A knowledge graph-augmented retrieval framework using Toxics Release Inventory source reduction records to support pollution prevention decisions
Year: 2026
Type: software
Abstract: Chemical industry stakeholders seek for pollution prevention (P2) strategies while safeguarding the technoeconomic feasibility of such process systems engineering endeavours. This requires case-by-case design, modeling, testing, and cost optimization of many possibilities for new and existing facilities. This study focuses on creating a graph-based learning from-data framework utilizing the source reduction activities reported by industrial facilities to the U.S. Toxics Release Inventory (TRI) program from 2014 to 2024. For a certain chemical and associated industrial and commercial condition of use (CoU), it provides a prioritized list of P2 initiatives that facilities have previously executed, achieving a cost-effective minimum reduction of 25%. The framework employs P2 alternative identification through surrogate analysis that integrates chemical structure with industrial sector and CoU information. Specifying the CoU made the recommendations more informative, giving engineers a shortlist specific to their processes, so feasibility studies can start from options that comparable facilities have already proven. In a first case study, the methodology identified supporting evidence for eleven chemicals of concern, six via surrogates or alternative uses. In a second case study, six chemicals that are not reported to TRI were checked against P2 activities documented in peer-reviewed studies; for four of the six, the documented activity was among the framework's ten highest-ranked suggestions. The framework transforms established proven industrial practices into a foundation for new chemical plant design and retrofits, focusing modeling and testing on technically and economically viable alternatives.
Language: English
Subfield: Environmental Chemistry
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7235656730
