V. Plevris, A.T. Al-Sayegh, J. Mir and A. Ahmad, “Nondestructive Evaluation of Hybrid Concrete Properties Using Image Processing and Machine Learning”, Structures, 79, Article ID 109423 (DOI: 10.1016/j.istruc.2025.109423), 2025.

Abstract:
Advancements in informatics, such as image processing (IP) and machine learning (ML), are increasingly being utilized to evaluate the mechanical properties of reinforced concrete structures. This study focuses on hybrid concrete (HC), which incorporates cement replacement materials (CRM) like fly ash and silica fume to enhance its mechanical performance while promoting sustainability. A novel methodology combining IP with supervised ML models—Support vector machine (SVM), boosted ensemble regression (BRE), and Gaussian process regression (GPR)—was developed to predict the compressive and tensile strengths of HC. A comprehensive dataset was created using 162 cylindrical specimens prepared with various mix ratios, CRM replacement levels, and curing durations. High-resolution images of both horizontal and vertical cuts of the specimens were analyzed, and statistical features were extracted to train the ML models. The results demonstrated the models’ high accuracy in predicting mechanical properties, with GPR emerging as the most reliable method. The findings confirm the effectiveness of integrating IP with ML as a nondestructive testing approach for concrete evaluation, offering a fast, cost-effective, and environmentally friendly alternative to traditional methods. This study underscores the potential of combining advanced computational techniques with sustainable materials to innovate in concrete technology.