Presented By: Weina Meng, Stevens Institute of Technology
Description: Geopolymers have been considered as promising alternatives to cement-based materials. However, the challenge lies in formulating geopolymers from solid wastes, mainly due to the huge differences in the physical and chemical properties of these wastes. To address this challenge, this paper introduced a knowledge graph-guided artificial intelligence (AI)-based approach to design geopolymer with wastes. This innovative approach aims to maximize mechanical properties, minimize material cost, and reduce carbon footprint, while substantially accelerating the pace of material discovery. The proposed approach integrates knowledge graph, machine learning, and multi-objective optimization to design ultra-high performance geopolymer (UHPG). There are two important novelties: (1) The integration of knowledge graphs provides critical domain expertise to the machine learning model, enhancing its interpretability. (2) The mixing of multiple solid wastes is facilitated by considering the different physical and chemical properties of raw materials. The outcomes revealed the efficacy of the AI model in predicting geopolymer properties. Moreover, the knowledge graph offered insightful interpretations of these predictions. Further, the UHPG design was optimized via multi-objective optimization of mechanical properties, material cost, and carbon footprint.
Негізгі бет Knowledge-Guided Data-Driven Design of Ultra-High-Performance Concrete (UHPC)
Пікірлер