Physics Closure Matters for Machine Olfaction: A Maxwell--Stefan Graph Solver for Identifiable Dynamic Gas Unmixing

2026-07-20

AI Summary

Gas demixing, as an underdetermined inverse problem, is hindered by low-dimensional, delayed, and coupled sensor responses, and traditional methods struggle to accurately infer constituent concentrations due to the lack of physically closed modeling. This work introduces, for the first time in machine olfaction, a multiphysics inverse framework that integrates the Maxwell–Stefan multicomponent transport PDEs, competitive adsorption ODEs, and nonlinear sensing transduction ODEs. To solve this formulation, we propose UnMixNet, an end-to-end differentiable graph neural solver that discretizes cross-diffusion dynamics via a graph structure, enabling local, flux-conserving, and differentiable inference while learning transferable physical fingerprints. Experiments demonstrate that UnMixNet significantly improves performance on SmellNet for single-odor identification and demixing of both seen and unseen gas mixtures, and achieves concentration estimates on the UCI dynamic dataset that closely match ground-truth values, confirming its physical consistency and generalization capability.

Abstract

Machine olfaction for gas unmixing is an underconstrained inverse problem in which gas compositions must be inferred from low-dimensional, delayed, and entangled sensor responses produced by interacting chemical transport, surface adsorption, and sensor transduction. One of the key obstacles is physics closure misspecification, where a neural network is designed to fit sensor traces rather than infer a physically closed olfactory process. In this work, we formulate gas unmixing as a multi-physics-constrained inverse problem governed by Maxwell--Stefan multicomponent transport PDEs, competitive adsorption ODEs, and nonlinear sensor transduction ODEs. Directly solving such a high-dimensional coupled system is computationally expensive and often numerically unstable. To this end, we propose UnMixNet, a physics-closed graph neural solver that embeds this multi-physics forward process into end-to-end gas unmixing. UnMixNet discretizes Maxwell--Stefan cross-diffusion on spatial graphs and formulates the multicomponent flux on each edge. This design enables local, differentiable, and flux-conservative inference for multicomponent cross-diffusion. Evaluations on SmellNet show improved single-odor recognition, seen-mixture unmixing, and unseen-mixture generalization. In addition, an external validation on UCI Dynamic Gas Mixtures shows that the inferred concentration process agrees with ground truth concentration set points under dynamic transitions. Process-consistency diagnostics further show that the proposed model learns transferable dynamic physical fingerprints that better satisfies transport, conservation, adsorption, and readout closure.

Problem

• machine olfaction
• gas unmixing
• physics closure
• inverse problem
• Maxwell-Stefan diffusion

Innovation

• physics-closed modeling
• Maxwell–Stefan diffusion
• graph neural solver
• multicomponent gas unmixing
• multi-physics-constrained inverse problem