Researchers have developed a new framework for real-time monitoring of inaccessible locations and unmeasurable parameters. They call it neural operator-based virtual sensing. The approach uses machine learning to recover interior fields from sparse boundary measurements. This is achieved through a multi-input, multi-output neural operator that fuses heterogeneous inputs and decodes coupled fields. The team tested their method on three engineering-grade evaluations and three independent real-world datasets. Their results show significant improvements over classical virtual-sensing baselines, with relative reconstruction error below 5% in most cases. This technology has the potential to enable real-time monitoring of safety-critical interior states in various systems. --- Why it matters: This matters because many complex systems, such as power plants and industrial processes, require real-time monitoring of internal states that are difficult or impossible to measure directly. The neural operator-based virtual sensing framework could provide a practical solution for these challenges, enabling engineers to make more accurate predictions and decisions. Source: https://arxiv.org/abs/2412.00107