Physics-informed neural ODE models for soft sensing in chemical manufacturing
Innovative soft sensing solution using physics-informed neural ODE models to optimize chemical manufacturing processes, enhancing robustness and uncertainty quantification.
Overview
This innovative solution leverages physics-informed neural ordinary differential equations (NODEs) to model chemical process units within manufacturing lines. By integrating physics information such as mass flows, continuity equations, and thermodynamic constraints, these models provide a robust framework for predicting and optimizing process outcomes. This approach enhances the model's robustness to extrapolation and provides valuable uncertainty quantification, facilitating informed decision-making in process optimization.
Technical specifications
Key features:
- Utilizes neural ODEs to model chemical processes, incorporating essential physics for enhanced accuracy.
- Models are trained using data from process units, such as thermocouple measurements, to predict performance metrics.
- Enables uncertainty quantification, aiding in active experiment design and optimization.
- Applicable to various chemical manufacturing processes, providing insights into spatially and temporally varying systems.
Technology readiness level
This solution is at Technology Readiness Level 5, as it has undergone validation in a relevant environment with plans for further development and optimization stages.
About Lehigh University
Lehigh University is a private, mid-sized research university in Bethlehem, Pennsylvania focused on interdisciplinary, hands-on problem solving. A multi-campus footprint includes a research campus with co-located labs, instrumentation, and prototyping spaces for rapid iteration with industry. Corporate engagement staff streamline sponsored research, capstones, internships, and short courses, and an on-campus incubator links partners to startup pipelines. Proximity to New York City and Philadelphia enables collaboration with regional manufacturers, logistics leaders, and growth-stage firms. Faculty secure competitive federal funding from agencies such as the NSF and DOE, and a technology transfer office supports IP, licensing, and new venture formation.
delivered to your inbox.