Mix design and performance prediction of EPS lightweight structural concrete based on orthogonal experimentation

This study explores mix proportion design and mechanical property prediction of EPS lightweight structural concrete using orthogonal experimentation and machine learning models. The research systematically analyzed the effects of EPS content, water-to-binder ratio, and POM fiber content on compressive strength, splitting tensile strength, thermal conductivity, and frost resistance. Key findings reveal that EPS content significantly enhances thermal insulation and frost resistance but reduces mechanical strength. POM fibers were shown to improve tensile strength and frost resistance by limiting crack propagation. A novel dataset was established and utilized in performance prediction using XGBoost, optimized with Seagull Optimization Algorithm (SOA), Whale Optimization Algorithm (WOA), and Particle Swarm Optimization (PSO). Among these, SOA-XGBoost achieved the highest prediction accuracy and stability. The optimal mix proportion, combining 35% EPS, 0.21 water-to-binder ratio, and 0.65% POM fiber content, was identified, providing an effective balance between mechanical and thermal performance. The proposed framework offers valuable insights and methodologies for optimizing lightweight concrete and serves as a reference for other composite materials in engineering applications.