Integrating Near-Infrared Spectroscopy and High-Fidelity Simulation for Paint drying Optimization

Abstract To address deficient online sensing and control in paint drying, this study constructed an integrated process model combining online detection and fluid simulation. An online moisture content prediction model (MCP-Model) was established, combining Near-Infrared Spectroscopy (NIRS) with a Particle Swarm Optimized Support Vector Machine (PSO-SVM). Leveraging moisture data from the MCP-Model, the dynamic moisture diffusion coefficient (D(t)) was inversely calculated based on mass transfer theory. The derived D(t) was subsequently employed as a critical boundary condition to build a high-fidelity Computational Fluid Dynamics model (CFD-Model). A stage-based dynamic control strategy (SDCS) was then proposed. The MCP-Model (Rp2=0.9877) showed superior predictive performance, surpassing traditional models. The derived D(t) showed a dynamic pattern (low, then sharp increase), which, with CFD analysis, revealed a rate-limiting step transition from external convection to internal diffusion. The SDCS adopts high air velocity during the initial (constant-rate) phase to reduce external resistance, then switches to high temperature during the later (falling-rate) phase to intensify internal diffusion. By dynamically matching parameters to the rate-limiting step, the drying process was effectively optimized. The proposed integrated framework (NIRS-driven MCP-Model, D(t) inversion, and D(t)-driven CFD-Model) provides a new methodology for optimizing paint drying, offering theoretical and engineering reference for the coatings industry.