AI-Powered Vision-Based Drying Control for Sustainable Building Materials

| AI-Powered Vision-Based Drying Control for Sustainable Building Materials Web Published: 11/12/2026 An autonomous manufacturing system that uses computer vision, artificial intelligence, and adaptive process control to optimize drying of biomass-based building materials. The technology predicts cracking before visible damage occurs and automatically adjusts drying conditions, reducing energy consumption while improving production quality. Biomass-based building materials, including engineered wood, wood fiber insulation, and biocomposite panels, offer sustainable alternatives to conventional construction materials. However, manufacturing these products often requires drying cycles lasting 48–72 hours, during which uneven moisture distribution can cause internal stress, cracking, and material rejection. Conventional drying processes typically rely on fixed temperature, humidity, and airflow settings with periodic manual inspection, limiting their ability to respond to changing material conditions, prevent defects, or avoid unnecessary energy consumption. University at Buffalo technology, Warden, is a closed-loop, robotic drying control system that integrates an overhead RGB camera, an NVIDIA Jetson Xavier NX edge-computing platform, and automated controls for temperature, airflow, and humidity. The camera captures images of biomass materials every 30 seconds, allowing an AI-based perception model to continuously assess drying progress and identify regions at risk of cracking. The system employs a dual-task deep learning architecture built on a ConvNeXt-Tiny encoder and Feature Pyramid Network to simultaneously generate spatial moisture distribution maps and predict crack-prone regions. Physics-informed cross-task coupling incorporates moisture gradients and material dryness into the crack prediction model, enabling the system to identify conditions associated with differential shrinkage and mechanical failure before visible defects develop. A finite-state controller uses these predictions to automatically regulate the drying environment. When elevated crack risk is detected, the system reduces temperature and airflow while increasing humidity to moderate drying intensity. Once the material reaches its target dryness, the controller shuts off active drying equipment to conserve energy. Source: Lotushaven, Wikimedia, CC0 1.0 Predictive identification of crack-prone regions before visible damage occurs 38.7% reduction in drying energy consumption Improved defect-free production success rate from 64.7% to 88.2% Simultaneous monitoring of spatial moisture distribution and crack risk Automatic adjustment of temperature, airflow, and humidity; continuous autonomous operation without human intervention Non-contact RGB imaging with on-device AI processing that does not require cloud connectivity Manufacturing of biomass-based construction materials Engineered wood and biocomposite panel production Wood fiber insulation manufacturing Automated industrial drying and quality control Potential adaptation to other manufacturing processes involving moisture removal, curing, or defect prevention Patent pending. Available for licensing or collaboration. Patent Information: | App Type | Country | Serial No. | Patent No. | Patent Status | File Date | Issued Date | Expire Date | |