659
IF: 3.8
2024-01-01
Materials Today Communications
Abstract:Composites have a wide range of applications across various industries due to their high strength-to-weight ratio, corrosion resistance, durability, versatility, and lightweight structures. However, manufacturing reinforced composites and the various tests they undergo for their appropriate applications are extensive and expensive. Because of this, many researchers have employed the machine learning (ML) technique to evaluate the significance of the process parameters and predict the properties for effective composite design and their widespread applications. Therefore, this study critically reviewed and compared the different ML models applied to predict the mechanical, thermal, tribological, acoustic, and electrical properties of different reinforced composites. ML models, their appropriate methods, database size and source, training and testing data, input and output parameters, and statistical index are also summarized. In addition, the performance evaluation of ML models and statistical indexes of different property predictions is critically analyzed based on several models' training and testing scores, which may help select appropriate ML models to predict reinforced composite properties. This review study can provide insights into applying ML techniques to reinforced composites to develop novel, innovative composites.