NSF Award
2412472
This project aims to create an AI-powered framework that helps designers automatically evaluate the environmental impact of their designs and optimize them for sustainability. This research overcomes several key challenges in sustainable design practices, including the scarcity of lifecycle inventory (LCI) data essential for lifecycle assessment (LCA), the lack of accurate tools for including product lifespan uncertainties in LCA prediction, and the absence of an inclusive decision-making framework that integrates environmental and technical considerations. This project develops advanced AI algorithms such as graph attention networks (GATs) and reinforcement learning (RL) to advance design science and national manufacturing prosperity. It augments designers’ abilities to quickly update design alternatives that align with sustainability practices and provides manufacturers with faster and more accurate tools to assess the environmental impacts of their products, aiding in the development of software packages to measure corporate social responsibility. This framework supports sustainable development within engineering design and extends its applications to decision-making throughout the product lifespan and supply chain. Furthermore, this work aligns with national interests by promoting scientific progress, supporting sustainable manufacturing, and advancing STEM education, particularly through undergraduate research experiences and K-12 activities.<br/><br/>The objective of this project is to establish a cutting-edge AI-driven framework for sustainable design that cohesively integrates advanced eco-design algorithms with product LCA and sustainability evaluation, setting a new standard in optimizing design alternatives. To achieve this objective, this research will implement several key innovations: (i) develop novel GAT algorithms to extract LCI data by analyzing similarities with existing designs that possess known LCI data, thereby addressing the issue of data scarcity essential for LCA; (ii) create advanced algorithms inspired by Markov process simulations to model the heterogeneity and stochasticity of product lifespans, thereby significantly improving the accuracy of LCA predictions; and (iii) establish a decision-making framework based on a novel non-lexicographic multi-criteria RL algorithm to facilitate the simultaneous evaluation of environmental and technical considerations and produce Pareto-optimal design alternatives. This integrative approach will combine the data extraction capabilities of GAT algorithms and the LCA prediction accuracy provided by Markov process simulations to transform sustainable design practices. This researched framework will be rigorously evaluated through its application in designing consumer electronics, including smartphones, tablets, and laptops. This collaborative project between the University of Florida and Florida International University combines expertise in engineering design, lifecycle assessment, circular economy, optimization, and artificial intelligence. Educational initiatives will include the organization of industry and academic webinars and workshops, provision of timely training for students, and promotion of STEM education with an emphasis on underrepresented groups.<br/><br/>This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
